TOHOKU NLP LAB - 東北大学 乾研究室TOHOKU NLP LAB - 東北大学 乾研究室


Research Topics【研究内容】

Research Topics / 研究内容

The most important means for communication are the languages that we use everyday, like Japanese and English. In this lab, we conduct research in the following areas: (i) theoretical research to clarify and model the mechanism of communication, namely, what it means to understand language and the conditions that make communication possible (ii) Natural Language Processing research on the development of software that automatically processes the information and knowledge that is represented and transmitted in language; and (iii) applied research supporting intelligent communication or information analysis for the benefit of mankind. We approach an understanding of human intelligence from the perspective of natural language.

本研究室では、自然言語で表現され、伝達され、蓄積される情報や人の知識をコンピュータで処理するための基礎理論、基盤技術、応用技術に関する研究を行っています。自然言語処理(natural language processing)、人工知能(artificial intelligence)、知識情報処理(knowledge processing)、計算言語学(computational linguistics)、コミュニケーション科学(communication science)などの領域が我々のフィールドです。

Currently, we are in a period in which anyone can obtain and accumulate large quantities of information due to the widespread popularity of the Internet. However, due to the excessive amount of information that is widely spread on the Internet, often times it is difficult to successfully find desired information, recognize where important information exists, and to be able to trust every bit of information. Now, as most of this information is composed of natural language, there is a strong demand for Natural Language Processing (NLP). If computers were to automatically collect, sort, and analysis a large quantity of language information, as well as automatically translate other languages and interactively express information to individuals, our surrounding language environment would drastically change. Therefore, the importance of Web information analysis, communication support, and knowledge cycle via Natural Language Processing would rapidly increase.

In order to fulfill such an objective, it is absolutely necessary to develop technology which has the capability of understanding human language. Of course, this is not an easy objective. Luckily, NLP technologies have been steadily progressing towards this goal. For example, little by little, we are beginning to see signs of major breakthroughs due to the possibility of computers decisively being able to automatically acquire lacking world knowledge from a large amount of data and use for semantic analysis and inference.

In this laboratory, we expand upon this work by developing software which supports in theoretically solving communication structure and modeling, human being’s intelligent communication, and information analysis. We aim for the wisdom of individuals by words. If you are looking for exciting research, then this laboratory is perfect.




当研究室への配属を検討している学生さんは『東北大学 乾研究室への配属を検討しているみなさんへ』のページもご覧ください。

東北大学サイエンスカフェ YouTube動画「言葉がわかるコンピューターはどこまでできたか ~言葉の不思議と自然言語処理の最前線」(スライド), 2013年2月

Main Research Themes / 主な研究テーマ

From fundamental theory to core and applied technologies of NLP and AI, we are working on a wide range of research topics.

  1. Natural Language Processing Technology High Performance and Robustness
  2. Large-scale Knowledge Acquisition and Flexible Inference for Deep Language Understanding
  3. Artificial Intelligence based on Big Data and Machine Learning
  4. Analysis and Compilation of Web and Social Media by Natural Language Processing
  5. Support for Disaster-Related Information and Risk Information
  6. Multi-language Processing for Machine Translation, Translation Support, and Language Learning Support
  7. Intelligent Robot Dialogue via Verbal Information, Non-verbal Information, and Deep Inference Integration
  8. Mathematical Models for Language, Understanding, and Communication


  1. 自然言語処理技術の高度化と頑健化
  2. 深い言語理解のための大規模な知識獲得と柔軟な推論
  3. ビッグデータと機械学習に基づく人工知能
  4. 言語情報・非言語情報・深い推論の統合による知能ロボット対話
  5. 自然言語処理によるウェブ・ソーシャルメディアの分析と編集
  6. 災害関連情報・リスク情報のコミュニケーション支援
  7. 機械翻訳・翻訳支援・言語学習支援などの多言語処理
  8. 言語・意味・コミュニケーションの数理モデル

Research Themes of Members / メンバーの研究テーマ



Step-QI School / Step-QIスクール アドバンスト創造工学


NLP and AI Applications / 応用技術のテーマ例

With the rapid advancement in our current information society, the Internet, day by day, is advancing with an increase in both an enormous amount of information and knowledge. As information and knowledge on the Internet is often be scattered and can be difficult to conveniently find, it is extremely vital to discover such information and organize it in a way to conveniently provide it to its users. Within our laboratory, we aim for developing software technologies which can automatically assist in rapidly discovering such information and knowledge.


Analyzing and Editing Information of the Web and Social Media / ウェブ・ソーシャルメディアの分析と編集

Due to words and their role of being actively used in our society, our technology is also quickly spreading to be used actively in society as well. For example, on the Internet in which anyone has the ability to freely send information, at the same time existing information is being spread, unreliable information and false information is simultaneously being overwhelmingly spread. In regards to this information reliability problem, within this laboratory, we resolve various information sources automatically and analyze information from both sides which includes discovering agreeing and conflicting information with our Statement Map Project.



Supporting Crisis Information Management / 災害関連情報・リスク情報のコミュニケーション支援

We currently develop Natural Language Processing technologies for bringing together communication between a disaster area, supporters, and various administrations.



Text Mining and Opinion/Experience Mining / テキストマイニング、意見・経験情報マイニング

From a collection of the vast amount of sentences on the Web (blogs, etc), we extract information regarding individual’s opinions and personal experiences and, as structural information, apply it to a database in order to determine a common ground between various individual’s personal experiences and knowledge and effectively take advantage of such information.


Natural Language Dialogue Systems / ロボット/エージェント対話

With the advancements of rapid language processing technologies and its information access, robots with the ability to communicate with humans and other software agents have been actively explored. For example, when using technology which is able to infer emotion by an individual’s utterance, appropriate responses, such as “That sounds fun” in response to “I went to Disney Land today” and “Are you all right?” in response to “I lost my wallet” can be created. When using information access technology, given a sentence such as “I like Apple products”, product information regarding the company Apple can be discovered on the Web and utterances such as “The iPhone5s camera seems to be high quality” and “The iPad is even being used in classrooms” can be said which allow for interesting conversation.


Fundamental Technologies / 基盤技術

Syntactic and Semantic Parsing / 構文解析・意味解析


Discourse Analysis / 談話解析・文脈解析


Knowledge Acquisition from Large-scale Text Data / 大規模言語データからの知識獲得


Artificial Intelligence: Knowledge and Inference / 人工知能:知識と推論


Fundamental Theory / 基礎理論

Mathematical Models of Language and Communication / 言語の数理モデル

Constructing statistical models that capture the properties of language and the mechanism of communication can be an effective way to incorporate semantic language analysis in applications such as advanced language understanding and information analysis.


Last-modified: 2021-03-16 (Tue) 15:15:07 (625d)

Recent Changes