• Search Research Projects
  • Search Researchers
  • How to Use
  1. Back to project page

2021 Fiscal Year Final Research Report

Sentence translation mechanism equipped with an explainable process based on real-world and linguistic knowledge

Research Project

  • PDF
Project/Area Number 19H04162
Research Category

Grant-in-Aid for Scientific Research (B)

Allocation TypeSingle-year Grants
Section一般
Review Section Basic Section 61030:Intelligent informatics-related
Research InstitutionTohoku University

Principal Investigator

Suzuki Jun  東北大学, データ駆動科学・AI教育研究センター, 教授 (80396150)

Co-Investigator(Kenkyū-buntansha) 松林 優一郎  東北大学, 教育学研究科, 准教授 (20582901)
乾 健太郎  東北大学, 情報科学研究科, 教授 (60272689)
赤間 怜奈  東北大学, データ駆動科学・AI教育研究センター, 助教 (70912533)
Project Period (FY) 2019-04-01 – 2022-03-31
Keywords自然言語処理 / 人工知能 / 機械学習 / 文章生成
Outline of Final Research Achievements

Realizing a text generation mechanism at the human level is one of the most critical and challenging unsolved problems in artificial intelligence and natural language processing research. While current methods based on deep neural networks can generate fluent sentences, a new problem, the erroneous generation problem, has been pointed out.
In this study, to solve the erroneous generation problem and reveal the mechanism of its occurrence, we have developed several methods, such as an example-based method to detect reasons for erroneous generation, and an improved natural language analysis system for analyzing erroneous generations.

Free Research Field

自然言語処理

Academic Significance and Societal Importance of the Research Achievements

文章生成における誤生成問題は一見小さな技術的課題のように思えるが,誹謗中傷,差別,卑猥な表現と捉えられるような,状況的,或いは,社会通念的に不適切な文章を実用システムが一度でも誤生成してしまうと致命的な社会問題となりえるという潜在的リスクがある.このことから,誤生成問題の解決に向けた研究成果は社会的に大きな意義がある.また,学術的にも,人間と同等レベルの文章生成技術の実現は,自然言語処理研究の古くからの最難関未解決課題の一つであり,達成に向けた研究成果の意義は高い.

URL: 

Published: 2023-01-30  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi