研究者データベース

坂地 泰紀(サカジ ヒロキ)
情報科学研究院 情報理工学部門 複合情報工学分野
准教授

基本情報

所属

  • 情報科学研究院 情報理工学部門 複合情報工学分野

職名

  • 准教授

学位

  • 博士(工学)(豊橋技術科学大学)

ホームページURL

科研費研究者番号

  • 70722809

ORCID ID

J-Global ID

研究キーワード

  • 情報抽出   テキストマイニング   自然言語処理   

研究分野

  • 情報通信 / 知能情報学 / 自然言語処理

職歴

  • 2023年11月 - 現在 北海道大学 大学院情報科学研究院 准教授
  • 2018年04月 - 2023年10月 東京大学 大学院工学系研究科システム創成学専攻 特任講師
  • 2017年04月 - 2018年03月 東京大学 大学院工学系研究科システム創成学専攻 助教
  • 2013年09月 - 2017年03月 成蹊大学 理工学部 助教

学歴

  • 2009年04月 - 2012年03月   豊橋技術科学大学大学院   工学研究科   電子・情報工学専攻
  • 2007年04月 - 2009年03月   豊橋技術科学大学大学院   工学研究科   知識情報工学専攻
  • 2003年04月 - 2007年03月   豊橋技術科学大学   工学部   知識情報工学科

所属学協会

  • IEEE   人工知能学会   言語処理学会   電子情報通信学会   情報処理学会   ACL   

研究活動情報

論文

  • Hajime Sasaki, Motomasa Fujii, Hiroki Sakaji, Shigeru Masuyama
    International Journal of Information Management Data Insights 4 1 100217 - 100217 2024年04月 [査読有り][通常論文]
  • Takeshi Sakumoto, Teruaki Hayashi, Hiroki Sakaji, Hirofumi Nonaka
    IEEE Access 1 - 1 2024年03月11日 [査読有り][通常論文]
  • Hiroki Sakaji, Noriyasu Kaneda
    2023 IEEE International Conference on Big Data (BigData) 5674 - 5683 2023年12月15日 [査読有り][通常論文]
  • Masahiro Suzuki, Masanori Hirano, Hiroki Sakaji
    2023 IEEE International Conference on Big Data (BigData) 2023年12月15日 [査読有り][通常論文]
  • Meiyun Wang, Hiroki Sakaji, Hiroaki Higashitani, Mitsuhiro Iwadare, Kiyoshi Izumi
    World Patent Information 75 2023年12月 [査読有り]
     
    Determining a technology's potential applications is crucial in assessing its level of innovation and evaluating its commercial viability. However, patent documents offer limited insights into a technology's full potential. As a solution, this study suggests an approach to explore a technology's applicability beyond what is explicitly stated in patents. The approach employs causal extraction to extract sentences expressing technologies and their applications from patents, followed by deep learning-based similarity analysis to compare the similarity of these sentences. Experimental results show its effectiveness in extracting sentences about technologies and applications and its superiority in terms of F1 score compared to benchmark models. This study enables cross-domain comparisons of technologies and applications, identifies multiple prospective applications for a given technology, and offers new opportunities for patent value analysis and intellectual property management in the industry. A cross-domain application network of the proposed method demonstrates how to find all potential cross-domain connections of a given data and we provide open access to the code.
  • Hiroki Sakaji, Kiyoshi Izumi
    New Generation Computing 41 4 839 - 857 2023年11月 [査読有り][通常論文]
  • Rei Taguchi, Hiroki Sakaji, Kiyoshi Izumi, Yuri Murayama
    New Generation Computing 2023年09月11日 [査読有り]
     
    Abstract This study demonstrates whether financial text is useful for the tactical asset allocation method using stocks. This can be achieved using natural language processing to create polarity indexes in financial news. We perform clustering of the created polarity indexes using the change point detection algorithm. In addition, we construct a stock portfolio and rebalanced it at each change point using an optimization algorithm. Consequently, the proposed asset allocation method outperforms the comparative approach. This result suggests that the polarity index is useful for constructing the equity asset allocation method.
  • Masanori Hirano, Masahiro Suzuki, Hiroki Sakaji
    Advances in Networked-based Information Systems 442 - 454 2023年08月24日 [査読有り]
  • Masahiro Suzuki, Hiroki Sakaji, Masanori Hirano, Kiyoshi Izumi
    Information Processing & Management 60 2 103194 - 103194 2023年03月 [査読有り][通常論文]
  • Hiroki Sakaji, Masahiro Suzuki, Kiyoshi Izumi, Hiroyuki Mitsugi
    2022 IEEE International Conference on Big Data (Big Data) 2022年12月17日 [査読有り]
  • Toshihiko Nanba, Kazuhito Ogawa, Naoki Watanabe, Teruaki Hayashi, Hiroki Sakaji
    2022 IEEE International Conference on Big Data (Big Data) 2022年12月17日 [査読有り]
  • Motomasa Fujii, Hiroki Sakaji, Shigeru Masuyama, Hajime Sasaki
    International Journal of Information Management Data Insights 2 2 100096 - 100096 2022年11月 [査読有り][通常論文]
  • Teruaki Hayashi, Hiroyasu Matsushima, Hiroki Sakaji, Yoshiaki Fukami, Takumi Shimizu
    Procedia Computer Science 207 1695 - 1704 2022年09月 [査読有り][通常論文]
  • Kei Nakagawa, Shingo Sashida, Hiroki Sakaji
    10th International Congress on Advanced Applied Informatics (SCAI2022) 287 - 292 2022年07月 [査読有り][通常論文]
  • Masahiro Suzuki, Hiroki Sakaji, Kiyoshi Izumi, Yasushi Ishikawa
    Frontiers in Artificial Intelligence 5 2022年06月07日 [査読有り][通常論文]
     
    This article proposes a methodology to forecast the movements of analysts' estimated net income and stock prices using analyst profiles. Our methodology is based on applying natural language processing and neural networks in the context of analyst reports. First, we apply the proposed method to extract opinion sentences from the analyst report while classifying the remaining parts as non-opinion sentences. Then, we employ the proposed method to forecast the movements of analysts' estimated net income and stock price by inputting the opinion and non-opinion sentences into separate neural networks. In addition to analyst reports, we input analyst profiles to the networks. As analyst profiles, we used the name of an analyst, the securities company to which the analyst belongs, the sector which the analyst covers, and the analyst ranking. Consequently, we obtain an indication that the analyst profile effectively improves the model forecasts. However, classifying analyst reports into opinion and non-opinion sentences is insignificant for the forecasts.
  • Attempt to Develop An Approach Based on BERT for Task of NTCIR-16 QA Lab-Poliinfo-3 Budget Argument Mining
    Akio Kobayashi, Hiroki Sakaji
    the 16th NTCIR Conference on Evaluation of Information Access Technologies 199 - 200 2022年06月 [査読有り][通常論文]
  • Rei Taguchi, Hikaru Watanabe, Hiroki Sakaji, Kiyoshi Izumi, Kenji Hiramatsu
    Frontiers in Artificial Intelligence 5 2022年05月18日 [査読有り][通常論文]
     
    This study demonstrates whether analysts' sentiments toward individual stocks are useful for stock investment strategies. This is achieved by using natural language processing to create a polarity index from textual information in analyst reports. In this study, we performed time series forecasting for the created polarity index using deep learning, and clustered the forecasted values by volatility using a regime switching model. In addition, we constructed a portfolio from stock data and rebalanced it at each change point of the regime. Consequently, the investment strategy proposed in this study outperforms the benchmark portfolio in terms of returns. This suggests that the polarity index is useful for constructing stock investment strategies.
  • Masanori Hirano, Hiroki Sakaji, Kiyoshi Izumi
    2022 IEEE Symposium on Computational Intelligence for Financial Engineering and Economics (CIFEr) 2022年05月 [査読有り][通常論文]
  • Implementation of Actual Data for Artificial Market Simulation
    Masanori Hirano, Kiyoshi Izumi, Hiroki Sakaji
    The 21st International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2022) 2022年05月 [査読有り][通常論文]
  • Graph Representation Learning of Banking TransactionNetwork with Edge Weight-Enhanced Attention and Textual Information
    Naoto Minakawa, Kiyoshi Izumi, Hiroki Sakaji, Hitomi Sano
    The 2nd Workshop on Financial Technology on the Web (FinWeb) 2022年05月 [査読有り][通常論文]
  • Masanori Hirano, Kiyoshi Izumi, Hiroki Sakaji
    New Generation Computing 40 1 7 - 24 2022年04月 [査読有り][通常論文]
  • Takehiro Takayanagi, Hiroki Sakaji, Kiyoshi Izumi
    IEEE International Conference on Big Data 2377 - 2382 2022年 [査読有り]
  • Rei Taguchi, Hiroki Sakaji, Kiyoshi Izumi
    IEEE International Conference on Big Data 2373 - 2376 2022年 [査読有り]
  • Naoto Minakawa, Kiyoshi Izumi, Hiroki Sakaji
    IEEE International Conference on Big Data 2327 - 2332 2022年 [査読有り]
     
    The gravity models has been studied to analyze interaction between two objects such as trade amount between a pair of countries, human migration between a pair of countries and traffic flow between two cities. Particularly in the international trade, predicting trade amount is instrumental to industry and government in business decision making and determining economic policies. Whereas the gravity models well captures such interaction between objects, the model simplifies the interaction to extract essential relationships or needs handcrafted features to drive the models. Recent studies indicate the connection between graph neural networks (GNNs) and the gravity models in international trade. However, to our best knowledge, hardly any previous studies in the this domain directly predicts trade amount by GNNs. We propose GGAE (Gravity-informed Graph Auto-encoder) and its surrogate model, which is inspired by the gravity model, showing trade amount prediction by the gravity model can be formulated as an edge weight prediction problem in GNNs and solved by GGAE and its surrogate model. Furthermore, we conducted experiments to indicate GGAE with GNNs can improve trade amount prediction compared to the traditional gravity model by considering complex relationships.
  • Teruaki Hayashi, Takumi Shimizu, Yoshiaki Fukami, Hiroki Sakaji, Hiroyasu Matsushima
    2021 IEEE International Conference on Big Data (Big Data) 2021年12月15日 [査読有り][通常論文]
  • Hiroki Sakaji, Teruaki Hayashi, Yoshiaki Fukami, Takumi Shimizu, Hiroyasu Matsushima, Kiyoshi Izumi
    2021 IEEE International Conference on Big Data (Big Data) 2021年12月15日 [査読有り][通常論文]
  • Rei Taguchi, Hikaru Watanabe, Masanori Hirano, Masahiro Suzuki, Hiroki Sakaji, Kiyoshi Izumi, Kenji Hiramatsu
    2021 IEEE International Conference on Big Data (Big Data) 2021年12月15日 [査読有り][通常論文]
  • Teruaki Hayashi, Hiroki Sakaji, Hiroyasu Matsushima, Yoshiaki Fukami, Takumi Shimizu, Yukio Ohsawa
    The Review of Socionetwork Strategies 15 2 521 - 534 2021年11月 [査読有り][通常論文]
     
    Abstract In recent years, rather than enclosing data within a single organization, exchanging and combining data from different domains has become an emerging practice. Many studies have discussed the economic and utility value of data and data exchange, but the characteristics of data that contribute to problem-solving through data combination have not been fully understood. In big data and interdisciplinary data combinations, large-scale data with many variables are expected to be used, and value is expected to be created by combining data as much as possible. In this study, we conducted three experiments to investigate the characteristics of data, focusing on the relationships between data combinations and variables in each dataset, using empirical data shared by the local government. The results indicate that even datasets that have a few variables are frequently used to propose solutions for problem-solving. Moreover, we found that even if the datasets in the solution do not have common variables, there are some well-established solutions to these problems. The findings of this study shed light on the mechanisms behind data combination for solving problems involving multiple datasets and variables.
  • Kiyoshi Izumi, Hitomi Sano, Hiroki Sakaji
    Proceedings of the 3rd Financial Narrative Processing Workshop, FNP 2021 19 - 25 2021年 [査読有り]
     
    This paper proposes a method that uses causal information extracted from textual data to predict economic indicators. The method automatically extracts causal information included in each sentence using machine learning and natural language processing methods. The extracted cause-effect expressions are stored in an economic causality database. Then, the method can generate causal chains from the given text using the word similarity between a result expression and a cause expression in the database. The causal chains are used to predict how the numerical values of economic indicators will change in the future due to spillover effects from the given text.
  • Iwao Maeda, Hiroyasu Matsushima, Hiroki Sakaji, Kiyoshi Izumi, David deGraw, Atsuo Kato, Michiharu Kitano
    International Journal of Smart Computing and Artificial Intelligence 5 1 1 - 18 2021年 [査読有り][招待有り]
  • Masanori Hirano, Hiroyasu Matsushima, Kiyoshi Izumi, Hiroki Sakaji
    Advances in Intelligent Systems and Computing 157 - 165 2021年 [査読有り]
  • Masanori Hirano, Hiroyasu Matsushima, Kiyoshi Izumi, Hiroki Sakaji
    PRIMA 2020: Principles and Practice of Multi-Agent Systems 3 - 18 2021年 [査読有り]
  • Hiroki Sakaji, Teruaki Hayashi, Kiyoshi Izumi, Yukio Ohsawa
    2020 IEEE International Conference on Big Data (Big Data) 2020年12月10日 [査読有り][通常論文]
  • Iwao Maeda, David deGraw, Michiharu Kitano, Hiroyasu Matsushima, Kiyoshi Izumi, Hiroki Sakaji, Atsuo Kato
    JOURNAL OF RISK AND FINANCIAL MANAGEMENT 13 11 2020年11月 [査読有り]
     
    While exchanges and regulators are able to observe and analyze the individual behavior of financial market participants through access to labeled data, this information is not accessible by other market participants nor by the general public. A key question, then, is whether it is possible to model individual market participants' behaviors through observation of publicly available unlabeled market data alone. Several methods have been suggested in the literature using classification methods based on summary trading statistics, as well as using inverse reinforcement learning methods to infer the reward function underlying trader behavior. Our primary contribution is to propose an alternative neural network based multi-modal imitation learning model which performs latent segmentation of stock trading strategies. As a result that the segmentation in the latent space is optimized according to individual reward functions underlying the order submission behaviors across each segment, our results provide interpretable classifications and accurate predictions that outperform other methods in major classification indicators as verified on historical orderbook data from January 2018 to August 2019 obtained from the Tokyo Stock Exchange. By further analyzing the behavior of various trader segments, we confirmed that our proposed segments behaves in line with real-market investor sentiments.
  • Yoshiyuki Suimon, Kiyoshi Izumi, Hiroki Sakaji, Takashi Shimada, Hiroyasu Matsushima
    2020 9th International Congress on Advanced Applied Informatics (IIAI-AAI) 2020年09月 [査読有り]
  • Masanori Hirano, Kiyoshi Izumi, Hiroyasu Matsushima, Hiroki Sakaji
    The Journal of Artificial Societies and Social Simulation 23 3 6  JASSS 2020年06月30日 [査読有り][通常論文]
     
    Recently financial markets have shown significant risks and levels of volatility. Understanding the sources of these risks require simulation models capable of representing adequately the real mechanisms of markets. In this paper, we compared data of the high-frequency-trader market-making (HFT-MM) strategy from both the real financial market and our simulation. Regarding the former, we extracted trader clusters and identified one cluster whose statistical indexes indicated HFT-MM features. We then analyzed the difference between these traders' orders and the market price. In our simulation, we built an artificial market model with a continuous double auction system, stylized trader agents, and HFT-MM trader agents based on prior research. As an experiment, we compared the distribution of the order placements of HFT-MM traders in the real and simulated financial data. We found that the order placement distribution near the market or best price in both the real data and the simulations were similar. However, the orders far from the market or best price differed significantly when the real data exhibited a wider range of orders. This indicates that in order to build more realistic simulation of financial markets, integrating fine-grained data is essential.
  • Masahiro Suzuki, Hiroki Sakaji, Kiyoshi Izumi, Hiroyasu Matsushima, Yasushi Ishikawa
    Information 11 6 292  2020年05月30日 [査読有り][通常論文]
     
    This paper proposes and analyzes a methodology of forecasting movements of the analysts’ net income estimates and those of stock prices. We achieve this by applying natural language processing and neural networks in the context of analyst reports. In the pre-experiment, we applied our method to extract opinion sentences from the analyst report while classifying the remaining parts as non-opinion sentences. Then, we performed two additional experiments. First, we employed our proposed method for forecasting the movements of analysts’ net income estimates by inputting the opinion and non-opinion sentences into separate neural networks. Besides the reports, we inputted the trend of the net income estimate to the networks. Second, we employed our proposed method for forecasting the movements of stock prices. Consequently, we found differences between security firms, which depend on whether analysts’ net income estimates tend to be forecasted by opinions or facts in the context of analyst reports. Furthermore, the trend of the net income estimate was found to be effective for the forecast as well as an analyst report. However, in experiments of forecasting movements of stock prices, the difference between opinion sentences and non-opinion sentences was not effective.
  • Yoshiyuki Suimon, Hiroki Sakaji, Kiyoshi Izumi, Hiroyasu Matsushima
    Journal of Risk and Financial Management 13 4 82  2020年04月23日 [査読有り][通常論文]
     
