Tree-structured hierarchical Dirichlet process

Md Hijbul Alam; Jaakko Peltonen; Jyrki Nummenmaa; Kalervo Järvelin

doi:10.1007/978-3-319-99608-0_33

Tree-structured hierarchical Dirichlet process

Md Hijbul Alam, Jaakko Peltonen^*, Jyrki Nummenmaa, Kalervo Järvelin

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Scientific › peer-review

1 Citation (Scopus)

Abstract

In many domains, document sets are hierarchically organized such as message forums having multiple levels of sections. Analysis of latent topics within such content is crucial for tasks like trend and user interest analysis. Nonparametric topic models are a powerful approach, but traditional Hierarchical Dirichlet Processes (HDPs) are unable to fully take into account topic sharing across deep hierarchical structure. We propose the Tree-structured Hierarchical Dirichlet Process, allowing Dirichlet process based topic modeling over a given tree structure of arbitrary size and height, where documents can arise at all tree nodes. Experiments on a hierarchical social message forum and a product reviews forum demonstrate better generalization performance than traditional HDPs in terms of ability to model new data and classify documents to sections.

Original language	English
Title of host publication	Distributed Computing and Artificial Intelligence, Special Sessions, 15th International Conference
Editors	Sara Rodríguez, Javier Prieto, Pedro Faria, María N. Moreno, Santiago Mazuelas, Elena M. Navarro, Slawomir Klos, Alberto Fernández, M. Dolores Jiménez-López
Pages	291-299
Number of pages	9
DOIs	https://doi.org/10.1007/978-3-319-99608-0_33
Publication status	Published - 1 Jan 2019
MoE publication type	A4 Article in a conference publication
Event	International Conference on Distributed Computing and Artificial Intelligence - Toledo, Spain Duration: 20 Jun 2018 → 22 Jun 2018 Conference number: 15

Publication series

Name	Advances in Intelligent Systems and Computing
Publisher	Springer
Volume	801
ISSN (Print)	2194-5357

Conference

Conference	International Conference on Distributed Computing and Artificial Intelligence
Abbreviated title	DCAI
Country	Spain
City	Toledo
Period	20/06/2018 → 22/06/2018

Keywords

Hierarchical Dirichlet Processes
Message forum
Topic modeling

Access to Document

10.1007/978-3-319-99608-0_33

Fingerprint Dive into the research topics of 'Tree-structured hierarchical Dirichlet process'. Together they form a unique fingerprint.

Cite this

Alam, M. H., Peltonen, J., Nummenmaa, J., & Järvelin, K. (2019). Tree-structured hierarchical Dirichlet process. In S. Rodríguez, J. Prieto, P. Faria, M. N. Moreno, S. Mazuelas, E. M. Navarro, S. Klos, A. Fernández, & M. D. Jiménez-López (Eds.), Distributed Computing and Artificial Intelligence, Special Sessions, 15th International Conference (pp. 291-299). (Advances in Intelligent Systems and Computing; Vol. 801). https://doi.org/10.1007/978-3-319-99608-0_33

Alam, Md Hijbul ; Peltonen, Jaakko ; Nummenmaa, Jyrki ; Järvelin, Kalervo. / Tree-structured hierarchical Dirichlet process. Distributed Computing and Artificial Intelligence, Special Sessions, 15th International Conference. editor / Sara Rodríguez ; Javier Prieto ; Pedro Faria ; María N. Moreno ; Santiago Mazuelas ; Elena M. Navarro ; Slawomir Klos ; Alberto Fernández ; M. Dolores Jiménez-López. 2019. pp. 291-299 (Advances in Intelligent Systems and Computing).

@inproceedings{97af25cb9207498fb271433f841a873b,

title = "Tree-structured hierarchical Dirichlet process",

abstract = "In many domains, document sets are hierarchically organized such as message forums having multiple levels of sections. Analysis of latent topics within such content is crucial for tasks like trend and user interest analysis. Nonparametric topic models are a powerful approach, but traditional Hierarchical Dirichlet Processes (HDPs) are unable to fully take into account topic sharing across deep hierarchical structure. We propose the Tree-structured Hierarchical Dirichlet Process, allowing Dirichlet process based topic modeling over a given tree structure of arbitrary size and height, where documents can arise at all tree nodes. Experiments on a hierarchical social message forum and a product reviews forum demonstrate better generalization performance than traditional HDPs in terms of ability to model new data and classify documents to sections.",

keywords = "Hierarchical Dirichlet Processes, Message forum, Topic modeling",

author = "Alam, {Md Hijbul} and Jaakko Peltonen and Jyrki Nummenmaa and Kalervo J{\"a}rvelin",

year = "2019",

month = jan,

day = "1",

doi = "10.1007/978-3-319-99608-0_33",

language = "English",

isbn = "9783319996073",

series = "Advances in Intelligent Systems and Computing",

publisher = "Springer",

pages = "291--299",

editor = "Sara Rodr{\'i}guez and Javier Prieto and Pedro Faria and Moreno, {Mar{\'i}a N.} and Santiago Mazuelas and Navarro, {Elena M.} and Slawomir Klos and Alberto Fern{\'a}ndez and Jim{\'e}nez-L{\'o}pez, {M. Dolores}",

booktitle = "Distributed Computing and Artificial Intelligence, Special Sessions, 15th International Conference",

note = "International Conference on Distributed Computing and Artificial Intelligence, DCAI ; Conference date: 20-06-2018 Through 22-06-2018",

