[3C3-GS-6] Language media processing: generation

[3C3-GS-6-01] A Study on Topic-Based Abstract Summarization Generation

〇Yuka Yokogawa¹, Ichiro Kobayashi¹ (1. Ochanomizu University)

Keywords:summarization, topic, Transformer

The objective of this research is to generate a summary of the contents of a document based on various topics in the text. As a method for summarizing sentences, we adopt the encoder part of PreSumm, which generates an abstractive summary with Transformer, and adopt the general-purpose language model GPT2 as a decoder. We also employ the Plug and Play Language Model to generate a topic-based summary, introducing particular topics into a generated summary.
Through experiments, we confirmed that our proposed method can make a summary reflecting provided topics.