Skip to main navigation Skip to search Skip to main content

T4S: Two-Stage Screenplay Synopsis Summary Generation with Turning Points

  • Depei Wang
  • , Wenyi Sun
  • , Cheng Luo
  • , Dachang Liu
  • , Ruibin Mao
  • , Ruifeng Xu*
  • *Corresponding author for this work
  • Ltd.
  • Harbin Institute of Technology Shenzhen
  • Hubei University
  • Shenzhen Stock Exchange

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

The recent advancements in neural network models and the availability of vast amounts of data, automatic summarization technology has become one of the primary solutions for dealing with information overload and pinpointing key information. Unlike narrative text, movie scripts consist of sequences of scene descriptions, and directly compressing the text may lead to truncation of plot-relevant content. Furthermore, movie script summarization tasks lack datasets that align script content with plot summaries. Therefore, this paper proposes a two-stage method T4S for generating plot summaries of movie scripts. First, the GraphTP model is employed as an extractor to extract key turning scenes from the scene text sequences. Second, an unsupervised text matching method is used to obtain text pairs that match scenes with plot summaries. Finally, a generator, utilizing an efficiently fine-tuned LLM, rewrites key scene text and concatenates it to form the final plot summary of the movie script. The results of the implementation show that the proposed method in this paper outperforms baseline methods.

Original languageEnglish
Title of host publicationCognitive Computing – ICCC 2023 - 7th International Conference Held as Part of the Services Conference Federation, SCF 2023, Proceedings
EditorsXiuqin Pan, Ting Jin, Liang-Jie Zhang
PublisherSpringer Science and Business Media Deutschland GmbH
Pages75-86
Number of pages12
ISBN (Print)9783031516702
DOIs
StatePublished - 2024
Externally publishedYes
Event7th International Conference on Cognitive Computing, ICCC 2023, Held as Part of the Services Conference Federation, SCF 2023 - Shenzhen, China
Duration: 17 Dec 202318 Dec 2023

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume14207
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference7th International Conference on Cognitive Computing, ICCC 2023, Held as Part of the Services Conference Federation, SCF 2023
Country/TerritoryChina
CityShenzhen
Period17/12/2318/12/23

Keywords

  • Abstractive Summarization
  • Data Mining
  • Natural Language Processing
  • Semantic Matching

Fingerprint

Dive into the research topics of 'T4S: Two-Stage Screenplay Synopsis Summary Generation with Turning Points'. Together they form a unique fingerprint.

Cite this