    Interest rates are representative indicators that reflect the degree of economic activity. The yield curve, which combines government bond interest rates by maturity, fluctuates to reflect various macroeconomic factors. Central bank monetary policy is one of the significant factors influencing interest rate markets. Generally, when the economy slows down, the central bank tries to stimulate the economy by lowering the policy rate to establish an environment in which companies and individuals can easily raise funds. In Japan, the shape of the yield curve has changed significantly in recent years following major changes in monetary policy. Therefore, an increasing need exists for a model that can flexibly respond to the various shapes of yield curves. In this research, we construct a three-factor model to represent the Japanese yield curve using the machine learning approach of an autoencoder. In addition, we focus on the model parameters of the intermediate layer of the neural network that constitute the autoencoder and confirm that the three automatically generated factors represent the “Level,” “Curvature,” and “Slope” of the yield curve. Furthermore, we develop a long–short strategy for Japanese government bonds by setting their valuation with the autoencoder, and we confirm good performance compared with the trend-follow investment strategy.
  • Kyoto Yono, Hiroki Sakaji, Hiroyasu Matsushima, Takashi Shimada, Kiyoshi Izumi
    Journal of Risk and Financial Management 13 4 79  2020年04月19日 [査読有り][通常論文]
     
    The uncertainty in the financial market, whether the US—China trade war will slow down the global economy or not, Federal Reserve Board (FRB) policy to increase the interest rates, or other similar macroeconomic events can have a crucial impact on the purchase or sale of financial assets. In this study, we aim to build a model for measuring the macroeconomic uncertainty based on the news text. Further, we proposed an extended topic model that uses not only news text data but also numeric data as a supervised signal for each news article. Subsequently, we used our proposed model to construct macroeconomic uncertainty indices. All these indices were similar to those observed in the historical macroeconomic events. The correlation was higher between the volatility of the market and uncertainty indices with larger expected supervised signal compared to uncertainty indices with the smaller expected supervised signal. We also applied the impulse response function to analyze the impact of the uncertainty indices on financial markets.
  • Masanori Hirano, Kiyoshi Izumi, Takashi Shimada, Hiroyasu Matsushima, Hiroki Sakaji
    Journal of Risk and Financial Management 13 4 75  2020年04月17日 [査読有り][通常論文]
     
    In this study, we assessed the impact of capital adequacy ratio (CAR) regulation in the Basel regulatory framework. This regulation was established to make the banking network robust. However, a previous work argued that CAR regulation has a destabilization effect on financial markets. To assess impacts such as destabilizing effects, we conducted simulations of an artificial market, one of the computer simulations imitating real financial markets. In the simulation, we proposed and used a new model with continuous double auction markets, stylized trading agents, and two kinds of portfolio trading agents. Both portfolio trading agents had trading strategies incorporating Markowitz’s portfolio optimization. Additionally, one type of portfolio trading agent was under regulation. From the simulations, we found that portfolio optimization as each trader’s strategy stabilizes markets, and CAR regulation destabilizes markets in various aspects. These results show that CAR regulation can have negative effects on asset markets. As future work, we should confirm these effects empirically and consider how to balance between both positive and negative aspects of CAR regulation.
  • Iwao Maeda, David deGraw, Michiharu Kitano, Hiroyasu Matsushima, Hiroki Sakaji, Kiyoshi Izumi, Atsuo Kato
    Journal of Risk and Financial Management 13 4 71  2020年04月 [査読有り][通常論文]
     
    Prediction of financial market data with deep learning models has achieved some level of recent success. However, historical financial data suffer from an unknowable state space, limited observations, and the inability to model the impact of your own actions on the market can often be prohibitive when trying to find investment strategies using deep reinforcement learning. One way to overcome these limitations is to augment real market data with agent based artificial market simulation. Artificial market simulations designed to reproduce realistic market features may be used to create unobserved market states, to model the impact of your own investment actions on the market itself, and train models with as much data as necessary. In this study we propose a framework for training deep reinforcement learning models in agent based artificial price-order-book simulations that yield non-trivial policies under diverse conditions with market impact. Our simulations confirm that the proposed deep reinforcement learning model with unique task-specific reward function was able to learn a robust investment strategy with an attractive risk-return profile.
  • Tomoki Ito, Kota Tsubouchi, Hiroki Sakaji, Tatsuo Yamashita, Kiyoshi Izumi
    Advances in Intelligent Systems and Computing 1009 AISC 183 - 191 2020年 [査読有り]
     
    Online reviews such as posts on financial micro-blogs are useful for decision making in the investment. However, to read all the posts should not be practical because the volume of the posts is sometimes very large. In this paper, we develop a novel word cloud based framework for visualizing online reviews. Using the LRP method, we visualize the online reviews in the form that we can visually catch-up the sentiments of reviews in cluster units. Images generated from the proposed framework in this paper should be useful in decision making in the investment.
  • Yoshiyuki Suimon, Hiroki Sakaji, Kiyoshi Izumi, Takashi Shimada, Hiroyasu Matsushima
    International Journal of Smart Computing and Artificial Intelligence 4 1 1 - 17 2020年 [査読有り]
  • 清水 たくみ, 早矢仕 晃章, 深見 嘉明, 松島 裕康, 坂地 泰紀, 大澤 幸生
    情報通信政策研究 4 1 103 - 123 総務省情報通信政策研究所 2020年 [査読有り]
     

    様々な産業・経済システムのデジタル化が進展する中、デジタルデータを活用した価値創造への期待が高まっている。その流れを受けて、異なる事業者間でデータを交換・取引するデータ流通エコシステムが萌芽し、新たなイノベーションの源泉として注目され始めている。このようにデータ流通を通じた価値創造への関心は高まっているものの、法制度設計と技術仕様が複雑に絡み合うことから、データ流通エコシステム全体を包括的に捉えた議論が進んでいるとは言い難い。加えて、データ流通プラットフォームの仕組みやサービスについてのシステム面・ビジネス面に絞られた形での議論は存在する一方で、実際に取り扱われるデータ自体の特徴やデータ間の関係性などのデータ流通メカニズムについて十分には明らかになっていない。そこで本論文では、データ流通プラットフォームにおけるデータ連携・相互作用メカニズムを、実証分析に基づいて検討した。プラットフォーム上のデータネットワーク分析から、1) 連携可能性を高めるデータの特徴、2) 連携を促進するデータ共有条件、及び3) データネットワーク構造の特徴に関する知見を得た。具体的には、「時間」や「場所」などに関わる変数が多様な異種データの連携可能性を高めることが示された。また、秘匿データと共有可能データの混在が密なネットワーク構造の構成を促し、異分野間データ連携の成立可能性を高めることが明らかになった。加えて、データネットワークは人間関係と似た「局所的に密、大域的に疎」な構造をしており、疎の部分を埋めるデータを提供することで、異分野をつなげるデータ活用を活性化できる可能性が示唆された。これら分析結果をもとに、データ流通プラットフォームの設計・運営及びデータエコノミー推進への示唆と課題を議論する。

  • Tomoki Ito, Kota Tsubouchi, Hiroki Sakaji, Tatsuo Yamashita, Kiyoshi Izumi
    Proceedings of the 2020 SIAM International Conference on Data Mining(SDM) 262 - 270 2020年 [査読有り]
  • Tomoki Ito, Kota Tsubouchi, Hiroki Sakaji, Tatsuo Yamashita, Kiyoshi Izumi
    The Thirty-Fourth AAAI Conference on Artificial Intelligence(AAAI) 4231 - 4238 2020年 [査読有り]
  • Tomoki Ito, Kota Tsubouchi, Hiroki Sakaji, Tatsuo Yamashita, Kiyoshi Izumi
    Data Sci. Eng. 5 2 180 - 192 2020年 [査読有り]
  • Kyoto Yono, Kiyoshi Izumi, Hiroki Sakaji, Takashi Shimada, Hiroyasu Matsushima
    Advances in Intelligent Systems and Computing 125 - 133 2020年 [査読有り]
  • Yuta Niki, Hiroki Sakaji, Kiyoshi Izumi, Hiroyasu Matsushima
    The 2nd International Workshop on Cross-disciplinary Data Exchange and Collaboration (CDEC 2019) 17 - 23 2019年11月 [査読有り][通常論文]
  • Hiroki Sakaji, Yasutomo Kimura, Izumi Kiyoshi, Hiroyasu Matsushima
    The 2nd International Workshop on Cross-disciplinary Data Exchange and Collaboration (CDEC 2019) 24 - 29 2019年11月 [査読有り][通常論文]
  • Ryo Hamawaki, Junichi Ozaki, Kiyoshi Izumi, Takashi Shimada, Hiroki Sakaji, Hiroyasu Matsushima
    IEEE International Conference on System, Man, and Cybernetics 3784 - 3790 2019年10月 [査読有り][通常論文]
  • Hiroki Sakaji, Akio Kobayashi, Masaki Kohana, Yasunao Takano, Kiyoshi Izumi
    The 8th International Workshop on Web Services and Social Media (WSSM-2019), in conjunction with the 22th International Conference on Network-Based Information Systems, (NBiS-2019) 705 - 714 2019年09月 [査読有り][通常論文]
  • Hiroki Sakaji, Ryota Kuramoto, Hiroyasu Matsushima, Kiyoshi Izumi, Takashi Shimada, Keita Sunakawa
    Proceedings of the First Workshop on Financial Technology and Natural Language Processing 40 - 46 2019年08月 [査読有り][通常論文]
  • Masanori Hirano, Hiroki Sakaji, Kiyoshi Izumi, Hiroyasu Matsushima
    Proceedings of the First Workshop on Financial Technology and Natural Language Processing 102 - 107 2019年08月 [査読有り][通常論文]
  • Analysis of Investment Effect using Multiplex Network Simulation of Banks and Firms
    Ryo Hamawaki, Junichi Ozaki, Kiyoshi Izumi, Takashi Shimada, Hiroyasu Matsushima, Hiroki Sakaji
    5th International Conference on Computational Social Science 2019年07月 [査読有り][通常論文]
  • Iwao Maeda, Hiroyasu Matsushima, Hiroki Sakaji, Kiyoshi Izumi, David Degraw, Atsuo Kato, Michiharu Kitano
    7th International Conference on Smart Computing and Artificial Intelligence (SCAI 2019) 673 - 678 2019年07月 [査読有り][通常論文]
  • Kyoto Yono, Kiyoshi Izumi, Hiroki Sakaji, Takashi Shimada, Hiroyasu Matsushima
    7th International Conference on Smart Computing and Artificial Intelligence (SCAI 2019) 661 - 666 2019年07月 [査読有り][通常論文]
  • Yoshiyuki Suimon, Hiroki Sakaji, Kiyoshi Izumi, Takashi Shimada, Hiroyasu Matsushima
    7th International Conference on Smart Computing and Artificial Intelligence (SCAI 2019) 649 - 654 2019年07月 [査読有り][通常論文]
  • Iwao Maeda, Hiroyasu Matsushima, Hiroki Sakaji, Kiyoshi Izumi, David Degraw, Hirokazu Tomioka, Atsuo Kato, Michiharu Kitano
    The International Conference on Decision Economics (DECON 2019) 210 - 218 Springer International Publishing 2019年06月 [査読有り][通常論文]
  • Kyoto Yono, Kiyoshi Izumi, Hiroki Sakaji, Hiroyasu Matsushima, Takashi Shimada
    IEEE Computational Intelligence for Financial Engineering & Economics (CIFEr) 2019 149 - 155 2019年05月 [査読有り][通常論文]
     
    For financial market participants, the current focused topic (Brexit, Federal Reserve Interest-Rate, U.S. and China trade war, etc.) and its sentiments (whether it is Risk-On or Risk-Off) is very important to decide investment strategies.In this study, we proposed extended topic model called supervised Joint Sentiment-Topic model (sJST) which using not only text data but also numeric data as a supervised signal to extract current focused topic and it's sentiment of market. By using the topic and sentiment weight of the market as a features, we apply several machine learning models to predict foreign exchange market price movement.Comparing the average accuracy over 32 currency pairs and prediction models, the result using sJST weight as features achieved 1.52% better performance than the results which use only historical prices as features.Furthermore, comparing the results limited to specific currency pairs which is difficult to predict when using only historical prices as features, the result using sJST weight as features achieved 3.64% better accuracy than the result which use only historical prices as features.
  • Atsuki Nakayama, Kiyoshi Izumi, Hiroki Sakaji, Hiroyasu Matsushima, Takashi Shimada, Kenta Yamada
    IEEE Computational Intelligence for Financial Engineering & Economics (CIFEr) 2019 84 - 89 2019年05月 [査読有り][通常論文]
     
    Predicting the price movements of stocks based on deep learning and high-frequency data has been studied intensively in recent years. Especially, limit order book which describes the supply-demand balance of the market is used as features of a neural network. However, these methods do not utilize the properties of market orders. On the other hand, this study encodes information of time and prices of orders into images. This encoding method can take advantage of these properties. Then, we apply machine learning methods, convolutional neural network (CNN) and logistic regression (LR), to order-based features to predict the direction of short-term price movements. The results show that the execution has the highest prediction power than the order and cancellation information. Moreover, the difference between CNN and LR are small and depends on kinds of stocks.
  • Yoshiyuki Suimon, Hiroki Sakaji, Kiyoshi Izumi, Takashi Shimada, Hiroyasu Matsushima
    IEEE Computational Intelligence for Financial Engineering & Economics (CIFEr) 2019 34 - 40 2019年05月 [査読有り][通常論文]
     
    In recent years, overseas financial system crises (e.g., Lehman shock and European debt crisis) exerted major influence on the Japanese interest rates market through global financial transactions, such as interest rates derivative contracts and many types of interest rates arbitrage strategies. In this research, we examined the effect of overseas interest rates movements to the Japanese rates market. Then, we developed a forecasting model of Japanese yield curve based on a variety of machine learning methods, by considering the information obtained from overseas markets. Finally, we confirmed that the prediction accuracy of Japanese long-term interest rate improved by using the US interest rates data in addition to the Japanese interest rates data for machine learning. Furthermore, we confirmed that the prediction accuracy increased by using US and Japanese rates markets data in recent years, particularly after 2006. This result suggests that information of overseas interest rates can be used to forecast Japanese rates market nowadays.
  • Hiroki Sakaji, Akio Kobayashi, Masaki Kohana, Yasunao Takano, Kiyoshi Izumi
    The 33rd International Conference on Advanced Information Networking and Applications (AINA-2019) 301 - 312 Springer 2019年03月 [査読有り][通常論文]
  • Improving Document Similarity Calculation Using Cosine-Similarity Graphs
    Yasunao Takano, Yusuke Iijima, Kou Kobayashi, Hiroshi Sakuta, Hiroki Sakaji, Masaki Kohana, Akio Kobayashi
    The 33rd International Conference on Advanced Information Networking and Applications (AINA-2020) 512 - 522 Springer 2019年03月 [査読有り][通常論文]
  • Hirofumi Yamamoto, Hiroki Sakaji, Hiroyasu Matsushima, Yuki Yamashita, Kyohei Osawa, Kiyoshi Izumi, Takashi Shimada
    The 33-rd International Conference on Advanced Information Networking and Applications (AINA-2019) 940 - 951 2019年03月 [査読有り][通常論文]
  • Daigo Tashiro, Hiroyasu Matsushima, Kiyoshi Izumi, Hiroki Sakaji
    Quantitative Finance 19 9 1499 - 1506 2019年 [査読有り][通常論文]
     
    Predicting the price trends of stocks based on deep learning and high-frequency data has been studied intensively in recent years. Especially, the limit order book which describes supply-demand balance of a market is used as the feature of a neural network; however these methods do not utilize the properties of market orders. On the other hand, the order-encoding method of our prior work can take advantage of these properties. In this paper, we apply some types of convolutional neural network architectures to order-based features to predict the direction of mid-price trends. The results show that smoothing filters which we propose to employ rather than embedding features of orders improve accuracy. Furthermore, inspection of the embedding layer and investment simulation are conducted to demonstrate the practicality and effectiveness of our model.
  • Chain Bankruptcy Size in Inter-bank Network: the Effects of Asset Price Volatility and the Network Structure
    Ryo Hamawaki, Kiyoshi Izumi, Hiroki Sakaji, Takashi Shimada, Hiroyasu Matsushima
    Journal of Computational Social Science 2 1 53 - 66 2019年 [査読有り][通常論文]
  • Masanori Hirano, Hiroki Sakaji, Shoko Kimura, Kiyoshi Izumi, Hiroyasu Matsushima, Shintaro Nagao, Atsuo Kato
    Information 10 3 102 - 102 2019年 [査読有り][通常論文]
  • Tomoki Ito, Kota Tsubouchi, Hiroki Sakaji, Tatsuo Yamashita, Kiyoshi Izumi
    2019 IEEE CONFERENCE ON COMPUTATIONAL INTELLIGENCE FOR FINANCIAL ENGINEERING & ECONOMICS (CIFER 2019) 27 - 33 2019年 [査読有り]
     
