Topic sequence kernel

Jian Xu; Qin Lu; Zhengzhong Liu; Junyi Chai

doi:10.1007/978-3-642-35341-3_41

Topic sequence kernel

Jian Xu, Qin Lu, Zhengzhong Liu, Junyi Chai

Research output: Chapter in book / Conference proceeding › Conference article published in proceeding or book › Academic research › peer-review

1 Citation (Scopus)

Abstract

This paper addresses the problem of classifying documents using the kernel approaches based on topic sequences. Previously, the string kernel uses the ordered subsequence of characters as features and the word sequence kernel is proposed to use words as the subsequences. However, they both face the problem of computational complexity because of the large amount of symbols (characters or words). This paper, therefore, proposes to use sequences of topics rather than characters or words to reduce the number of symbols, thus increasing the computational efficiency. Documents that exhibit similar posterior topic proportions are expected to have similar topic sequence and then should be classified into the same category. Experiments conducted on the Reuters-21578 datasets have proven this hypothesis.

Original language	English
Title of host publication	Information Retrieval Technology - 8th Asia Information Retrieval Societies Conference, AIRS 2012, Proceedings
Pages	457-466
Number of pages	10
DOIs	https://doi.org/10.1007/978-3-642-35341-3_41
Publication status	Published - 31 Dec 2012
Event	8th Asia Information Retrieval Societies Conference, AIRS 2012 - Tianjin, China Duration: 17 Dec 2012 → 19 Dec 2012

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	7675 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	8th Asia Information Retrieval Societies Conference, AIRS 2012
Country/Territory	China
City	Tianjin
Period	17/12/12 → 19/12/12

Keywords

Classification
String kernel
Topic sequence

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

More information

10.1007/978-3-642-35341-3_41

Cite this

Xu, J., Lu, Q., Liu, Z., & Chai, J. (2012). Topic sequence kernel. In Information Retrieval Technology - 8th Asia Information Retrieval Societies Conference, AIRS 2012, Proceedings (pp. 457-466). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7675 LNCS). https://doi.org/10.1007/978-3-642-35341-3_41

@inproceedings{0a33de0d84fb45b0bf5fd6dde4eb3532,

title = "Topic sequence kernel",

abstract = "This paper addresses the problem of classifying documents using the kernel approaches based on topic sequences. Previously, the string kernel uses the ordered subsequence of characters as features and the word sequence kernel is proposed to use words as the subsequences. However, they both face the problem of computational complexity because of the large amount of symbols (characters or words). This paper, therefore, proposes to use sequences of topics rather than characters or words to reduce the number of symbols, thus increasing the computational efficiency. Documents that exhibit similar posterior topic proportions are expected to have similar topic sequence and then should be classified into the same category. Experiments conducted on the Reuters-21578 datasets have proven this hypothesis.",

keywords = "Classification, String kernel, Topic sequence",

author = "Jian Xu and Qin Lu and Zhengzhong Liu and Junyi Chai",

year = "2012",

month = dec,

day = "31",

doi = "10.1007/978-3-642-35341-3_41",

language = "English",

isbn = "9783642353406",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

pages = "457--466",

booktitle = "Information Retrieval Technology - 8th Asia Information Retrieval Societies Conference, AIRS 2012, Proceedings",

note = "8th Asia Information Retrieval Societies Conference, AIRS 2012 ; Conference date: 17-12-2012 Through 19-12-2012",

}

Xu, J, Lu, Q, Liu, Z & Chai, J 2012, Topic sequence kernel. in Information Retrieval Technology - 8th Asia Information Retrieval Societies Conference, AIRS 2012, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 7675 LNCS, pp. 457-466, 8th Asia Information Retrieval Societies Conference, AIRS 2012, Tianjin, China, 17/12/12. https://doi.org/10.1007/978-3-642-35341-3_41

Topic sequence kernel. / Xu, Jian; Lu, Qin; Liu, Zhengzhong et al.
Information Retrieval Technology - 8th Asia Information Retrieval Societies Conference, AIRS 2012, Proceedings. 2012. p. 457-466 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7675 LNCS).

Research output: Chapter in book / Conference proceeding › Conference article published in proceeding or book › Academic research › peer-review

TY - GEN

T1 - Topic sequence kernel

AU - Xu, Jian

AU - Lu, Qin

AU - Liu, Zhengzhong

AU - Chai, Junyi

PY - 2012/12/31

Y1 - 2012/12/31

N2 - This paper addresses the problem of classifying documents using the kernel approaches based on topic sequences. Previously, the string kernel uses the ordered subsequence of characters as features and the word sequence kernel is proposed to use words as the subsequences. However, they both face the problem of computational complexity because of the large amount of symbols (characters or words). This paper, therefore, proposes to use sequences of topics rather than characters or words to reduce the number of symbols, thus increasing the computational efficiency. Documents that exhibit similar posterior topic proportions are expected to have similar topic sequence and then should be classified into the same category. Experiments conducted on the Reuters-21578 datasets have proven this hypothesis.

AB - This paper addresses the problem of classifying documents using the kernel approaches based on topic sequences. Previously, the string kernel uses the ordered subsequence of characters as features and the word sequence kernel is proposed to use words as the subsequences. However, they both face the problem of computational complexity because of the large amount of symbols (characters or words). This paper, therefore, proposes to use sequences of topics rather than characters or words to reduce the number of symbols, thus increasing the computational efficiency. Documents that exhibit similar posterior topic proportions are expected to have similar topic sequence and then should be classified into the same category. Experiments conducted on the Reuters-21578 datasets have proven this hypothesis.

KW - Classification

KW - String kernel

KW - Topic sequence

UR - http://www.scopus.com/inward/record.url?scp=84871566194&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-35341-3_41

DO - 10.1007/978-3-642-35341-3_41

M3 - Conference article published in proceeding or book

SN - 9783642353406

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 457

EP - 466

BT - Information Retrieval Technology - 8th Asia Information Retrieval Societies Conference, AIRS 2012, Proceedings

T2 - 8th Asia Information Retrieval Societies Conference, AIRS 2012

Y2 - 17 December 2012 through 19 December 2012

ER -

Topic sequence kernel

Abstract

Publication series

Conference

Keywords

ASJC Scopus subject areas

More information

Other files and links

Fingerprint

Cite this