The study of artificial intelligence to generate documents

2022 Fiscal Year Final Research Report

• Project/Area Number: 20K11958
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 61030:Intelligent informatics-related
• Research Institution: Osaka Metropolitan University (2022); Osaka Prefecture University (2020-2021)
• Principal Investigator: Okada Makoto 大阪公立大学, 大学院情報学研究科, 助教 (40336813)
• Project Period (FY): 2020-04-01 – 2023-03-31
• Keywords: 深層学習 / 自然言語処理 / 文意一貫性 / 文書生成 / 意外性

Outline of Final Research Achievements

While attention has recently been focused on the automatic generation of data by artificial intelligence, the field of language processing has not achieved sufficient results compared to other media. I have developed a new method for estimating the emotion of dialogues in comic books, estimating paragraph boundaries in novels to explore the possibility of understanding the consistency of sentence meaning, and a method based on the Conditional Variational Autoencoder (CVAE), a kind of deep generation method, for the purpose of building an artificial intelligence that can freely generate sentences while cooperating with human beings. A method for bi-directional sentence generation from the beginning and the end of a sentence was proposed.

Free Research Field

自然言語処理，自然言語理解，知識処理，機械学習

Academic Significance and Societal Importance of the Research Achievements

人工知能によるデータの自動生成が申請の際よりも大きく注目が集まる中，他のメディアと比較すると十分な成果が得られていなかった言語処理分野において，他に先んじて人工知能の高度な言語能力の獲得を目標として「意味一貫性」と創造的意外性としての「意味破綻」の双方をコントロール可能かつ協調しながら自由に文生成可能な人工知能の構築を目的とした本研究のが深層生成手法の一種である Conditional Variational Autoencoder (CVAE) を基にした先頭と最後から双方向に文を生成する手法はなど提案できたことは一定の成果を残したといえる．