    It has a great demand for automatically visualizing word-level sentiment scores in financial documents in the form that even non-experts can briefly understand documents. In this paper, we aim to develop a method for automatically visualizing the original word-level sentiment (i.e., word-level sentiment before considering the contexts in a document) and the contextual wordlevel sentiment (i.e., word-level sentiment after considering the contexts in a document) of each term in a document.To achieve this aim, we develop a method for assigning both original and contextual word-level sentiment scores to words using the Layer-wise Relevance Propagation (LRP) method. The LRP based approach can consider the contextual information in assigning original and contextual sentiments to words, in contrast to the other approaches. Using synthetic and real financial textual datasets, we demonstrated the validity of our LRP based approach.Moreover, we propose two types of novel text-visualization frameworks: local word-level sentiment visualization (LWSV) and global word-level sentiment visualization (GWSV). The LWSV visualizes both original and contextual word-level sentiment of each term in a document. The GWSV visualizes both original and contextual word-level sentiments of documents in concept units. These types of text-visualization should be helpful for understanding financial documents quickly.
  • Tomoki Ito, Kota Tsubouchi, Hiroki Sakaji, Kiyoshi Izumi, Tatsuo Yamashita
    2019 19TH IEEE INTERNATIONAL CONFERENCE ON DATA MINING (ICDM 2019) 1126 - 1131 2019年 [査読有り]
     
    Although deep neural networks are excellent for text sentiment analysis, their applications in real-world practice are occasionally limited owing to their black-box property. In response, we propose a novel neural network model called contextual sentiment neural network (CSNN) model that can explain the process of its sentiment analysis prediction in a way that humans find natural and agreeable. The CSNN has the following interpretable layers: the word-level original sentiment layer, word-level sentiment shift layer, word-level local contextual sentiment layer, word-level global importance layer, and word-level global contextual sentiment layer. Because of these layers, this network can explain the process of its document-level sentiment analysis results in a human-like way using these layers. Realizing the interpretability of each layer in the CSNN is a crucial problem in the development of this CSNN because the general back-propagation method cannot realize such interpretability. To realize this interpretability, we propose a novel learning strategy called initialization propagation (IP) learning. Using real textual datasets, we experimentally demonstrate that the proposed IP learning is effective for improving the interpretability of each layer in CSNN. We then experimentally demonstrate that both the predictability and explanation ability of the CSNN are high.
  • Tomoki Ito, Hiroki Sakaji, Kiyoshi Izumi
    Advances in Artificial Intelligence - Selected Papers from the Annual Conference of Japanese Society of Artificial Intelligence (JSAI 2019)(JSAI) 215 - 226 2019年 [査読有り]
  • Kiyoshi Izumi, Hiroki Sakaji
    Artificial Intelligence. IJCAI 2019 International Workshops - Macao(IJCAI) 23 - 35 2019年 [査読有り]
  • Yoshinori Tanaka, Syunya Kodera, Fumihito Sato, Hiroki Sakaji, Kiyoshi Izumi
    8th International Congress on Advanced Applied Informatics(IIAI-AAI) 679 - 684 2019年 [査読有り]
  • Masahiro Suzuki, Toshiya Katagi, Hiroki Sakaji, Kiyoshi Izumi, Yasushi Ishikawa
    2019 International Conference on Data Mining Workshops 30 - 36 2019年 [査読有り]
  • Kei Nakagawa, Shingo Sashida, Hiroki Sakaji, Kiyoshi Izumi
    8th International Congress on Advanced Applied Informatics(IIAI-AAI) 655 - 660 2019年 [査読有り][通常論文]
  • 坂地泰紀, 和泉潔, 松島裕康, 川瀬和哉, 林寛
    日本知能情報ファジィ学会誌 31 2 626 - 635 日本知能情報ファジィ学会 2019年 [査読有り][通常論文]
     

    本論文では,地方銀行が業務利用のために蓄積しているテキストを用いて,その地域の景況感を示すインデックスを生成する手法を提案する. 銀行内には業務で扱っている様々なテキストデータがある. 我々は,その中でも接触履歴に着目した. 接触履歴は,行員が顧客と何かしらのやり取りを行った際に記録されるデータであり,様々なことが記述されている. そのような接触履歴を解析すれば,その地域の景況感が分かるのではないかと,我々は考えた. そこで,本研究では,接触履歴から地域景況インデックスを生成した. まず,景況インデックス作成に最適なモデルを景気ウォッチャー調査と接触履歴を用いて検討した. その後,そのモデルを用いて景況インデックスを作成し,既存の指標と比較することで,性能を評価し,高い性能を示した.

  • Enriching folksonomy for online videos
    Hiroki Sakaji, Masaki Kohana, Akio Kobayashi, Hiroyuki Sakai
    International Journal of Grid and Utility Computing 10 3 258 - 264 2019年 [査読有り][通常論文]
  • File Assignment Control for a Web System of Contents Categorization
    Kohana Masaki, Hiroki Sakaji, Akio Kobayashi, Shusuke Okamoto
    Transactions on Computational Collective Intelligence,Lecture Notes in Computer Science 11610 89 - 102 2019年 [査読有り][通常論文]
  • 地方議会会議録における発言文推定手法の性能評価
    桧森 拓真, 木村 泰知, 坂地 泰紀, 荒木 健治
    日本知能情報ファジィ学会誌 31 2 645 - 652 2019年 [査読有り][通常論文]
  • 複数企業からの関連企業の抽出と事業内容に基づく分類
    田中 瑞竜, 酒井 浩之, 坂地 泰紀, 北島 良三
    日本知能情報ファジィ学会誌 31 1 546 - 562 2019年 [査読有り][通常論文]
  • Masanori Hirano, Hiroki Sakaji, Shoko Kimura, Kiyoshi Izumi, Hiroyasu Matsushima, Shintaro Nagao, Atsuo Kato
    The 1st International Workshop on Cross-disciplinary Data Exchange and Collaboration (CDEC) in 18th IEEE International Conference on Data Mining Workshops (ICDMW2018) 191 - 198 2018年11月 [査読有り][通常論文]
  • Impact Assessments of the CAR Regulation using Artificial Markets
    Masahiro Hirano, Kiyoshi Izumi, Hiroki Sakaji, Takashi Shimada, Hiroyasu Matsushima
    International Workshop on Artificial Market 2018 (IWAM2018) in The 21st International Conference on Principles and Practice of Multi-Agent Systems (PRIMA2018) 43 - 58 2018年10月 [査読有り][通常論文]
  • Impact on Financial Markets of Dark Pools, Large Investor, and HFT
    Shin Nishioka, Kiyoshi Izumi, Wataru Matsumoto, Takashi Shimada, Hiroki Sakaji, Hiroyasu Matsushima
    International Workshop on Artificial Market 2018 (IWAM2018) in The 21st International Conference on Principles and Practice of Multi-Agent Systems (PRIMA2018) 4 - 16 2018年10月 [査読有り][通常論文]
  • Wang Zhouhao, Liu Enda, Sakaji Hiroki, Ito Tomoki, Izumi Kiyoshi, Tsubouchi Kota, Yamashita Tatsuo
    JOURNAL OF RISK AND FINANCIAL MANAGEMENT 11 1 2018年03月 [査読有り][通常論文]
     
    In this research, two estimation algorithms for extracting cross-lingual news pairs based on machine learning from financial news articles have been proposed. Every second, innumerable text data, including all kinds news, reports, messages, reviews, comments, and tweets are generated on the Internet, and these are written not only in English but also in other languages such as Chinese, Japanese, French, etc. By taking advantage of multi-lingual text resources provided by Thomson Reuters News, we developed two estimation algorithms for extracting cross-lingual news pairs from multilingual text resources. In our first method, we propose a novel structure that uses the word information and the machine learning method effectively in this task. Simultaneously, we developed a bidirectional Long Short-Term Memory (LSTM) based method to calculate cross-lingual semantic text similarity for long text and short text, respectively. Thus, when an important news article is published, users can read similar news articles that are written in their native language using our method.
  • Shiori Kitamori, Hiroyuki Sakai, Hiroki Sakaji
    2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings 2018- 1 - 7 2018年02月02日 [査読有り][通常論文]
     
    The proportion of individual investors in security markets has been rising at a rapid rate in Japan in recent times. As a result, the need for technology to support individual investors in making smarter investment decisions is ever increasing. When individual investors make investment decisions, what is the most important information they should hunt for? It is to be able to sense the company's future performance forecast. This is because even if the current performance is in the red (unfavorable) zone, stock prices will tend to rise if it shows that the future performance will recover from the enterprise side. Therefore, in this research, we focus on the financial results as an information source including performance forecasts. The summaries of financial statements include sentences concerning business performance forecast of companies and other associated sentences predicting the upcoming economic conditions. In this study, we extract and classify sentences indicating business performance forecasts and economic forecasts from summaries of financial statements. This allows individual investors to easily grasp the company's performance forecast. Our approach consists of two steps. First, our method extracts sentences relating to 'prediction' by using clue expressions. Second, we automatically classify these prediction oriented sentences into two classes by using deep learning. Our method is capable of automatically generating training data for deep learning and greatly omits efforts to create the training data set by hand.
  • Tomoki Ito, Hiroki Sakaji, Kiyoshi Izumi, Kota Tsubouchi, Tatsuo Yamashita
    2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings 2018- 1 - 8 2018年02月02日 [査読有り][通常論文]
     
    Extracting useful information for market trend analysis automatically from textual data is an important issue in the financial field. Textual data from Yahoo finance bulletin board posts is one of the valuable resources. From these posts, we can obtain the useful information such as investors' sentiments and discussions between investors. On these boards, the number of untagged posts is more than ten times that of posts with sentiment tags. Thus, we believe that sentiment indicators extracted from both unlabeled and labeled posts can represent investors' sentiments better than those extracted from only labeled posts. However, no other previous work had focused on using unlabeled posts. In this paper, we propose a framework for extracting sentiment indicators from both labeled and unlabeled posts on stock message boards. First, we develop models for predicting sentiment tags on posts using tagged posts as training data. Then, we assign polarity scores to unlabeled posts by using the prediction models and develop sentiment indicators using the polarity scores of all the stock message board posts. Our experimental results show that the proposed sentiment indicators are more strongly correlated with stock prices than those developed using only the labeled posts. We also develop a novel indicator called GAP using the proposed sentiment indicator and demonstrate experimentally that GAP is helpful for evaluating the reliability of the sentiment tags for stock message board posts.
  • Hiroki Sakaji, Risa Murono, Hiroyuki Sakai, Jason Bennett, Kiyoshi Izumi
    2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings 2018- 1 - 7 2018年02月02日 [査読有り][通常論文]
     
    What would happen if temperatures were subdued and result in a cool summer? One can easily imagine that air conditioner, ice cream or beer sales would be suppressed as a result of this. Less obvious is that agricultural shipments might be delayed, or that sound proofing material sales might decrease. The ability to extract such causal knowledge is important, but it is also important to distinguish between cause-effect pairs that are known and those that are likely to be unknown, or rare. Therefore, in this paper, we propose a method for extracting rare causal knowledge from Japanese financial statement summaries produced by companies. Our method consists of three steps. First, it extracts sentences that include causal knowledge from the summaries using a machine learning method based on an extended language ontology. Second, it obtains causal knowledge from the extracted sentences using syntactic patterns. Finally, it extracts the rarest causal knowledge from the knowledge it has obtained.
  • Hiroki Sakaji, Atsuya Miyazaki, Hiroyuki Sakai, Kiyoshi Izumi
    ADVANCES IN NETWORK-BASED INFORMATION SYSTEMS, NBIS-2017 7 1117 - 1125 2018年 [査読有り]
     
    In this paper, we propose a method for extracting laboratory front pages from university websites. There are more than 779 universities and colleges in Japan. For selecting a university or a college, some high school students want to know what laboratories these universities or colleges have. To learn about these laboratories, high school students have to search the laboratory front pages from the university websites. However, sometimes it is difficult to find a laboratory front page because they are sometimes buried deep in the hierarchy of university websites. Our method extracts laboratory front pages by using a support vector machine model and applying certain rules. We also developed a laboratory search system that can be used to retrieve laboratory front pages extracted with our method. We evaluated our method and confirmed that is attained 85.0% precision and 65.5% recall.
  • Tomoki Ito, Hiroki Sakaji, Kiyoshi Izumi, Kota Tsubouchi, Tatsuo Yamashita
    International Journal of Data Science and Analytics 9 4 431 - 445 2018年 [査読有り][通常論文]
  • Ototake Hokuto, Sakaji Hiroki, Takamaru Keiichi, Kobayashi Akio, Uchida Yuzu, Kimura Yasutomo
    INTERNATIONAL JOURNAL OF WEB INFORMATION SYSTEMS 14 3 357 - 371 2018年 [査読有り][通常論文]
  • Masaki Kohana, Hiroki Sakaji, Akio Kobayashi, Shusuke Okamoto
    32nd International Conference on Advanced Information Networking and Applications Workshops, AINA 2018 workshops, Krakow, Poland, May 16-18, 2018 40 - 44 IEEE Computer Society 2018年 [査読有り][通常論文]
  • Tomoki Ito, Hiroki Sakaji, Kota Tsubouchi, Kiyoshi Izumi, Tatsuo Yamashita
    Advances in Knowledge Discovery and Data Mining - 22nd Pacific-Asia Conference, PAKDD 2018, Melbourne, VIC, Australia, June 3-6, 2018, Proceedings, Part III 247 - 259 Springer 2018年 [査読有り][通常論文]
  • 和泉 潔, 坂地 泰紀
    年金と経済 36 4 23 - 27 年金総合研究センター 2018年01月 [査読無し][招待有り]
  • 高野 海斗, 酒井 浩之, 坂地 泰紀, 和泉 潔, 岡田 奈奈, 水内 利和
    自然言語処理 25 1 3 - 31 一般社団法人 言語処理学会 2018年 [査読有り][通常論文]
     
    <p>本論文では,テキストマイニング技術を用いて,株主招集通知の情報をデータベースに格納する業務の効率化を実現するための応用システムの研究について述べる.効率化したい業務とは,株主招集通知に記載されている議案の開始ページを予測し,その開始ページにおける議案の議案タイトルと議案内容を分類する業務である.本研究では,これらの業務を株主招集通知のテキスト情報を用いて自動的に行うシステムを開発し,実際に運用している.本研究によって実装したシステムと従来の人手による作業の比較実験の結果,作業時間は 1/10 程度に短縮された.議案分類の手法としては,学習データから抽出した特徴語の重みを用いた分類,多層ニューラルネットワーク(深層学習)を用いた分類,抽出した議案タイトルを用いた分類の三手法を用いた.さらに,各手法の評価を行い,各手法の議案ごとの有効性を確認した.</p>
  • Tomoki Ito, Hiroki Sakaji, Kiyoshi Izumi, Kota Tsubouchi, Tatsuo Yamashita
    IEEE International Conference on Data Mining Workshops, ICDMW 2017- 1122 - 1131 2017年12月15日 [査読有り][通常論文]
     
    Automatic creation of polarity dictionaries is an important issue, as explanations of prediction models are often required in the financial industry. This paper proposes a novel method of developing an interpretable and predictable neural network model. The neural network model we built can extract polarity scores of concepts from documents. Furthermore, we can detect pairwise interactions between concepts, and create polarity concept dictionaries using our neural network model. The model was built using vector representations of words and polarity scores for about 100 words provided by financial professionals, and we obtained about a hundred times more polarity scores for unknown words through backpropagation. First, we analyze the properties of our method from a theoretical point of view. We then confirm its capabilities by conducting simulations of assigning polarity scores to unknown words and detecting interactions using artificial data. We subsequently estimate sentiment tags using real financial textual datasets. Compared with other conventional methods, the proposed approach can forecast sentiments with higher F1 scores. Finally, we develop a polarity concept dictionary based on Yahoo! Finance board textual data.
  • 和泉 潔, 坂地 泰紀, 伊藤 友貴, 伊藤 諒
    証券アナリストジャーナル = Securities analysts journal 55 10 28 - 36 日本証券アナリスト協会 2017年10月 [査読無し][招待有り]
  • A Method for Extracting Correct Links from Automatic Created Links on Folksonomy
    Akio Kobayashi, Hiroki Sakaji, Masaki Kohana
    The 6th International Workshop on Web Services and Social Media (WSSM-2017), in conjunction with the 20th International Conference on Network-Based Information Systems (NBiS-2018) 1144 - 1150 Springer 2017年08月 [査読有り][通常論文]
  • A Topic Trend on P2P Based Social Media
    M. Kohana, H. Sakaji, A. Kobayashi, S. Okamoto
    2017 6th International Workshop on Web Services and Social Media, in conjunction with 2017 20th International Conference on Network-Based Information Systems 1136 - 1143 2017年08月 [査読有り][通常論文]
  • A Web-Based Visualization System for Interdisciplinary Research Using Japanese Local Political Corpus
    Hokuto Ototake, Hiroki Sakaji, Keiichi Takamaru, Akio Kobayashi, Yuzu Uchida, Yasutomo Kimura
    Advances in Network-Based Information Systems, The 20th International Conference on Network-Based Information Systems (NBiS-2017) 2017年08月 [査読有り][通常論文]
  • 北森詩織, 酒井浩之, 坂地泰紀
    電子情報通信学会論文誌 D J100-D 2 150‐161  2017年02月 [査読有り][通常論文]
  • Masaki Kohana, Hiroki Sakaji, Akio Kobayashi, Shusuke Okamoto
    2017 IEEE 31ST INTERNATIONAL CONFERENCE ON ADVANCED INFORMATION NETWORKING AND APPLICATIONS (AINA) 614 - 620 2017年 [査読有り][通常論文]
     