}

Alam, MH, Peltonen, J, Nummenmaa, J & Järvelin, K 2019, Tree-structured hierarchical Dirichlet process. in S Rodríguez, J Prieto, P Faria, MN Moreno, S Mazuelas, EM Navarro, S Klos, A Fernández & MD Jiménez-López (eds), Distributed Computing and Artificial Intelligence, Special Sessions, 15th International Conference. Advances in Intelligent Systems and Computing, vol. 801, pp. 291-299, International Conference on Distributed Computing and Artificial Intelligence, Toledo, Spain, 20/06/2018. https://doi.org/10.1007/978-3-319-99608-0_33

Tree-structured hierarchical Dirichlet process. / Alam, Md Hijbul; Peltonen, Jaakko; Nummenmaa, Jyrki; Järvelin, Kalervo.

Distributed Computing and Artificial Intelligence, Special Sessions, 15th International Conference. ed. / Sara Rodríguez; Javier Prieto; Pedro Faria; María N. Moreno; Santiago Mazuelas; Elena M. Navarro; Slawomir Klos; Alberto Fernández; M. Dolores Jiménez-López. 2019. p. 291-299 (Advances in Intelligent Systems and Computing; Vol. 801).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Scientific › peer-review

TY - GEN

T1 - Tree-structured hierarchical Dirichlet process

AU - Alam, Md Hijbul

AU - Peltonen, Jaakko

AU - Nummenmaa, Jyrki

AU - Järvelin, Kalervo

N1 - Conference code: 15

PY - 2019/1/1

Y1 - 2019/1/1

N2 - In many domains, document sets are hierarchically organized such as message forums having multiple levels of sections. Analysis of latent topics within such content is crucial for tasks like trend and user interest analysis. Nonparametric topic models are a powerful approach, but traditional Hierarchical Dirichlet Processes (HDPs) are unable to fully take into account topic sharing across deep hierarchical structure. We propose the Tree-structured Hierarchical Dirichlet Process, allowing Dirichlet process based topic modeling over a given tree structure of arbitrary size and height, where documents can arise at all tree nodes. Experiments on a hierarchical social message forum and a product reviews forum demonstrate better generalization performance than traditional HDPs in terms of ability to model new data and classify documents to sections.

AB - In many domains, document sets are hierarchically organized such as message forums having multiple levels of sections. Analysis of latent topics within such content is crucial for tasks like trend and user interest analysis. Nonparametric topic models are a powerful approach, but traditional Hierarchical Dirichlet Processes (HDPs) are unable to fully take into account topic sharing across deep hierarchical structure. We propose the Tree-structured Hierarchical Dirichlet Process, allowing Dirichlet process based topic modeling over a given tree structure of arbitrary size and height, where documents can arise at all tree nodes. Experiments on a hierarchical social message forum and a product reviews forum demonstrate better generalization performance than traditional HDPs in terms of ability to model new data and classify documents to sections.

KW - Hierarchical Dirichlet Processes

KW - Message forum

KW - Topic modeling

UR - http://www.scopus.com/inward/record.url?scp=85061708779&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-99608-0_33

DO - 10.1007/978-3-319-99608-0_33

M3 - Conference contribution

AN - SCOPUS:85061708779

SN - 9783319996073

T3 - Advances in Intelligent Systems and Computing

SP - 291

EP - 299

BT - Distributed Computing and Artificial Intelligence, Special Sessions, 15th International Conference

A2 - Rodríguez, Sara

A2 - Prieto, Javier

A2 - Faria, Pedro

A2 - Moreno, María N.

A2 - Mazuelas, Santiago

A2 - Navarro, Elena M.

A2 - Klos, Slawomir

A2 - Fernández, Alberto

A2 - Jiménez-López, M. Dolores

T2 - International Conference on Distributed Computing and Artificial Intelligence

Y2 - 20 June 2018 through 22 June 2018

ER -

Alam MH, Peltonen J, Nummenmaa J, Järvelin K. Tree-structured hierarchical Dirichlet process. In Rodríguez S, Prieto J, Faria P, Moreno MN, Mazuelas S, Navarro EM, Klos S, Fernández A, Jiménez-López MD, editors, Distributed Computing and Artificial Intelligence, Special Sessions, 15th International Conference. 2019. p. 291-299. (Advances in Intelligent Systems and Computing). https://doi.org/10.1007/978-3-319-99608-0_33

Tree-structured hierarchical Dirichlet process

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Cite this