Development of writer support system using the user's intention extraction and automatic text generation

2022 Fiscal Year Final Research Report

• Project/Area Number: 20K19878
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 61030:Intelligent informatics-related
• Research Institution: National Institute of Advanced Industrial Science and Technology
• Principal Investigator: Watanabe Kento 国立研究開発法人産業技術総合研究所, 情報・人間工学領域, 研究員 (50828324)
• Project Period (FY): 2020-04-01 – 2023-03-31
• Keywords: 歌詞情報処理 / 自然言語処理 / 音楽情報処理 / ヒューマンコンピュータインタラクション / 創作支援

Outline of Final Research Achievements

We have developed a lyric writing support system to assist users who have a clear idea of the content they want to include in their lyrics, but are struggling to come up with appropriate phrases. This system generates and suggests lyrics that reflect the user's intentions, providing inspiration for the creative process. In developing this support system, we found that removing information related to the wording of the lyrics and focusing only on the meaning was effective in enabling the computer to grasp the intended message. To accomplish this, we investigated methods for removing wording information and discovered that by converting the user's input text into an image, we could extract only the intended meaning. By generating text from this image, we were able to improve the accuracy of the user's intent in the generated text. In addition, we developed a lyric generation method that reduces the risk of plagiarism, which further enhances the usefulness of the system.

Free Research Field

自然言語処理

Academic Significance and Societal Importance of the Research Achievements

本研究によって、画像を用いた情報抽出手法が歌詞生成の性能向上に有効であることも示され、自然言語処理の分野において、意味情報のみを取り出す手法の一つとして、有用な知見を提供することを意味する。また、本研究成果が音楽制作において、作詞者や音楽プロデューサーの創造性を支援することが期待される。特に、作詞に苦手意識を持つ人や初心者にとって、意図を反映した歌詞生成が手助けとなると考えられる。また、著作権侵害の問題が深刻化するGenerative AIにおいて、著作物の剽窃リスクを下げる歌詞生成手法が開発されたことは、AI業界と音楽業界にとっても大きな意義を持つ。