    This paper describes a distributed calculation scheme for scoring relationship among documents. This scheme categorizes documents by using an algorithm which calculates a score value for the relationship between a category and a word in a document. The longer calculation time becomes when increasing the number of documents. Therefore, our scheme uses multiple machines. A master node divides a document set into several subsets, and it distributes them to each calculation nodes. Using this distributed calculation makes the calculation time short, and also makes the memory usage low.
  • Ryo Ito, Kiyoshi Izumi, Hiroki Sakaji, Shintaro Suda
    Journal of Mathematical Finance 7 4 869 - 907 2017年 [査読有り][通常論文]
  • Yousuke Kawamoto, Yasuki Nishiyama, Akio Kobayashi, Hiroki Sakaji, Shigeru Masuyama
    4th IGNITE Conference and 2016 International Conference on Advanced Informatics: Concepts, Theory and Application, ICAICTA 2016 2016年12月30日 [査読有り][通常論文]
     
    Communication via comments is carried out actively in video sharing sites. Recently several studies on analysis of communication through comments and collaborative creative activities are conducted. When it becomes possible to analyze comments, these studies become possible to acquire useful information on user's behavior, involvement of users and video works. The interest of the user, an opinion, and the reaction to other users influence on comments. Moreover, the expression refelecting Web's original culture make noise removal difficult and the object to mention does not exist. Thus, useful information is hard to obtain in each study. Moreover, accurate analysis of the expression is difficult, due to unique expressions that appear in comments. In this paper, we created a dictionary where unique expressions that appear in a video sharing site are registered in 2the dictionary, and propose an analysis method of these comments using the dictionary.
  • Yasutomo Kimura, Keiichi Takamaru, Takuma Tanaka, Akio Kobayashi, Hiroki Sakaji, Yuzu Uchida, Hokuto Ototake, Shigeru Masuyama
    Proceedings of the 12th Workshop on Asian Language Resources (ALR12), The COLING 2016 Organizing Committee 78 - 85 2016年12月 [査読有り][通常論文]
  • Hiroki Sakaji, Masaki Kohana, Akio Kobayashi, Hiroyuki Sakai
    PROCEEDINGS OF 2016 19TH INTERNATIONAL CONFERENCE ON NETWORK-BASED INFORMATION SYSTEMS (NBIS) 550 - 553 2016年 [査読有り][通常論文]
     
    In this paper, we propose a method that estimates new tags from user comments on videos on the Nico Nico Douga website. On Nico Nico Douga, users can post tags and comments on videos. However, users cannot post more than 12 tags on a video; therefore, there are some important tags that could be posted but are sometimes missed. We present a technique to acquire some of these missing tags by choosing new tags that score well in a scoring method developed by us.
  • Hiroki Sakaji, Junya Ishibuchi, Hiroyuki Sakai
    IJSSC 6 3 165 - 172 2016年 [査読有り][通常論文]
  • 決算短信PDFからの原因・結果表現の抽出
    坂地 泰紀, 酒井 浩之, 増山 繁
    電子情報通信学会論文誌D J98-D 5 811 - 822 2015年05月 [査読有り][通常論文]
  • 酒井 浩之, 西沢 裕子, 松並 祥吾, 坂地 泰紀
    人工知能学会論文誌 30 1 172 - 182 The Japanese Society for Artificial Intelligence 2015年01月06日 [査読有り][通常論文]
     
    In this paper, we propose a method of extracting causal information from PDF files of the summary of financial statements of companies, e.g., The sales of smart phones was expanded continually. Cause information is useful for investors in selecting companies to invest. We downloaded 106,885 PDF files of the summary of financial statements of companies from Web pages of the companies automatically. Our method extracts causal information from the PDF files by using clue expressions (e.g., was expanded) and keywords relevant to a company. The clue expressions are extracted from the PDF files of the summary of financial statements of companies and articles concerning business performance of companies automatically. We developed the search system which is able to retrieve causal informations extracted by our method. The search system shows causal information containing a keyword inputted by users, and the summary of financial statements containing the retrieved causal information. We evaluated our method and it attained 83.91% precision and 55.04% recall, respectively. Moreover, we compared our method with Sakai et als method originally proposed for extracting causal information from financial articles concerning business performance of companies and experimental results showed that our method outperforms Sakai et als method.
  • Hiroki Sakaji, Junya Ishibuchi, Hiroyuki Sakai
    PROCEEDINGS 2015 18TH INTERNATIONAL CONFERENCE ON NETWORK-BASED INFORMATION SYSTEMS (NBIS 2015) 669 - 672 2015年 [査読有り][通常論文]
     
    In this paper, we propose a method that extracts positive comments (e.g. (motto hyouka sarerubeki: should be valued)") and negative comments (e.g. (hetakuso: dub)") from Nico Nico Douga automatically. For example, positive comments extracted by our method are beneficial for analysis of campaign broadcast.
  • 坂地 泰紀, 増山 繁
    電子情報通信学会論文誌. D, 情報・システム 94 8 1496 - 1506 一般社団法人電子情報通信学会 2011年08月 [査読有り][通常論文]
     
    本論文では,新聞記事から因果関係を含む文を自動的に抽出する手法を提案する.現在,ウェブページや新聞記事を含む大規模な機械可読文書が入手可能であり,その中には実アプリケーションに役立つ様々な情報が存在し,テキストマイニング技術を用いることで獲得することが可能である.そのような情報の一つに因果関係があり,本研究では因果関係の存在を示す手掛りとなる表現に基づいた因果関係を含む文の抽出を行った.その結果,人手により作られた辞書やパターンを用いず,自動的に因果関係を含む文を抽出することができた.本手法は,素性として構文的な素性と,意味的な素性を用いた.また,追加学習データを自動的に獲得することができる.その結果,性能が向上し,F値0.797を達成した.
  • 谷口 将太, 坂地 泰紀, 酒井 浩之, 増山 繁
    電子情報通信学会論文誌. D, 情報・システム 94 6 1039 - 1043 一般社団法人電子情報通信学会 2011年06月 [査読有り][通常論文]
     
    経済新聞記事から抽出した景気動向を示唆する表現を対象として,その表現が景気の回復を示唆する表現なのか,景気の悪化を示唆する表現なのかのいずれかに自動的に分類する研究を行った.
  • 坂地 泰紀, 野中 尋史, 酒井 浩之, 増山 繁
    電子情報通信学会論文誌. D, 情報・システム 93 6 742 - 755 一般社団法人電子情報通信学会 2010年06月 [査読有り][通常論文]
     
    特許文書から直接的なユーザの便益に相当する表現と,技術上の解決課題を示す表現を自動的に抽出するアルゴリズム「Cross-Bootstrapping」を提案する.特許出願件数は年間40万件にものぼり,1文書当りの文章量も膨大であるため,出願動向調査に有用なパテントマップ(特許出願動向を可視化したもの)を手作業で作成するには多大な時間とコストを要するため,その作成に役立つ情報を自動的に抽出する技術が求められている.そこで,本研究ではパテントマップの作成に役立つ「直接的なユーザの便益に相当する表現」と「技術上の解決課題を示す表現」を自動的に抽出する.本手法は,二つの手がかりと統計随報を用いて,ブートストラップ的に表現対を抽出する.また,辞書や人手により作成したパターンを用いず,自動的に表現を抽出することができる.最後に本手法の評価実験を行い,F値0.89と高い性能を達成したことを確認した.
  • Hirofumi Nonaka, Akio Kobayahi, Hiroki Sakaji, Yusuke Suzuki, Hiroyuki Sakai, Shigeru Masuyama
    40th International Conference on Computers and Industrial Engineering: Soft Computing Techniques for Advanced Manufacturing and Service Systems, CIE40 2010 63 105 - 111 2010年 [査読有り][通常論文]
     
    A patent map, a visual representation of related patent information, is an effective strategic tool for analysis of patent application trends. In particular, an effect-technology type patent map is commonly used as patent examinations are based on technology and effect terms of the inventions. However, conventional patent mining tools are not realized to make an effect-technology type patent map. In order to generate effect-technology type patent maps automatically, we propose a method for extracting information on the effect and the technological terms from patent documents. Our proposed method is described by the following steps. (Step 1) Extract expressions containing the information on the technical effect by using frequent expressions and clue expressions that are automatically extracted by using statistical information and initial clue expressions. (Step 2) Extract technology terms in the above expressions by character-string comparison between patent claims and the expressions. (Step 3) Extract effect terms using grammar patterns that are composed of technology terms, frequent expressions and clue expressions. Our method achieves high precision (85.0%) by using frequent expressions, clue expressions, and the grammar patterns.
  • Hiroki Sakaji, Satoshi Sekine, Shigeru Masuyama
    PRACTICAL ASPECTS OF KNOWLEDGE MANAGEMENT, PROCEEDINGS 5345 111 - + 2008年 [査読有り][通常論文]
     
    This paper proposes a method to extract causal knowledge (cause and effect relations) using clue phrases and syntactic patterns from Japanese newspaper articles concerning economic trends. For example, a sentence fragment "World economy recession due to the subprime loan crisis..." contains causal knowledge in which "World economy recession" is an effect phrase and "the subprime loan crisis" is its cause phrase. These relations are found by clue phrases, such as "(sic) (tame: because)" and "(sic) (niyori: due to)". We, first, investigated newspaper corpus by annotating causal knowledge and clue phrases. We found that some specific syntactic patterns axe useful to improve accuracy to extract causal knowledge. Finally, we developed our system using the clue phrases and the syntactic patterns and showed the evaluation results on a large corpus.
  • Hiroki Sakaji, Hiroyuki Sakai, Shigeru Masuyama
    Advances in Knowledge Discovery and Data Mining, 12th Pacific-Asia Conference, PAKDD 2008, Osaka, Japan, May 20-23, 2008 Proceedings 977 - 984 Springer 2008年 [査読有り][通常論文]

書籍

講演・口頭発表等

  • テキストマイニングを用いた地方景況感の分析  [招待講演]
    坂地泰紀
    第54回サイバーワールド(CW)研究会 2023年08月
  • 経済因果チェーンの構築とその応用  [招待講演]
    坂地泰紀
    Workshop on Microeconomic Analysis of Social Systems and Institutions: Theory, Experiment, and Empirical Studies 2023年03月
  • BERTと因果抽出を用いた気候変動ナラティブの可視化/指数化  [招待講演]
    坂地泰紀, 金田規靖
    日本銀行金融研究所ファイナンス・ワークショップ 2022年11月
  • Economic Causal-Chain Search using Text Mining Technology  [招待講演]
    Hiroki Sakaji
    Knowledge Graphs in Finance and Economics, AKBC Conference 2022 Workshop 2022年11月
  • 決算短信定性的情報の活用方法・分析ポイントについて~決算短信への期待~  [招待講演]
    坂地泰紀
    決算短信からのデータ抽出に関するハンズオン「JPX総研(日本取引所グループ)/AWS共催」 2022年10月
  • 地域経済の未来をデータで読み解く!  [招待講演]
    坂地泰紀, 砂川恵太, 上島邦彦
    日本データ取引所主催シンポジウム「データ流通市場の歩き方」 2022年01月 公開講演,セミナー,チュートリアル,講習,講義等
  • 金融テキストマイニングの最新技術とその応用  [招待講演]
    坂地泰紀
    日本経済研究センター主催「AI・ビッグデータ経済モデル研究会」 2021年01月 公開講演,セミナー,チュートリアル,講習,講義等
  • テキストを用いた経済・経営分析のための人工知能技術  [招待講演]
    坂地泰紀
    長岡市主催「ビジネスのためのAI・IT・IoT活用セミナー」 2020年12月 公開講演,セミナー,チュートリアル,講習,講義等
  • 因果チェーンとネットワーク学習によるソースごとのCOVID19に関する捉え方の違いの抽出  [招待講演]
    坂地泰紀, 久野遼平
    東京大学金融教育研究センター・日本銀行調査統計局共催 ビッグデータフォーラム 2020年11月 公開講演,セミナー,チュートリアル,講習,講義等
  • 多様なテキストデータを対象にした金融テキストマイニング  [招待講演]
    坂地泰紀
    第293回MPTフォーラム 2019年05月 公開講演,セミナー,チュートリアル,講習,講義等
  • 言語処理技術を用いた金融テキストマイニング  [招待講演]
    坂地泰紀
    北海道大学講演会(電子情報通信学会・日本知能情報ファジィ学会) 2018年07月 公開講演,セミナー,チュートリアル,講習,講義等

