A topical VAEGAN-IHMM approach for automatic story segmentation

Feature representations with rich topic information can greatly improve the performance of story segmentation tasks. VAEGAN offers distinct advantages in feature learning by combining variational autoencoder (VAE) and generative adversarial network (GAN), which not only captures intricate data repre...

Full description

Saved in:
Bibliographic Details
Main Authors: Jia Yu, Huiling Peng, Guoqiang Wang, Nianfeng Shi
Format: Article
Language:English
Published: AIMS Press 2024-07-01
Series:Mathematical Biosciences and Engineering
Subjects:
Online Access:https://www.aimspress.com/article/doi/10.3934/mbe.2024289?viewType=HTML
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846163749260492800
author Jia Yu
Huiling Peng
Guoqiang Wang
Nianfeng Shi
author_facet Jia Yu
Huiling Peng
Guoqiang Wang
Nianfeng Shi
author_sort Jia Yu
collection DOAJ
description Feature representations with rich topic information can greatly improve the performance of story segmentation tasks. VAEGAN offers distinct advantages in feature learning by combining variational autoencoder (VAE) and generative adversarial network (GAN), which not only captures intricate data representations through VAE's probabilistic encoding and decoding mechanism but also enhances feature diversity and quality via GAN's adversarial training. To better learn topical domain representation, we used a topical classifier to supervise the training process of VAEGAN. Based on the learned feature, a segmentor splits the document into shorter ones with different topics. Hidden Markov model (HMM) is a popular approach for story segmentation, in which stories are viewed as instances of topics (hidden states). The number of states has to be set manually but it is often unknown in real scenarios. To solve this problem, we proposed an infinite HMM (IHMM) approach which utilized an HDP prior on transition matrices over countably infinite state spaces to automatically infer the state's number from the data. Given a running text, a Blocked Gibbis sampler labeled the states with topic classes. The position where the topic changes was a story boundary. Experimental results on the TDT2 corpus demonstrated that the proposed topical VAEGAN-IHMM approach was significantly better than the traditional HMM method in story segmentation tasks and achieved state-of-the-art performance.
format Article
id doaj-art-33e620b48815430991f415ce1142f7c0
institution Kabale University
issn 1551-0018
language English
publishDate 2024-07-01
publisher AIMS Press
record_format Article
series Mathematical Biosciences and Engineering
spelling doaj-art-33e620b48815430991f415ce1142f7c02024-11-19T01:27:16ZengAIMS PressMathematical Biosciences and Engineering1551-00182024-07-012176608663010.3934/mbe.2024289A topical VAEGAN-IHMM approach for automatic story segmentationJia Yu 0Huiling Peng1Guoqiang Wang 2Nianfeng Shi 31. School of Computer and Information Engineering, Luoyang Institute of Science and Technology, China 2. Software Research Institute, Technological University of Shannon, Ireland1. School of Computer and Information Engineering, Luoyang Institute of Science and Technology, China1. School of Computer and Information Engineering, Luoyang Institute of Science and Technology, China1. School of Computer and Information Engineering, Luoyang Institute of Science and Technology, ChinaFeature representations with rich topic information can greatly improve the performance of story segmentation tasks. VAEGAN offers distinct advantages in feature learning by combining variational autoencoder (VAE) and generative adversarial network (GAN), which not only captures intricate data representations through VAE's probabilistic encoding and decoding mechanism but also enhances feature diversity and quality via GAN's adversarial training. To better learn topical domain representation, we used a topical classifier to supervise the training process of VAEGAN. Based on the learned feature, a segmentor splits the document into shorter ones with different topics. Hidden Markov model (HMM) is a popular approach for story segmentation, in which stories are viewed as instances of topics (hidden states). The number of states has to be set manually but it is often unknown in real scenarios. To solve this problem, we proposed an infinite HMM (IHMM) approach which utilized an HDP prior on transition matrices over countably infinite state spaces to automatically infer the state's number from the data. Given a running text, a Blocked Gibbis sampler labeled the states with topic classes. The position where the topic changes was a story boundary. Experimental results on the TDT2 corpus demonstrated that the proposed topical VAEGAN-IHMM approach was significantly better than the traditional HMM method in story segmentation tasks and achieved state-of-the-art performance.https://www.aimspress.com/article/doi/10.3934/mbe.2024289?viewType=HTMLgenerative adversarial networkvariational autoencoderhdphidden markov modelstory segmentation
spellingShingle Jia Yu
Huiling Peng
Guoqiang Wang
Nianfeng Shi
A topical VAEGAN-IHMM approach for automatic story segmentation
Mathematical Biosciences and Engineering
generative adversarial network
variational autoencoder
hdp
hidden markov model
story segmentation
title A topical VAEGAN-IHMM approach for automatic story segmentation
title_full A topical VAEGAN-IHMM approach for automatic story segmentation
title_fullStr A topical VAEGAN-IHMM approach for automatic story segmentation
title_full_unstemmed A topical VAEGAN-IHMM approach for automatic story segmentation
title_short A topical VAEGAN-IHMM approach for automatic story segmentation
title_sort topical vaegan ihmm approach for automatic story segmentation
topic generative adversarial network
variational autoencoder
hdp
hidden markov model
story segmentation
url https://www.aimspress.com/article/doi/10.3934/mbe.2024289?viewType=HTML
work_keys_str_mv AT jiayu atopicalvaeganihmmapproachforautomaticstorysegmentation
AT huilingpeng atopicalvaeganihmmapproachforautomaticstorysegmentation
AT guoqiangwang atopicalvaeganihmmapproachforautomaticstorysegmentation
AT nianfengshi atopicalvaeganihmmapproachforautomaticstorysegmentation
AT jiayu topicalvaeganihmmapproachforautomaticstorysegmentation
AT huilingpeng topicalvaeganihmmapproachforautomaticstorysegmentation
AT guoqiangwang topicalvaeganihmmapproachforautomaticstorysegmentation
AT nianfengshi topicalvaeganihmmapproachforautomaticstorysegmentation