TY - GEN
T1 - T4S
T2 - 7th International Conference on Cognitive Computing, ICCC 2023, Held as Part of the Services Conference Federation, SCF 2023
AU - Wang, Depei
AU - Sun, Wenyi
AU - Luo, Cheng
AU - Liu, Dachang
AU - Mao, Ruibin
AU - Xu, Ruifeng
N1 - Publisher Copyright:
© 2024, The Author(s), under exclusive license to Springer Nature Switzerland AG.
PY - 2024
Y1 - 2024
N2 - 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.
AB - 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.
KW - Abstractive Summarization
KW - Data Mining
KW - Natural Language Processing
KW - Semantic Matching
UR - https://www.scopus.com/pages/publications/85181983943
U2 - 10.1007/978-3-031-51671-9_6
DO - 10.1007/978-3-031-51671-9_6
M3 - 会议稿件
AN - SCOPUS:85181983943
SN - 9783031516702
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 75
EP - 86
BT - Cognitive Computing – ICCC 2023 - 7th International Conference Held as Part of the Services Conference Federation, SCF 2023, Proceedings
A2 - Pan, Xiuqin
A2 - Jin, Ting
A2 - Zhang, Liang-Jie
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 17 December 2023 through 18 December 2023
ER -