作品等

その他活動・業績

  • 指田 昌樹, 和泉 潔, 坂地 泰紀 人工知能学会第二種研究会資料 2023 (FIN-031) 55 -60 2023年10月10日 
    責任投資の原則(PRI)のもと機関投資家は、ESGの取り組みを取り入れた投融資をすることが求められている。しかし、現在日本においてはESG及びSDGsの取り組みに関する開示は、主にPDFベースのサステナビリティレポートや統合報告書で行なわれており、統一的なフォームは存在していない。そのため、ESG投資の責任者は手作業で各企業の取り組みの状況を評価する必要が生じ、多くの手がかかっている。本論文では、サステナビリティレポート等のPDFベースの情報からSDG関連の文を自動で抽出し、17のゴールに分類を行なうモデルの構築した。分類モデルは、事前学習済みのsentence-BERTモデルをファインチューニングして作成した。作成の過程で、SDGsのゴールのうち、「貧困」等に取り組んでいる企業が日本では少なく、学習データとなる事例が少ないという問題が生じた。そこで、ChatGPTを活用し、「貧困」等に関する文章を生成しそれを学習データとして用いることで、その課題の解決を図った。
  • 上田 翼, 和泉 潔, 坂地 泰紀 人工知能学会第二種研究会資料 2023 (FIN-030) 40 -44 2023年03月04日 
    COVID-19の流行以降、サプライチェーンの混乱が経済や資産市場に大きな影響を及ぼしている。政策当局や金融市場関係者の間で、供給関連指標に対する関心は高まっているが、データの粒度や迅速性の点で課題が残る。そこで、本研究では、オルタナティブデータと深層学習手法を用いて、自動車サプライチェーンの異常度をリアルタイムで測定する指数の構築を試みた。構築した指数は、ミクロ的な生産障害を把握する上で有用であり、既存の統計指標とも一定の関係性があることを確認した。
  • Rei Taguchi, Hikaru Watanabe, Hiroki Sakaji, Kiyoshi Izumi, Kenji Hiramatsu Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 13859 LNAI 171 -181 2023年 
    In this study, we demonstrate whether analysts’ sentiment toward individual stocks is useful for stock market analysis. This can be achieved by creating a polarity index in analyst reports using natural language processing. In this study, we calculated anomaly scores for the created polarity index using anomaly detection algorithms. The results show that the proposed method is effective in detecting the turning point of the polarity index.
  • 王 美蘊, 坂地 泰紀, 東谷 泰明, 岩垂 光宏, 和泉 潔 人工知能学会全国大会論文集 JSAI2023 2N1GS1004 -2N1GS1004 2023年 
    本研究では,因果関係抽出と類似度分析を組み合わせ,知識発見のための知的財産管理に対する新しいアプローチを提案した.既存の研究は特定の分野で技術トレンドの予測や特許テキストのマイニングに限られている.我々の提案した方法は,技術の潜在的なアプリケーションを多くの分野に広げ,知的財産の二次開発を可能にすることができる.具体的には,多分野からの特許テキストデータの情報を抽出するアプローチを実施した.その後,事前学習言語モデルを使用して特許テキストの分析を行い,最終的に知識発見のための因果チェーンを構築した.専門家による評価データに基づき,我々の方法が既存の深層学習方法よりもロバストであることを示した.
  • 鈴木 雅弘, 坂地 泰紀, 和泉 潔 人工知能学会全国大会論文集 JSAI2023 1E3GS602 -1E3GS602 2023年 
    投資や企業分析において,伝統的に用いられてきた財務情報や経済指標などのデータに対して,今まで利活用の少なかったテキストデータの利用が進んでいる.金融ドメインにおける自然言語処理の研究では,ある一時点の文書の分析や評価が多く,時系列にわたる分析は一時点の評価を並べたものが主になっている.決算短信や有価証券報告書など,金融ドメインでは定期的に様々な文書が発行・公開されている.投資家を始めとしたこれらの文書の読者は,企業について継続的に注目し,同じ企業の文書について前回発行時からの変化に着目することが多い.多くの銘柄を運用する投資家にとって,それぞれの企業について二時点に存在する文書を隅々まで読み込み,変化した点を見つけることは容易ではない.本研究では,時系列に並ぶ2つの文書に対し,前回発行時の文書との差分や変化を抽出するタスクを提案する.具体的には,同じ企業について書かれた2つの金融文書を用い,同じ項目について書かれた箇所を抽出する.抽出した箇所について,人手によって2文書間の差分を抽出する.また提案したタスクに対し,事前学習言語モデルを実際に適用した評価実験を行う.
  • 田口 怜, 坂地 泰紀, 和泉 潔 人工知能学会全国大会論文集 JSAI2023 2T1GS1001 -2T1GS1001 2023年 
    本研究では、金融テキストが株式の戦術的アセットアロケーションに有用であるかどうかを実証する。これは、自然言語処理を用いて金融ニュースから極性指標を作成することで実現できる。作成された極性指標に対して、変化点検知アルゴリズムを用いてクラスタリングを行う。さらに、株式ポートフォリオを構築し、最適化アルゴリズムを用いて各変化点でリバランスを行う。その結果、提案手法は、比較手法を上回る性能を示した。この結果は、極性指標が株式アセットアロケーションの構築に有用であることを示唆している。
  • 坂地 泰紀, 和泉 潔 人工知能学会全国大会論文集 JSAI2023 3A5GS601 -3A5GS601 2023年 
    本論文では,多言語テキストデータから金融の因果関係を表す知識を抽出する方法を提案する.金融分野では,ファンドマネージャや金融アナリストなどが,業務上因果的な知識を必要としている.既存の言語処理技術は人間が認識した因果関係を抽出するのに非常に有効であるが,既存の手法には2つの大きな問題点がある.第一に,多言語での因果関係抽出はこれまで確立されていない.第二に,入れ子の因果関係など,複雑な因果構造を抽出する技術が不十分である.そこで,これらの問題を解決するために,手がかり(becauseやdue toなど)とUniversal Dependenciesに基づいて,入れ子型の因果関係を抽出する手法を提案する.提案する金融因果関係抽出手法を金融分野の多言語テキストデータで評価した結果,提案モデルが既存モデルを上回る性能を持つことが実証された.
  • Cong Liu, 坂地 泰紀, 和泉 潔, 早川 正亮, 塚本 和哉, 加藤 大輔 人工知能学会第二種研究会資料 2022 (FIN-029) 28 -31 2022年10月08日 
    近年、中国経済の躍進に伴い、中国の各国経済に与える影響が高まっている。そのため、米国経済を中心に把握するだけではなく、中国経済の動向を把握することがより重要になっている。しかしながら、中国経済に言及した英語記事は中国語媒体よりも少なく、また、中国語で記載された中国経済に関する膨大な記事から選別してトピックを抽出することは現状難しい。そこで本研究では、中国語記事と英語記事の両方からセンチメントを獲得し、これらを合わせて利用することで、中国市場インデックスを予測する新たなモデルを提案する。
  • 坂地 泰紀, 加藤 大輔, 吉田 佑輔, 渡辺 剛史, 早川 正亮, 和泉 潔 人工知能学会第二種研究会資料 2022 (FIN-028) 113 2022年03月12日 
    本研究では,テキストマイニング技術を用いた中央銀行の金融政策変更を予測するフレームワークを提案する.中央銀行による金融政策の変更は為替市場や株式市場,国債市場などにも影響を与えることから,企業や銀行は金融政策変更の予兆を捉えようと試みているが,中央銀行によって公開される情報の量が限られることから非常に困難である.そこで,我々はテキストマイニング技術を用いて,金融政策変更の予兆に関する情報をニュース記事から抽出することを試みる.この問題を解決するために,我々は因果関係に基づく金融政策変更を予測する新たなフレームワークを提案する.本フレームワークはトピックモデル,機械学習に基づく文選択,機械学習に基づく期待値予測から構成される.
  • 佐野 仁美, 皆川 直人, 坂地 泰紀, 和泉 潔 人工知能学会第二種研究会資料 2022 (FIN-028) 105 2022年03月12日 
    近年,局地的大雨の事象は年々増加し,水災害により甚大な経済的ダメージを被る事例が増えている.特に,被災地では当座の生活における安定的な決済手段が必要であり,これまでは現金決済が最も有用な手段の一つと考えられてきた.しかし,現在は,国内でキャッシュレス決済が推進され,都心部を中心に現金決済の比率が減少傾向であり,災害時における現金需要の行動様式にも変化をもたらす可能性が考えられる.このため,本研究では,自然災害のうち大雨の事例にフォーカスし,降雨情報(国土交通省のXRAIN)と金融情報(銀行ATM 統計情報)の異分野データを互いの位置情報を利用して融合する新たな手法により,降雨発生時における現金需要の動向を分析した.
  • 鈴木 雅弘, 坂地 泰紀, 平野 正徳, 和泉 潔 人工知能学会第二種研究会資料 2022 (FIN-028) 132 2022年03月12日 
    本研究では決算短信や有価証券報告書を用い,言語モデルのBERT とELECTRA について,事前学習や追加で事前学習(追加事前学習) を行いモデルを構築する.構築したモデルについて,金融ドメインのタスクによって汎用コーパスを用いたモデルとの性能を比較する.その際,ファインチューニングを行う層の数などパラメーターによる性能についても比較を行う.構築した一部のモデルについては一般に公開する.
  • 皆川 直人, 和泉 潔, 坂地 泰紀, 佐野 仁美 人工知能学会第二種研究会資料 2022 (FIN-028) 124 2022年03月12日 
    金融機関の取引データを活用すると,企業の活動をリアルタイムで把握することが可能である.これらのデータを有効活用すれば与信管理をはじめ,CRM 等にも活用使途が見出せる.一方,主要な取引先以外の企業については取引量が少ない等,データ活用上は注意を要する.すなわち,潜在的な取引を見逃している可能性がある.本問題に対し,取引主体をノード,取引有無をエッジとした企業間ネットワークを構成すれば,取引有無の予測は本ネットワーク上のリンク予測問題として定式化できる.昨今のグラフ上の深層学習の進展に伴い,より精度の高いリンク予測が可能となった.本研究では,最近のグラフ深層学習手法を使った潜在的な取引予測について紹介する.
  • Masanori Hirano 0001, Hiroki Sakaji, Kiyoshi Izumi CoRR abs/2204.13338 2022年
  • 高柳 剛弘, 坂地 泰紀, 和泉 潔 人工知能学会全国大会論文集 JSAI2022 2A6GS601 -2A6GS601 2022年 
    本研究では,新たな銘柄埋め込みを用いて,あるテーマに関連する銘柄群を抽出する手法を提案する.様々な投資信託 会社は,テーマ型ファンドと呼ばれるテーマに関連したファンドを商品として扱っている.テーマ型ファンドは,その 時代の流行に合わせて作成され,売り出されているが,テーマに関連する銘柄を人手で選択しており,構築にはコスト がかかっている.そこで,本研究では,近年注目されている埋め込み技術を用いて,銘柄を埋め込み表現に変換し,そ れを用いて,あるテーマに関する銘柄を抽出する手法を提案する.評価実験の結果,本提案手法が比較手法よりも高い 性能で銘柄を抽出できることを示した.
  • 佐野 仁美, 皆川 直人, 坂地 泰紀, 和泉 潔 人工知能学会全国大会論文集 JSAI2022 4N1GS302 -4N1GS302 2022年 
    近年、局地的大雨の事象は年々増加し、水災害により甚大な経済的ダメージを被る事例が増えている。しかし、こうした気候変動が経済や金融システムの安定に及ぼす影響を評価する有意な手法は未だ得られていない。このため、本研究では水災害要因の一つである局地的大雨にフォーカスし、気象情報(降雨観測情報)と金融情報(ATMからの引出しに関する統計情報)を互いの位置情報を利用して融合する手法を開発した。本稿では、局地的大雨発生時に変化が生じたATMの現金引出動向に関する分析結果を紹介する。
  • MINAKAWA Naoto, IZUMI Kiyoshi, SAKAJI Hiroki, SANO Hitomi Proceedings of the Annual Conference of JSAI JSAI2022 4S1IS2f01 -4S1IS2f01 2022年 
    Transaction data owned by financial institutions can be alternative source of information to comprehend real-time corporate activities. Such transaction data can be applied to predict stock price and macroeconomic indicator as well as to sophisticate credit management, customer relationship management, and etc. However, it needs attention when a financial institution uses transaction data for aforementioned applications since occurrence of transactions depends on miscellaneous factors including customer loyalty, implying missing potential transactions. To solve this issue, we can predict occurrence of transactions by formulating the problem as a link prediction task in a transaction network with bank accounts as nodes and transaction volume as edges. With the recent advances in deep learning on graphs, we can expect better link prediction. We introduce an approach to predict transaction occurrence by using graph neural network with a special attention mechanism and textual industry information, analyzing the effectiveness of the proposed attention mechanism.
  • 鈴木 雅弘, 坂地 泰紀, 平野 正徳, 和泉 潔 人工知能学会第二種研究会資料 2021 (FIN-027) 05 2021年10月09日 
    BERT を始めとする事前学習言語モデルは,様々な自然言語処理のタスクにおいて成果を上げている.これらのモデルの多くはWikipedia やニュース記事などの一般的なコーパスを用いているため,専門的な単語が使用される金融分野においては十分な効果が得られない.本研究では決算短信や有価証券報告書から事前学習言語モデルを構築する.また金融ドメインのタスクによって汎用モデルとの性能を比較する.
  • 水門 善之, 和泉 潔, 坂地 泰紀 人工知能学会第二種研究会資料 2021 (FIN-027) 101 2021年10月09日 
    国債のイールドカーブ情報には中央銀行の金融政策の方針が反映されるほか,市場参加者の物価や景気の見通しが織り込まれている.また金利水準自体は各種経済主体にとっての借り入れコストとなることから,先行きの経済活動に影響を与えうる.これらを踏まえ,本研究では,企業や家計などの経済主体による景気に先行的な経済活動をとらえた各種経済統計に加えて,国債イールドカーブの情報を特徴量として用いた,ニューラルネットワークベースの機械学習手法に基づく短期経済予測モデルを構築した.その結果,深層学習手法の一種であり再帰的なネットワーク構造を持つRNN(リカレントニューラルネットワーク)ベースのモデルにおいて,相対的に高い予測精度を確認した.更に,経済統計のみをモデルの特徴量として用いた場合に比べて,イールドカーブの情報も学習に用いた場合に,先行きの経済予測の精度が改善する傾向を確認した.このことは,経済予測において,イールドカーブに織り込まれる情報の有用性を示す結果と考える.
  • 松原 冬樹, 和泉 潔, 坂地 泰紀 人工知能学会第二種研究会資料 2021 (FIN-026) 69 2021年03月06日 
    Predicting the movement of stock price is an important issue for market participants. Recently, there have been many attempts applying machine learning techniques in financial time series prediction. However, overfitting presents a huge challenge when machine learning approaches are used in financial time series prediction. In this paper, we propose a stock price prediction method utilizing limit order book data from stocks other than target stocks by stratifying the data and holding a multi-phase pre-training considering market liquidity. Experimental results shows that the proposed approach enhances prediction performance.
  • 和泉潔, 坂地泰紀, 佐野仁美 人工知能学会全国大会(Web) 35th 2021年
  • 坂地 泰紀, 和泉 潔, 酒井 浩之 自然言語処理 = Journal of natural language processing 27 (4) 951 -955 2020年12月
  • 坂地 泰紀, 近藤 洋平, 和泉 潔, 長尾 慎太郎, 加藤 惇雄 人工知能学会第二種研究会資料 2020 (FIN-025) 34 2020年10月10日 
    資産運用会社や証券会社などの多くの金融機関では,定期的な(月次,週次等)市況コメントを作成し,顧客に提供している.この市況コメントを作成する際には,ニュース記事や公的機関発表の統計など,様々な情報源にあたり事実関係を確認しながら,実務者の知識や経験に照らして原稿を作成することになる.一方で,こうした市況コメントの調査・執筆の作業は時間とコストが伴うものである.もし,主要な情報源から,自然言語処理をはじめとした機械的なプロセスによって,信頼できる市況コメントを作成することができれば,それは金融機関における当該業務のコストの削減,あるいは処理の迅速化やカバレッジ拡大による顧客サービス向上に資することになる.そこで,本研究では,まずニュース記事に含まれる原因・結果を含む因果文を抽出することで,市況コメントを自動的に生成することを試みた.
  • Fuyuki Matsubara, Kiyoshi Izumi, Hiroki Sakaji, Hiroyasu Matsushima The 34th Annual Conference of the Japanese Society for Artificial Intelligence, 2020 1K4-ES-2-05 2020年06月09日 [査読無し][通常論文]
     
    There have been many studies seeking to predict excess returns in financial time series data. Nevertheless, not many studies have focused on applying machine learning approaches among factors in different asset classes. The main objective of this paper is to analyze whether a predictability of return in financial products could be improved by considering factors obtained from other asset classes, and to indicate the effectiveness of machine learning in financial time series prediction. We targeted 10-year Japanese Government Bond(10-year JGB), and Nikkei Stock Average Index, implementing non-linear machine learning approaches as well as conventional multiple linear regression models to predict returns in both assets. The results suggest that considering factors from other asset classes could improve return prediction both in 10-year JGB and Nikkei Stock Average, especially when using non-linear approaches.
  • STBM : Stochastic Trading Behavior Model for Financial Markets Based on Long Short-Term Memory
    Masanori HIRANO, Hiroyasu MATSUSHIMA, Kiyoshi IZUMI, Hiroki SAKAJI The 34th Annual Conference of the Japanese Society for Artificial Intelligence 2020 1K4-ES-2-04 2020年06月09日 [査読無し][通常論文]
  • 平野 正徳, 坂地 泰紀, 松島 裕康, 和泉 潔 言語処理学会 第26回年次大会 26th 569 -572 2020年03月16日 [査読無し][通常論文]
  • 五十嵐 光秋, 坂地 泰紀, 和泉 潔, 島田 尚, 松島 裕康, 須田 真太郎 言語処理学会 第26回年次大会 26th 2020年03月16日 [査読無し][通常論文]
  • 仁木 裕太, 坂地 泰紀, 松島 裕康, 和泉 潔 言語処理学会 第26回年次大会 26th 2020年03月16日 [査読無し][通常論文]
  • 高嶺 航, 坂地 泰紀, 和泉 潔, 松島 裕康, 島田 尚 言語処理学会 第26回年次大会 26th 2020年03月16日 [査読無し][通常論文]
  • 小寺 俊哉, 佐藤 史仁, 坂地 泰紀, 和泉 潔 人工知能学会第二種研究会資料 2020 (FIN-024) 50 2020年03月14日 
    近年,個別株式のリターン予測において,様々なファクターの中から予測に有用な特徴量を自動で抽出することのできる深層学習技術の応用研究がなされている.しかしながら,深層学習は計算の過程が複雑で,人間にはその予測根拠の把握が難しく,意思決定に理由が求められる実務での利用において,解釈の困難さが課題視されることがある.一方,深層学習の解釈手法についても研究が行われており,深層学習において研究が盛んな画像分類等のタスクだけでなく株価指数や株式個別銘柄等の資産価格リターン予測を行う深層学習モデルに対しても解釈を行う研究が行われている.本稿では,モデルの解釈に焦点を当て,個別銘柄のリターン予測をタスクとした深層学習モデルについて,LRP と呼ばれる深層学習の解釈手法を用いて,各入力値の寄与度を個別銘柄レベルで確認した.さらに,深層学習モデルの入力値に個別銘柄属性だけでなくマーケット指数等の市場情報を用いることで,銘柄属性と市場トレンドの2 つの側面での分析を行った.
  • 中川 慧, 指田 晋吾, 坂地 泰紀, 和泉 潔 人工知能学会第二種研究会資料 2020 (FIN-024) 171 2020年03月14日 
    A lead-lag effect in stock markets describes the situation where one (leading) stock return is cross-correlated with another (lagging) stock return at later times. There are various methods for stock return forecasting based on such a lead-lag effect. One of the most representative methods is based on the supply chain network. In this research, we propose a stock return forecasting method with an economic causal chain. The economic causal chain refers to a cause and effect network structure constructed by extracting a description indicating a causal relationship from the texts of Japanese financial statement summaries. We examine the following lead-lag effect. (1) whether lead-lag effect spreads to the 'effect' stock group when there is a large stock uctuation in the 'cause' stock group in the causal chain. (2) whether lead-lag effect spreads to the 'cause' stock group when there is a large stock uctuation in the 'effect' stock group in the causal chain. We confirm the existence of the both side of lead-lag effect and the evidence of stock return predictability across causally linked firms in the Japanese stock market.
  • 平野 正徳, 坂地 泰紀, 木村 祥子, 和泉 潔, 松島 裕康, 長尾 慎太郎, 加藤 惇雄 第24回金融情報学研究会資料 226 -233 2020年03月14日 [査読無し][通常論文]
  • 水門 善之, 坂地 泰紀, 和泉 潔, 島田 尚, 松島 裕康 第24回金融情報学研究会資料 66 -69 2020年03月14日 [査読無し][通常論文]
  • 坂地 泰紀, 蔵本 涼太, 和泉 潔, 松島 裕康, 島田 尚, 砂川 恵太 第24回金融情報学研究会資料 98 -102 2020年03月14日 [査読無し][通常論文]
  • 小林 弘幸, 和泉 潔, 松島 裕康, 坂地 泰紀, 島田 尚 第24回金融情報学研究会資料 129 -133 2020年03月14日 [査読無し][通常論文]
  • 伊藤友貴, 坪内孝太, 山下達雄, 坂地泰紀, 和泉潔 言語処理学会年次大会発表論文集(Web) 26th 2020年
  • 鈴木雅弘, 堅木聖也, 坂地泰紀, 和泉潔, 石川康 言語処理学会年次大会発表論文集(Web) 26th 2020年
  • 仁木裕太, 坂地泰紀, 和泉潔, 松島裕康 人工知能学会全国大会(Web) 34th 2020年
  • 前田巌, 松島裕康, 坂地泰紀, 和泉潔, DEGRAW David, 加藤惇雄, 北野道春 人工知能学会全国大会(Web) 34th 2020年
  • 若杉亮, 和泉潔, 松島裕康, 坂地泰紀 人工知能学会全国大会(Web) 34th 2020年
  • 酒井浩之, 坂地泰紀, 和泉潔, 松井藤五郎, 入江圭太郎 人工知能学会全国大会(Web) 34th 2020年
  • 鈴木雅弘, 坂地泰紀, 和泉潔, 松島裕康, 石川康 人工知能学会全国大会(Web) 34th 2020年
  • 高頻度電力需要データを用いた製造業活動のナウキャスティングモデルの構築
    水門 善之, 和泉 潔, 坂地 泰紀, 島田 尚, 松島 裕康 第35回社会におけるAI研究会(SIG-SAI), 人工知能学会合同研究会 2019 1 -5 2019年11月 [査読無し][通常論文]
  • 前田 巌, 松島 裕康, 坂地 泰紀, 和泉 潔, ディグロー デビット, 富岡 博和, 加藤 惇雄, 北野 道春 人工知能学会全国大会論文集 JSAI2019 (0) 4I2J201 -4I2J201 2019年06月 [査読無し][通常論文]
     

    近年,社会の様々な場面で機械学習・深層学習手法による予測が活用されている.深層学習手法を用いて学習したモデルは高い精度で予測を行うことができるが,予測信頼性を十分に考慮できておらず,予測の困難な外挿データに対しても高い確信を持って予測を行ってしまう危険がある.本研究では画像識別タスクに対して,通常の深層学習手法および近年提案されている不確かさを考慮した深層学習手法を適用し,外挿データに対するモデルの頑健性を検証した.通常の深層学習手法により学習したモデルはモデル学習用データに存在しない特徴を持つデータに対して高い確信度で予測を行ってしまうが,不確かさを考慮した深層学習手法により学習したモデルはそのようなデータに対し確信度を低く出力し,誤った予測を回避することが可能となる.実験結果より深層学習手法における不確かさ評価の重要性が示唆された.

  • 松浦 出, 和泉 潔, 坂地 泰紀, 松島 裕康, 島田 尚 人工知能学会全国大会論文集 JSAI2019 (0) 1P2-J-13-01 -1P2J1301 2019年06月 [査読無し][通常論文]
     

    インデックス投資が市場の価格形成に与える影響を調べるため,証券市場とその参加者,価格決定をモデル化した.そのモデル上でインデックス投資が価格形成にほとんど影響を与えないことをシミュレーション実験により示した.

  • 余野 京登, 和泉 潔, 坂地 泰紀, 島田 尚, 松島 裕康 人工知能学会全国大会論文集 JSAI2019 (0) 4J3J1302 -4J3J1302 2019年06月 [査読無し][通常論文]
     

    市場参加者にとって、各国の金融政策や財政政策、あるいは2国間の貿易摩擦などのマクロ経済の不確実性は、意思決定をする際に重要な役割を果たす。本研究では、ニューステキストを用いてマクロ経済の不確実性指数の構築を目指す。トピックモデルを拡張し、テキストデータのみならず、数値データを教師信号として用いたモデルの適用を試みた。また、構築した4つのマクロ経済の不確実性指数を対象にした定性的・定量的な分析を行った.

  • 五十嵐 光秋, 坂地 泰紀, 和泉 潔, 島田 尚, 松島 裕康, 須田 真太郎 人工知能学会全国大会論文集 JSAI2019 (0) 2O3J1304 -2O3J1304 2019年06月 [査読無し][通常論文]
     

    決算短信の因果関係ネットワークを構築する実験を行った.因果関係を決算短信から抽出し,表現の類似度を算出してネットワークを構築した.類似度の計算には日本語版ウィキペディアのコーパスから作成したword2vecモデルを使用し,単語の重要度を表すidf値の組み合わせることで,単語ベクトルから因果関係表現のベクトルを獲得した.また、極性辞書を用いて因果表現の極性の反転を捉えることで,word2vecモデルにより同一視される対義語と類義語による違いを検出した.実験によってネットワークから,因果関係どうしを結ぶ100個のエッジを無作為に選択し、目視により評価した。 その結果、妥当な接続と考えられる割合は84%であり、さらにその中で極性の反転が正しく捉えられたエッジの割合は86%であった。

  • 平野 正徳, 和泉 潔, 松島 裕康, 坂地 泰紀, 島田 尚 人工知能学会全国大会論文集 JSAI2019 (0) 2O1J1304 -2O1J1304 2019年06月 [査読無し][通常論文]
     

    本研究は,金融市場における高頻度取引(HFT)のマーケットメイク(MM)戦略と呼ばれる注文行動について分析を行うことを目的とした.株式会社日本取引所グループより提供を受けた,東京証券取引所の注文データを使用し,仮想サーバーの名寄せを前処理として行なった.その結果得られた,取引主体別の注文データを,いくつか指標を使うことで,クラスター分析を行い,高頻度マーケットメイク戦略(HFT-MM)を取っている取引主体を抽出し,それらの注文が,直近約定価格から何ティック離れたところに置かれているかについて計算した.その結果,HFT-MMとされる行動主体は,直近約定価格からかなり離れた位置(5-10ティック)のところにも注文を置いていることが明らかになった.この結果は,HFT-MMとされる取引主体が,マーケットメイク戦略だけではなく,他の戦略も採用している可能性を示唆しており,さらに確認すると,価格が急変した際には,不安定化効果を引き起こす可能性をも示唆していることがわかった.

  • 仁木 裕太, 和泉 潔, 松島 裕康, 坂地 泰紀, 島田 尚 人工知能学会全国大会論文集 JSAI2019 (0) 2O1J1303 -2O1J1303 2019年06月 [査読無し][通常論文]
     

    イタリアの銀行間預金オンライン取引システムe-MIDで行われた取引データから, 1999年から2016年の期間について営業日ごとの銀行間ネットワークを構築した. 構築した銀行間ネットワークについて, ノードのネットワーク特徴量を計算し, それを入力データとしてクラスタリング構造変化検知による分析を行った. この分析の結果, リーマンショック後のダイナミクスの特徴が銀行の種類によって異なることが明らかになった.

  • 高嶺 航, 和泉 潔, 坂地 泰紀, 松島 裕康, 島田 尚, 清水 康弘 人工知能学会全国大会論文集 JSAI2019 (0) 1P2J1305 -1P2J1305 2019年06月 [査読無し][通常論文]
     

    本論文では、新聞記事からのテキストマイニングによる因果関係を考慮したアナリストレポートの要約文生成手法の提案を行う。因果関係抽出本手法を用いれば、機関投資家の大量の文書を熟読しなければならない負担を軽減し、証券アナリストの肩が投資判断に関わる重要な経済情報を簡単に収集することができる。そこで今回は、本手法における因果関係表現抽出手法の妥当性を野村證券株式会社のテキストデータを用いて評価及び分析を行う。全アナリストレポートの7割程度で、予想の根拠情報を取得することができ、本手法における因果関係抽出手法の有用性を示した。また、評価をするにあたり、証券アナリストの予想の書きぶりも示すことができた。

  • 濱脇 諒, 尾崎 順一, 和泉 潔, 島田 尚, 松島 裕康, 坂地 泰紀 人工知能学会全国大会論文集 JSAI2019 (0) 2O3J1302 -2O3J1302 2019年06月 [査読無し][通常論文]
     

    近年経済のグローバル化によりある金融機関の破綻の影響がその業界や国を超えて国際的なものとなっている。 関連する分野の先行研究は、銀行単体もしくは企業単体でのエージェントシミュレーションを行うものがほとんどである。 そこで、本研究では銀行と企業の相互作用に焦点を当てた多層ネットワークのモデルを提案し、銀行から企業への投資の有無が企業の成長率に与える影響を調べる。 具体的には、投資を受けているかどうかによって成長率の変動幅を変化させ、最終的な企業の規模の分布の変化について議論する。 銀行が持つ貸借対照表は全国銀行協会が公表している全国銀行財務諸表分析に基づいて作成した。 シミュレーションの結果として、投資は企業の成長の平均には影響を与えなかったが、企業間の格差が広がった。 これは投資を受けた企業のうち、投資を活かして成長することができたものと失敗したものに別れてしまったことが原因だと考えられる。

  • 不動産分野へのデータ解析と人工知能技術の応用
    和泉潔, 諏訪博彦, 大西立顕, 坂地泰紀, 松島裕康 不動産テックを考える,赤木正幸,浅見泰司,谷山智彦編,プログレス 75 -89 2019年05月 [査読無し][通常論文]
  • 酒井 浩之, 坂地 泰紀, 和泉 潔, 松井 藤五郎, 入江 圭太郎 人工知能学会第二種研究会資料 2019 (FIN-022) 61 2019年03月03日 
    本研究では、経済新聞記事などの経済テキストから、日経平均株価などの市況について言及している記事を抽出し、それらの内容を自動的に要約することによりマーケットレポートにおける市況分析コメントを自動生成する手法の開発を行う。しかし,日経平均株価の市況について言及している記事のみでは,指定した期間において重要な内容について言及している文の数が少なく,そのために重要な内容が市況分析コメントに含まれないことがある.そこで本研究では,ある期間において日経平均株価に影響を与えたイベントを推定し,そのイベントについて述べた記事を関連記事として自動的に検索し,検索された関連記事をも使用することで,生成する市況分析コメントの精度を高める手法を提案する.
  • 堅木 聖也, 坂地 泰紀, 和泉 潔, 石川 康, 笠岡 恒平 人工知能学会第二種研究会資料 2019 (FIN-022) 42 2019年03月03日 
    In this paper, we propose a methodology to forecast the direction and extent of volatility in mid-to-long term excess return of stock price by applying natural language processing and neural networks on the context of analyst reports. Analyst reports are prepared by analysts in research department in stock brokerage firms and we consider the content of reports include usefull information to forecast movements in stock prices. First, our method extracts opinion sentences from analyst reports, while the remaining parts correspond to non-opinion sentences. Second, our method predicts stock price movements by inputting opinion sentences and non-opinion sentences to neural networks separately.
  • 松浦 出, 和泉 潔, 坂地 泰紀, 松島 裕康, 島田 尚 人工知能学会第二種研究会資料 2019 (FIN-022) 116 2019年03月03日 
    In this paper, we modeled stock markets to investigate the effect of index investing on stock price formation. We showed that index investing has little effect on stock price formation in our stock markets model by analyzing results from experiments with various market settings.
  • 水門 善之, 坂地 泰紀, 和泉 潔, 島田 尚, 松島 裕康 第22回金融情報学研究会資料 81 -87 2019年03月 [査読無し][通常論文]
  • 山本 寛史, 坂地 泰紀, 松島 裕康, 山下 雄己, 大澤 恭平, 和泉 潔, 島田 尚 第22回金融情報学研究会資料 25 -30 2019年03月 [査読無し][通常論文]
  • 高嶺 航, 坂地 泰紀, 和泉 潔, 松島 裕康, 島田 尚, 清水 康弘 第22回金融情報学研究会資料 48 -52 2019年03月 [査読無し][通常論文]
  • 濱脇 諒, 和泉 潔, 坂地 泰紀, 島田 尚, 松島 裕康 第22回金融情報学研究会資料 104 -107 2019年03月 [査読無し][通常論文]
  • 松浦 出, 和泉 潔, 坂地 泰紀, 松島 裕康, 島田 尚 第22回金融情報学研究会資料 116 -119 2019年03月 [査読無し][通常論文]
  • 曽根 泰平, 和泉 潔, 坂地 泰紀, 松島 裕康, 島田 尚 第22回金融情報学研究会資料 131 -138 2019年03月 [査読無し][通常論文]
  • 水門 善之, 坂地 泰紀, 和泉 潔, 島田 尚, 松島 裕康 SIG-SAI 34 (4) 1 -6 2019年03月 [査読無し][通常論文]
  • 和泉 潔, 松島 裕康, 坂地 泰紀 情報処理学会研究報告 2019-ICS-194 (4) 1 -4 2019年03月 [査読無し][通常論文]
  • シミュレーションによるバリューチェーン交渉制度の安定性評価
    和泉 潔, 松島 裕康, 坂地 泰紀 第18回社会システム部会研究会 159 2019年03月 [査読無し][通常論文]
  • 平野 正徳, 坂地 泰紀, 木村 笙子, 和泉 潔, 松島 裕康, 長尾 慎太郎, 加藤 惇雄 言語処理学会第25回年次大会(NLP2019) 25th 597 -600 2019年03月 [査読無し][通常論文]
  • 日米イールドカーブの連動性に基づく機械学習を用いた日本の金利変動モデルの構築
    水門 善之, 坂地 泰紀, 和泉 潔, 島田 尚, 松島 裕康 第50回 日本金融・証券計量・工学学会 大会 2019年02月 [査読無し][通常論文]
  • 和泉潔, 坂地泰紀 人工知能学会全国大会(Web) 33rd 2019年
  • 前田 巌, 松島 裕康, 坂地 泰紀, 和泉 潔, David deGraw, 富岡 博和, 加藤 惇雄 第21回金融情報学研究会資料 21 50 -52 2018年10月 [査読無し][通常論文]
  • 坂地 泰紀, 和泉 潔, 松島 裕康 第21回金融情報学研究会資料 59 -60 2018年10月 [査読無し][通常論文]
  • 酒井 浩之, 坂地 泰紀, 和泉 潔, 松井 藤五郎, 入江 圭太郎 人工知能学会第二種研究会資料 2018 (FIN-020) 44 2018年03月20日 
    本研究では,経済新聞記事などの経済テキストから、日経平均株価などの市況について 言及している文書のみを抽出し,それらの内容を自動的に要約することにより,ファンドの運用報告 書における市況分析コメントを自動生成する手法の開発を行う.本手法では,まず経済新聞記事から 深層学習により日経平均株価の市況について言及している記事を抽出する.次に抽出された記事の 中から例えば「ギリシャへの金融支援協議が難航していることや、中国・上海株の値動きへの警戒感 から、投資家のリスクオフの動きが強まった。」のような日経平均が大幅に変動した理由について言 及している文を抽出する.そして,抽出された文を時系列順に並べることで市況分析コメントを自動 生成する.
  • 丸澤 英将, 和泉 潔, 坂地 泰紀, 田村 浩道, 本廣 守 人工知能学会第二種研究会資料 2018 (FIN-020) 74 2018年03月20日 
    Recently, with the increase of individual investors, the necessity of investment support technologies is increasing. Although analyst reports on which professional securities analysts forecast business performances or stock prices of companies are regarded as important investment decision materials, writing an analyst report is heavy burden. In this research, we summarize newspaper articles and support the generation of analyst reports by using knowledge of information features which are referred to as reasons for analysts' forecasts of business performances or stock prices in analyst reports.
  • 伊藤 友貴, 坂地 泰紀, 和泉 潔 人工知能学会第二種研究会資料 2018 (FIN-020) 61 2018年03月20日 
    経済文書のような専門的な文書は非専門家にとって読みにくい場合が多い.そのため,非専門家を対象に経済文書上のセンチメントを単語単位で可視化するようなサポートシステムを構築することには一定の需要があると思われる.経済文書上のセンチメントを可視化する手段の一つとして近年提案された「ニューラルネットワークモデルの解釈」に関する手法,LRP (Layer-wiseRepresentation Propagation) を用いるという手段がある.しかし現状 LRP が日本語の経済文書の可視化に有用かどうかは調査されておらず,その性質についての詳細な分析もあまりされていない.また,LRP の Attention RNN への適用方法は未だ提案されていない.本報告では LRP の AttentionRNN への適用方法を提案し,また,LRP が日本語金融テキストの可視化に有用かどうかを検証する.さらに,実データを用いた検証の中で LRP を用いた日本語文書可視化の性質について分析する.
  • 西村 弘平, 坂地 泰紀, 和泉 潔 人工知能学会第二種研究会資料 2018 (FIN-020) 50 2018年03月20日 
    複数のテキストデータから経済・金融事象を背景知識まで含めて可視化することは, 経済・金融事象の理解の助けになり有用である. しかしながら, 経済・金融事象の連鎖を手動で抽出することは非常に時間とコストがかかる. そこで, 本研究では経済・金融事象の連鎖を因果関係として扱い, 各事象を表したベクトル間の類似度を用いて因果関係の連鎖を構築する手法を提案する. また,提案手法における問題点から今後の提案をまとめる.
  • 余野 京登, 和泉 潔, 坂地 泰紀 人工知能学会第二種研究会資料 2018 (FIN-020) 67 2018年03月20日 
    金融市場におけるテキストデータは、投資家にとって分析対象の一つであり、その重要性は日に日に増している。決算短信をはじめとする企業の業績について書かれたドキュメント、証券会社のアナリストが書いた個別企業についてのアナリストレポート、更にはツイッター等のSNSで発信している個人投資家のつぶやき等、多種多様なテキストデータが存在する。定性的な投資判断の材料になるとともに、これらのテキストを用いて、モデルを構築し、定量的な分析を行うこともある。本研究では、中央銀行が発行する議事録やステートメントなどの公的な文章を対象に、定量的な数値化を目的としている。具体的には、日本銀行の発行する金融政策決定会合の議事要旨から物価、生産、雇用等の各トピックに対するセンチメント指数の構築を目指す。
  • 坂地 泰紀, 和泉 潔 電子情報通信学会技術研究報告 = IEICE technical report : 信学技報 117 (468) 21 -24 2018年03月02日
  • 地方議会会議録における発言文の推定
    桧森拓真, 木村泰知, 坂地泰紀 言語処理学会第24回年次大会(NLP2018) P1-3 125 -128 2018年03月 [査読無し][通常論文]
  • 佐藤 史仁, 佐久間 洋明, 小寺 俊哉, 田中 良典, 坂地 泰紀, 和泉 潔 人工知能学会全国大会論文集 JSAI2018 2O404 -2O404 2018年 
    有価証券報告書には,業績の他,リスク対策や企業の施策等,決算短信にはない情報の記載もある.また,先行研究では,多くの情報から重要な文を効率よく抽出する方法として,因果関係文を重要文とした手法が提案されている.しかし,抽出対象を決算短信等とした報告はあるが有価証券報告書とした報告はない.そこで本稿では,この手法を応用し,有価証券報告書専用の因果関係文を抽出する判別モデルを提案した.そして,判別モデルの評価等を行い,高い性能であることなどを示した.この判別モデルにより有価証券報告書独自の投資判断に有益な情報の効率的な抽出が期待できる.
  • 坂地泰紀, 和泉潔 電子情報通信学会技術研究報告 117 (468(AI2017 43-51)) 2018年
  • 余野京登, 和泉潔, 坂地泰紀 人工知能学会全国大会論文集(CD-ROM) 32nd 2018年
  • 坂地泰紀, BENNETT Jason, 宮尾祐介, 和泉潔 人工知能学会全国大会論文集(CD-ROM) 32nd 2018年
  • 西村弘平, 坂地泰紀, 和泉潔 人工知能学会全国大会論文集(CD-ROM) 32nd 2018年
  • 濱脇諒, 和泉潔, 坂地泰紀, 米納弘渡 人工知能学会全国大会論文集(CD-ROM) 32nd 2018年
  • 丸澤 英将, 和泉 潔, 坂地 泰紀, 田村 浩道 人工知能学会第二種研究会資料 2017 (FIN-019) 71 2017年10月14日 
    These days, a growing number of individual investors is attracting public attentions even in Japan, and securities companies are actively providing them with investment informations. Especially, analyst reports written by professional security analysts are important investment judgment materials, but their timing of publication varies by brands. In this paper, using the structures of causal relationships of sentences in analyst reports, the ways of security analysts paying attention to cause informations concerning business performances were learned, and newspaper articles including similar causal relationships were extracted. We aim to realize a real-time investment supporting system.
  • 伊藤 諒, 坂地 泰紀, 和泉 潔, 須田 真太郎 人工知能学会第二種研究会資料 2017 (FIN-019) 78 2017年10月14日 
    In recent years, textual information, which is unstructured data attracts attention as new analytical data in the financial and economic fields and it is expected to structure knowledge on this domain. One such knowledge is a sentiment polarity dictionary in which each word is representing positive or negative. In building the dictionary, it is costly to add the polarity value to a vast number of words manually. Therefore, in this research, we propose a the dictionary construction model especially considering the synonymity and symmetry of words. As a result of the experiment, the proposed method is a more accurate than the model of the previous research. In addition, we extended the conventional dictionary using the proposed method, and we showed that the extended dictionary has higher accuracy than the dictionary which is not extended.
  • 室野 莉沙, 酒井 浩之, 坂地 泰紀, ベネット ジェイスン 電子情報通信学会技術研究報告 = IEICE technical report : 信学技報 117 (207) 93 -98 2017年09月07日
  • 地方議会の議案収集に向けた議案一覧抽出の試み
    田中琢真, 坂地泰紀, 小林暁雄, 木村泰知, 増山繁 電子情報通信学会技術研究報告 117 (207) 41 -46 2017年09月 [査読無し][通常論文]
  • 木村 泰知, 小林 暁雄, 坂地 泰紀, 内田 ゆず, 高丸 圭一, 乙武 北斗, 吉田 光男, 荒木 健治 第31回人工知能学会全国大会 2017年05月 [査読無し][通常論文]
  • 高野 海斗, 酒井 浩之, 坂地 泰紀, 和泉 潔, 岡田 奈奈, 水内 利和 人工知能学会第二種研究会資料 2017 (FIN-018) 10 2017年03月10日 
    In this research, we aim to predict start pages of proposals stated in notice of the meeting of shareholders and classify which proposal the page is. We propose two methods that classification method of proposals. The first method heuristically predicts the page on which the proposal is described. Moreover our method extracts specialized terms of each proposal and assigns weights to them. After that, our method classifies proposals by specialized terms. The second method classifies proposals using deep learning. Each methods were evaluated, and the effectiveness of each methods was verified.
  • 都道府県議会会議録からの意見や意志を表す発言の抽出
    坂地泰紀, 酒井浩之, 小林暁雄, 内田ゆず, 乙武北斗, 高丸圭一, 木村泰知 言語処理学会第23回年次大会(NLP2017) 426 -429 2017年03月 [査読無し][通常論文]
  • 木村泰知, 小林暁雄, 坂地泰紀, 内田ゆず, 高丸圭一, 乙武北斗, 吉田光男, 川浦昭彦 言語処理学会第23回年次大会 54 -57 2017年03月 [査読無し][通常論文]
  • 宮崎 敦也, 酒井 浩之, 坂地 泰紀 成蹊大学理工学研究報告 53 (2) 25 -28 2016年12月 
    In this paper, we propose a method that extracts laboratory front pages from university web sites. Our method extracts the laboratory front pages by using SVM and applying some rules. Moreover, we developed the laboratory search system which is able to retrieve laboratory front pages extracted by our method. We evaluated our method and it attained 85.0% precision and 65.5% recall, respectively
  • 都道府県議会会議録を用いた地方政治コーパスの構築の試み
    田中琢真, 小林暁雄, 坂地泰紀, 内田ゆず, 乙武北斗, 高丸圭一, 木村泰知 情報処理北海道シンポジウム 2016 131 -132 2016年10月 [査読無し][通常論文]
  • 田中瑞竜, 酒井浩之, 坂地泰紀 電子情報通信学会技術研究報告 116 (213(NLC2016 13-28)) 19‐24 2016年09月01日 [査読無し][通常論文]
  • 吉岡晋作, 坂地泰紀, 酒井浩之 電子情報通信学会技術研究報告 115 (445(NLC2015 44-53)) 1‐5 2016年01月28日 [査読無し][通常論文]
  • 坂地泰紀, 所佳祐, 酒井浩之 人工知能学会全国大会論文集(CD-ROM) 30th ROMBUNNO.1H2‐4 2016年 [査読無し][通常論文]
  • 北森詩織, 酒井浩之, 坂地泰紀 言語処理学会年次大会発表論文集(Web) 22nd C3‐5 (WEB ONLY) 2016年 [査読無し][通常論文]
  • 坂地泰紀, 小花聖輝, 小林暁雄, 酒井浩之 ファジィシステムシンポジウム講演論文集(CD-ROM) 32nd ROMBUNNO.WD3‐3 2016年 [査読無し][通常論文]
  • 田中 琢真, 小林 暁雄, 坂地 泰紀, 内田 ゆず, 乙武 北斗, 高丸 圭一, 木村 泰知 日本知能情報ファジィ学会 ファジィ システム シンポジウム 講演論文集 32 (0) 251 -254 2016年 [査読無し][通常論文]
     

  • 宮崎敦也, 酒井浩之, 坂地泰紀 電子情報通信学会技術研究報告 115 (222(NLC2015 17-33)) 37 -41 2015年09月03日 [査読無し][通常論文]
  • 坂地泰紀, 酒井浩之 電子情報通信学会技術研究報告 115 (222(NLC2015 17-33)) 31 -35 2015年09月03日 [査読無し][通常論文]
  • Akihito Ikeda, Akio Kobayashi, Hiroki Sakaji, Shigeru Masuyama PROCEEDINGS 2015 18TH INTERNATIONAL CONFERENCE ON NETWORK-BASED INFORMATION SYSTEMS (NBIS 2015) 673 -678 2015年 [査読有り][通常論文]
     
    Online videos have become indispensable to people's daily lives. Everyday, millions of people watch online videos at video websites such as Nico Nico Douga and they comment on these websites. In Nico Nico Douga, comments are useful for online advertisements and video retrieval methods. Therefore, it is important to annotate comments in detail for them. In this paper, we annotate comments based on referring contents automatically.
  • 坂地 泰紀, 酒井 浩之, 増山 繁 成蹊大学理工学研究報告 51 (2) 23 -27 2014年12月 
    This paper proposes a method that extracts causal knowledge from news paper articles via clue expressions. Our method decides whether a sentence includes causal knowledge or not when the method extracts it. Therefore, our method can extract causal knowledge accurately. Furthermore, the advantage of our decision method is to extract causal knowledge manually without dictionaries and patterns.
  • 中山 大, 坂地 泰紀, 勝田 研一郎 成蹊大学理工学研究報告 51 (2) 53 -60 2014年12月 [査読無し][通常論文]
     
    This paper proposes a method of extracting important articles that influence the stock price of companies from financial articles. In the first step, our method obtains dates that have large difference from previous day at stock price of a selected company. In the second step, our method acquires articles that are published around the obtained dates and concern the selected company from financial articles. In the third step, our method extracts articles that influence the stock price of the selected company by using SVM as a machine learning method from the acquired articles. Finally, we evaluated our method. As a result, our method achieved 69.8% precision and 53.1% recall.
  • 石淵 準也, 坂地 泰紀, 酒井 浩之 電子情報通信学会技術研究報告 = IEICE technical report : 信学技報 114 (211) 17 -21 2014年09月11日 [査読無し][通常論文]
     
    本研究では,ニコニコ動画サイトに投稿された動画のコメントから,例えば「もっと評価されるべき」などのようなポジティブなコメントを自動的に抽出することを行った.ポジティブなコメントを自動的に抽出できることにより,例えば政見放送へのコメントの中からポジティブなコメントを得ることができ,政見放送の反応が即時に分析できるようになる.
  • 酒井 浩之, 坂地 泰紀 電子情報通信学会技術研究報告 = IEICE technical report : 信学技報 114 (211) 41 -45 2014年09月11日 [査読無し][通常論文]
     
    本論文では,企業Webページを対象とした企業検索システムにおいて,検索クエリに関連したタグを自動的に推定し,検索された企業ごとに推定されたタグを付与する手法を提案する.例えば検索クエリが「テキストマイニング」であれば,「開発」「提供」「導入」などのタグを推定する.そして,推定したタグを検索された企業に付与することで,例えば「テキストマイニング」技術を「開発」している企業を検索することができるようになる.
  • 池田 晃人, 小林 暁雄, 坂地 泰紀 電子情報通信学会技術研究報告 = IEICE technical report : 信学技報 114 (211) 47 -52 2014年09月11日 [査読無し][通常論文]
     
    日本の動画共有サイトであるニコニコ動画の最大の特徴は投稿コメントである.この投稿コメントは,再生中の動画上の任意の時間,場所に重ねて表示することができ,それがニコニコ動画の動画をより一層ユーザにとって楽しいもの,価値のあるものにしている.ニコニコ動画のコメントは,その動画の1つの特徴であるといえる.動画の特徴であるコメントの情報を用いることで,動画推薦や動画検索などの様々な応用が考えられる.本研究では,ニコニコ動画のコメントから情報を得ることを容易にするために,コメントにアノテーションを行う.ここでは,アノテーションを行うのための分類を行う.
  • 柴原 啓太, 坂地 泰紀, 酒井 浩之 電子情報通信学会技術研究報告 = IEICE technical report : 信学技報 114 (211) 75 -79 2014年09月11日 [査読無し][通常論文]
     
    近年,インターネットの発達により,動画共有サービスや動画配信サービスなどの動画サイトが普及している.本研究では,動画サイトの一つであるニコニコ動画を対象に,動画に付与されているタグ情報を用いて,動画に最適な日経プレスリリース記事を関連付けることで,効果的かつ効率的なインターネット広告の配信を目的とする.
  • 田原 如菜, 坂地 泰紀, 酒井 浩之 電子情報通信学会技術研究報告 = IEICE technical report : 信学技報 113 (429) 5 -10 2014年02月06日 [査読無し][通常論文]
     
    本研究では,キャラクターに対応する印象表現をTwitterから抽出する手法を提案する.具体的には,キャラクターとして,ゆるキャラを対象とし,キャラクター名に対応する印象表現(「可愛い」や「癒される」等)を抽出する.人手で印象表現辞書(喜,怒,哀,怖,恥,好,嫌,昂,驚,安,楽,その他のカテゴリーから成る363表現)を作成し,キャラクター名と印象表現辞書に存在する印象表現が含まれているツイートを取得する.そして,キャラクターごとに印象表現を抽出し,実際にアンケートで印象調査した結果と,本手法で抽出した印象表現を比べて評価した.
  • 田原 如菜, 坂地 泰紀, 酒井 浩之 電子情報通信学会技術研究報告. NLC, 言語理解とコミュニケーション 113 (429) 5 -10 2014年01月30日 [査読無し][通常論文]
     
    本研究では,キャラクターに対応する印象表現をTwitterから抽出する手法を提案する.具体的には,キャラクターとして,ゆるキャラを対象とし,キャラクター名に対応する印象表現(「可愛い」や「癒される」等)を抽出する.人手で印象表現辞書(喜,怒,哀,怖,恥,好,嫌,昂,驚,安,楽,その他のカテゴリーから成る363表現)を作成し,キャラクター名と印象表現辞書に存在する印象表現が含まれているツイートを取得する.そして,キャラクターごとに印象表現を抽出し,実際にアンケートで印象調査した結果と,本手法で抽出した印象表現を比べて評価した.
  • 坂地 泰紀, 酒井 浩之, 増山 繁 人工知能学会全国大会論文集 28 1 -4 2014年 [査読無し][通常論文]
  • 坂地 泰紀, 増山 繁 電子情報通信学会技術研究報告. NLC, 言語理解とコミュニケーション 111 (119) 7 -10 2011年06月30日 [査読無し][通常論文]
     
    本研究では,手がかり表現を用いて,タグの付されていない新聞記事から因果関係を表す表現(結果と原因の組)を抽出する手法を提案する.本手法は,因果関係を表す表現を抽出する際に,文が因果関係を含んでいるか否かを判定し,因果関係を含んでいると判定された文から因果関係を表す表現を抽出することで,精度の向上を図っている.さらに,判定手法は人手により作られた辞書やパターンを用いず,自動的に因果関係が含まれているか否かを判定することができる.
  • 坂地 泰紀, 増山 繁, 酒井 浩之 情報処理学会研究報告 2010 (2) 4p 2010年08月 [査読無し][通常論文]
  • 坂地 泰紀, 増山 繁, 酒井 浩之 電子情報通信学会技術研究報告. NLC, 言語理解とコミュニケーション 110 (142) 47 -50 2010年07月15日 [査読無し][通常論文]
     
    本論文では,新聞記事中の文が因果関係を含むか否かの判定を行う手法を提案する.手がかり表現を伴う因果関係に限定し,有無の判定を行う.因果関係を示す手がかり表現は,因果関係以外の意味を持っている場合がある.例えば,手がかり表現「から」は因果関係以外に,出発する位置を表す意味があり,「京都から大阪」などには因果関係は含まれない.このような因果関係を示す手がかり表現が含まれていても,因果関係以外の意味であるものを除くために,本手法を提案する.意味的な素性として拡張言語オントロジーを用いることと,学習データを自動的に増やすことにより,本手法の性能が向上した.
  • 坂地泰紀 言語処理学会第16回年次大会, 2010 371 -374 2010年
  • 野中 尋史, 小林 暁雄, 坂地 泰紀 人工知能学会全国大会論文集 24 1 -4 2010年 [査読無し][通常論文]
  • 坂地 泰紀, 野中 尋史, 酒井 浩之 電子情報通信学会技術研究報告 109 (142) 85 -92 2009年07月22日 [査読無し][通常論文]
  • 坂地 泰紀, 野中 尋史, 酒井 浩之, 増山 繁 情報処理学会研究報告. 自然言語処理研究会報告 2009 (14) 1 -8 2009年07月15日 [査読無し][通常論文]
     
    特許文書から直接的なユーザの便益に相当する表現と,技術上の解決課題を示す表現を自動的に抽出するアルゴリズム 「Cross-Bootstrapping」 を提案する.抽出した直接的なユーザの便益に相当する表現と,技術上の解決課題を示す表現はパテントマップを生成するために役立つ.本手法は,二つの手がかりと統計情報を用いて,ブートストラップ的に表現対を抽出する.また,辞書や人手により作成したパターンを用いず,自動的に表現を抽出することができる.最後に本手法の評価実験を行い,パテントマップを自動生成するために,十分な性能を達成したことを確認した.
  • 坂地 泰紀, 酒井 浩之, 増山 繁 日本オペレーションズ・リサーチ学会秋季研究発表会アブストラクト集 2007 194 -195 2007年09月27日 [査読無し][通常論文]
  • 坂地 泰紀, 酒井 浩之, 増山 繁 情報処理学会研究報告. 自然言語処理研究会報告 2007 (76) 151 -156 2007年07月24日 [査読無し][通常論文]
     
    本研究では,新聞記事から景気動向を示す「根拠となる表現」を統計的手法を用いて自動的に抽出する手法を提案する.また,抽出された景気動向を示す「根拠となる表現」を景気が回復することを示すPositive表現と悪化することを示すNegative表現に分類する手法も併せて提案する.企業や投資家にとって,株価や商品の売行きを予測するために,景気動向を知ることは重要なことである.そこで,我々は景気動向に関する記事から景気動向を示す「根拠となる表現」を抽出し,それを用いることにより,景気動向の予測が可能ではないかと考えた.今回,景気動向を予測するための材料として,景気動向の根拠となる表現を抽出し,それが景気が回復することを示すのか,悪化することを示すのかを判定する.我々は1990年から2005年の日経新聞を用いて実験し,景気動向を示す「根拠となる表現」の抽出手法と分類手法をそれぞれ評価した.その結果,景気動向を示す「根拠となる表現」の抽出手法に関しては適合率が71.43%,再現率が33.33%となり,分類手法に関してはPositive表現分類のF値が0.695, Negative表現分類のF値が0.849となった.
  • 坂地 泰紀, 酒井 浩之, 増山 繁 電子情報通信学会技術研究報告. NLC, 言語理解とコミュニケーション 107 (158) 151 -156 2007年07月17日 [査読無し][通常論文]
     
    本研究では,新聞記事から景気動向を示す「根拠となる表現」を統計的手法を用いて自動的に抽出する手法を提案する.また,抽出された景気動向を示す「根拠となる表現」を景気が回復することを示すPositive表現と悪化することを示すNegative表現に分類する手法も併せて提案する.企業や投資家にとって,株価や商品の売行きを予測するために,景気動向を知ることは重要なことである.そこで,我々は景気動向に関する記事から景気動向を示す「根拠となる表現」を抽出し,それを用いることにより,景気動向の予測が可能ではないかと考えた.今回,景気動向を予測するための材料として,景気動向の根拠となる表現を抽出し,それが景気が回復することを示すのか,悪化することを示すのかを判定する.我々は1990年から2005年の日経新聞を用いて実験し,景気動向を示す「根拠となる表現」の抽出手法と分類手法をそれぞれ評価した.その結果,景気動向を示す「根拠となる表現」の抽出手法に関しては適合率が71.43%,再現率が33.33%となり,分類手法に関してはPositive表現分類のF値が0.695,Negative表現分類のF値が0.849となった.

特許

  • 特許6615392:情報処理装置、情報処理方法、およびプログラム  
    田中良典, 佐久間洋明, 佐藤史仁, 小寺俊哉, 和泉潔, 坂地泰紀
  • 特許6596565:抽出システムおよびプログラム  
    平野正徳, 坂地泰紀, 松島裕康, 和泉潔, 加藤惇雄, 森岡嗣人, 長尾慎太郎, 木村笙子

受賞

  • 2023年07月 14th IIAI International Congress on Advanced Applied Informatics Competitive Paper Award
     Policy Gradient Stock GAN for Realistic Discrete Order Data Generation in Financial Markets 
    受賞者: Masanori Hirano;Hiroki Sakaji;Kiyoshi Izumi
  • 2022年07月 10th International Congress on Advanced Applied Informatics Outstanding Paper Award
     Investment Strategy via Lead Lag Effect using Economic Causal Chain and SSESTM Model 
    受賞者: Kei Nakagawa;Shingo Sashida;Hiroki Sakaji
  • 2022年05月 The 2nd Workshop on Financial Technology on the Web Best Paper Award
     Graph Representation Learning of Banking TransactionNetwork with Edge Weight-Enhanced Attention and Textual Information 
    受賞者: Naoto Minakawa;Kiyoshi Izumi;Hiroki Sakaji;Hitomi Sano
  • 2022年 人工知能学会 研究会優秀賞
     因果情報を用いた経済数値予測 
    受賞者: 和泉潔;坂地泰紀;佐野仁美
  • 2020年 人工知能学会 全国大会優秀賞
     STBM+: Advanced Stochastic Trading Behavior Model for Financial Markets based on Residual Blocks or Transformers 
    受賞者: Masanori Hirano;Kiyoshi Izumi;Hiroki Sakaji
  • 2019年08月 The First Workshop on Financial Technology and Natural Language Processing (FinNLP) Best Paper Award
     Economic Causal-Chain Search using Text Mining Technology 
    受賞者: Kiyoshi Izumi;Hiroki Sakaji
  • 2019年 人工知能学会 研究会優秀賞
     高頻度電力需要データを用いた製造業活動のナウキャスティングモデルの構築 
    受賞者: 水門善之;和泉潔;坂地泰紀;島田尚;松島裕康
  • 2019年 人工知能学会 全国大会優秀賞
     経済因果チェーン検索のシステム紹介と応用 
    受賞者: 和泉潔;坂地泰紀
  • 2017年12月 IDRユーザフォーラム2017 企業賞「Sansan株式会社」
     動画サイトのコメントを用いたタグ推定に関する研究 
    受賞者: 坂地泰紀;小林暁雄;小花聖輝
  • 2014年10月 金融情報学研究会 優秀論文賞
     
    受賞者: 坂地 泰紀

共同研究・競争的資金等の研究課題

  • AI農業社会実装プロジェクト
    農林水産省:農林水産省における研究開発とSociety5.0 との橋渡しプログラム(BRIDGE)
    研究期間 : 2023年09月 -2026年03月
  • 因果情報を用いた経済ナラティブシミュレーション
    戦略的創造研究推進事業:さきがけ
    研究期間 : 2022年10月 -2026年03月 
    代表者 : 坂地 泰紀
  • 日本学術振興会:科学研究費助成事業 基盤研究(C)
    研究期間 : 2021年04月 -2024年03月 
    代表者 : 坂地 泰紀
     
    本年度は、まず、因果関係インスタンスを取得するために、決算短信に対してタグ付与を行った。その後、タグ付けを行った決算短信データ、タグ付与済みの英語ロイターニュース記事、FinCausalのデータセットを用いて実験を行い、日英の文書から因果関係インスタンスを抽出できる手法の開発に取り組んだ。結果的に、BERTとグラフニューラルネットワークを組み合わせることで既存の手法よりも高い精度で因果関係インスタンスを抽出できる手法の開発に成功した。具体的には、全てのデータセットにおいて、F値が0.75以上で因果関係インスタンスを抽出することができた。この結果を論文としてまとめて国際会議に投稿したが、残念ながら不採録となった。そのため、2022年度は論文のブラッシュアップを行い、再度、投稿を行う。 因果関係インスタンスを抽出するためにドメイン特化のBERTモデルの構築も行った。モデル 構築にあたり、グラフィックカードであるNvidiaのV100を購入予定であったが、V100よりも価格が安いうえに性能が高いNvidiaのA6000が発売されていたことから、こちらを2個購入し、モデル構築や実験に利用した。 作成した事前学習モデルは、Web上で公開しており、誰でも無料で利用可能となっている。全ての公開したモデルのダウンロード数を合わせると、現時点で6,700件以上あり、多くの方に利用して頂いている。こちらの研究に関しては、SIGFINなどの国内研究会で発表済みである。こちらの研究に関しては、SIGFINなどの国内研究会で発表済みである。
  • 日本学術振興会:科学研究費助成事業 基盤研究(B)
    研究期間 : 2020年04月 -2024年03月 
    代表者 : 早矢仕 晃章, 清水 たくみ, 松島 裕康, 深見 嘉明, 坂地 泰紀
     
    本研究の目的は、データ流通市場の健全な発展を目指し、マーケットにおける異種のデータ・法規制・ヒトの動的な相互作用の解明とそれに基づくデータ流通の制度設計である。 2年目となる本年度は、昨年度整理したデータ流通市場の構成要素とその関係性を検討し、市場シミュレータ開発のためのモデル構築に取り組んだ。また、IEEE International Conference on Big Data 2021にて国際ワークショップCross-disciplinary Data Exchange and Collaborationを企画・運営し、国際的なコミュニティ作りにも注力した。 はじめに、(1)行政情報システム研究所で公開しているメタデータサイト及び、(2)データ分析コンペティションサービスKaggleの2つのデータ交換プラットフォームを分析した。(1)では組み合わせによって価値を持つデータのモデル化の手掛かりを得た。(2)では、COVID-19関連データを対象とし、データ公開初期のユーザ行動がコミュニティ形成とデータ利活用を促進することが分かった。 続いて、Kaggleデータセットを昨年度開発したデータ類似性計算手法に適用した。その結果、プラットフォームに依らない構造的特徴の一部が明らかとなった。 これらの解析から得られたデータとユーザの特徴から複数のモデルを作成し、データ生成器とデータ市場シミュレータを試作した。これにより、観測可能な相互作用が限定されるデータ流通市場の理解に向けた次年度の進展が期待できる。 また、近年注目されるIDアーキテクチャの分散化がデータ流通にもたらす影響と効率性を評価するために、分散型と集中型IDアーキテクチャを分析した。今後は、データ流通のインセンティブや導入コストの観点から、マルチエージェントシミュレーションに向けたパラメータを整理と検証を進めていく。
  • データ取引の市場設計:被験者実験による研究
    関西大学ソシオネットワーク戦略研究機構:
    研究期間 : 2023年04月 -2024年03月
  • データ流通市場の設計:新しい情報技術の人工市場における実験研究
    関西大学ソシオネットワーク戦略研究機構:
    研究期間 : 2021年04月 -2023年03月 
    代表者 : 渡邊 直樹, 難波 敏彦, 早矢仕 晃章, 坂地 泰紀, 松島 裕康, 清水 たくみ, 深見 嘉明
  • 日本学術振興会:科学研究費助成事業 基盤研究(B)
    研究期間 : 2016年04月 -2020年03月 
    代表者 : 木村 泰知, 川浦 昭彦, 乙武 北斗, 荒木 健治, 河村 和徳, 高丸 圭一, 湯淺 墾道, 吉田 光男, 坂地 泰紀, 内田 ゆず, 渋木 英潔, 小林 暁雄
     
    本研究では、地方政治に関する研究の活性化・学際的応用を目指して、「議論の背景」「議論の過程」「議論の結果」を関連づけるコーパスを構築する。具体的には、新聞記事(議論の背景)、地方議会会議録(議論の過程)、条例(議論の結果)の3つの言語資源を地域、時間、課題の観点から関連づけることを目指す。本研究の目的は以下の3点である。 A)3つの言語資源(地方議会会議録・条例・新聞記事)の収集・整理 B)地域・時間・課題などの観点から3つの言語資源を関連づけるコーパスの構築 C)上記の成果を用いた、政治学・経済学・言語学・情報工学における応用研究
  • 日本学術振興会:科学研究費助成事業 基盤研究(C)
    研究期間 : 2015年04月 -2019年03月 
    代表者 : 酒井 浩之, 増山 繁, 坂地 泰紀
     
    本研究では,決算短信や企業WEBサイトから,企業の事業に関連する情報を自動的に抽出し,その情報を検索対象とした企業検索システムを開発することを目的とした.研究成果として,決算短信から業績要因を抽出する手法,因果関係(原因と結果)を抽出する手法,業績予測文を抽出する手法,関連企業を推定する手法等を開発した.また,本手法により抽出された業績要因等を検索対象とした4つの企業検索システムを開発した.開発した企業検索システムは,WEBサイト(http://www.ci.seikei.ac.jp/sakai/)にて公開されている.

大学運営

委員歴

  • 2023年 - 現在   Special Session on Applied Informatics in Finance and Economics (AIFE)   Organizer
  • 2023年 - 現在   IEEE Computational Finance and Economics   Technical Committee Member
  • 2023年 - 現在   人工知能学会 第二種研究会 データ指向構成マイニングとシミュレーション研究会 (DOCMAS)   幹事
  • 2023年 - 現在   情報処理学会 編集委員会「知能グループ」   編集委員
  • 2023年 - 現在   言語処理学会 会誌編集委員会   編集委員
  • 2022年 - 現在   電子情報通信学会 言語理解とコミュニケーション研究会   副委員長
  • 2022年 - 現在   人工知能学会 第二種研究会 金融情報学研究会   主幹事
  • 2022年 - 現在   IEEE International Conference on Big Data   Program Committee
  • 2021年 - 現在   International Workshop on Financial Technology on the Web (FinWeb)   Program Committee
  • 2021年 - 現在   人工知能学会 編集委員会   編集委員
  • 2021年 - 現在   AAAI Conference on Artificial Intelligence   Program Committee
  • 2020年 - 現在   International Workshop on Financial Technology and Natural Language Processing (FinNLP)   Program Committee
  • 2015年 - 現在   International Workshop on Web Services and Social Media (WSSM)   Program Committee
  • 2021年 - 2023年03月   言語処理学会 言語処理学会年次大会   プログラム委員
  • 2023年 - 2023年   Special Session on Understanding New Markets by Data Science, Social Science, and Economics   Organizer
  • 2023年 - 2023年   The 61st Annual Meeting of the Association for Computational Linguistics (ACL2023)   Area Chair
  • 2021年 - 2021年   人工知能学会 特集:「ファイナンスにおける人工知能応用」   ゲストエディタ
  • 2021年 - 2021年   International Workshop on Economics and Natural Language Processing (ECONLP)   Program Committee
  • 2021年 - 2021年   The 29th International Joint Conference on Artificial Intelligence (IJCAI)   Technical Session Chair
  • 2020年 - 2021年   電子情報通信学会 言語理解とコミュニケーション研究会   幹事
  • 2018年 - 2021年   電子情報通信学会 情報・システムソサイエティ学術奨励賞選定委員会   投票委員
  • 2018年 - 2021年   人工知能学会 第二種研究会 金融情報学研究会 (SIG-FIN)   幹事
  • 2018年 - 2021年   International Workshop on Cross-disciplinary Data Exchange and Collaboration (CDEC)   Program Committee
  • 2020年 - 2020年   The 9th International Congress on Advanced Applied Informatics   Program Commitee
  • 2019年 - 2020年   NLP若手の会運営委員会   運営委員
  • 2018年 - 2020年   電子情報通信学会 情報・システムソサイエティ誌編集委員会   編集委員
  • 2018年 - 2020年   電子情報通信学会 言語理解とコミュニケーション研究会   幹事補佐
  • 2019年 - 2019年   電子情報通信学会総合大会 情報・システム   編集委員
  • 2019年 - 2019年   The 33rd International Conference on Advanced Information Networking and Applications (AINA-2019)   Program Committee
  • 2019年 - 2019年   The 7th International Conference on Smart Computing and Artificial Intelligence (SCAI 2019)   Special Session Organizer
  • 2019年 - 2019年   International Conference on Smart Computing and Artificial Intelligence (SCAI)   Program Committee
  • 2019年 - 2019年   The 14th International Conference on P2P, Parallel, Grid, Cloud and Internet Computing (3PGCIC)   Program Committee
  • 2018年 - 2018年   The 21st International Conference on Principles and Practice of Multi-Agent Systems (PRIMA2018)   Workshop/Tutorial Chairs
  • 2017年 - 2017年   電子情報通信学会 言語理解とコミュニケーション研究会   専門委員


Copyright © MEDIA FUSION Co.,Ltd. All rights reserved.