Analysis of Textual Variation by Latent Tree Structures

Teemu Teppo Roos; Yuan Zou

doi:10.1109/ICDM.2011.24

Analysis of Textual Variation by Latent Tree Structures

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

Abstract

We introduce Semstem, a new method for the
reconstruction of so called stemmatic trees, i.e., trees encoding
the copying relationships among a set of textual variants.
Our method is based on a structural expectation-maximization
(structural EM) algorithm. It is the first computer-based
method able to estimate general latent tree structures, unlike
earlier methods that are usually restricted to bifurcating trees
where all the extant texts are placed in the leaf nodes. We
present experiments on two well known benchmark data
sets, showing that the new method outperforms current stateof-
the-art both in terms of a numerical score as well as
interpretability.

Original language	English
Title of host publication	2011 IEEE 11th International Conference on Data Mining (ICDM 2011)
Editors	Diane Cook, Jian Pei, Wei Wang, Osmar Zaïane, Xindong Wu
Number of pages	10
Publisher	IEEE Computer Society
Publication date	11 Dec 2011
Pages	567-576
ISBN (Print)	9781457720758
DOIs	https://doi.org/10.1109/ICDM.2011.24
Publication status	Published - 11 Dec 2011
MoE publication type	A4 Article in conference proceedings
Event	Unknown host publication - , Canada Duration: 1 Jan 1800 → …

Fields of Science

113 Computer and information sciences

Access to Document

10.1109/ICDM.2011.24

Stemmatologian algoritmiset menetelmät
Roos, T., Myllymäki, P., Heikkilä, T. & Zou, Y.
Helsingin yliopiston rahastot
01/01/2009 → 31/12/2011
Project: Research project

Cite this

@inproceedings{fa043cd1ef1f41ae882e4b1177d1ddbc,

title = "Analysis of Textual Variation by Latent Tree Structures",

abstract = "We introduce Semstem, a new method for thereconstruction of so called stemmatic trees, i.e., trees encodingthe copying relationships among a set of textual variants.Our method is based on a structural expectation-maximization(structural EM) algorithm. It is the first computer-basedmethod able to estimate general latent tree structures, unlikeearlier methods that are usually restricted to bifurcating treeswhere all the extant texts are placed in the leaf nodes. Wepresent experiments on two well known benchmark datasets, showing that the new method outperforms current stateof-the-art both in terms of a numerical score as well asinterpretability.",

keywords = "113 Computer and information sciences",

author = "Roos, {Teemu Teppo} and Yuan Zou",

year = "2011",

month = dec,

day = "11",

doi = "10.1109/ICDM.2011.24",

language = "English",

isbn = "9781457720758",

pages = "567--576 ",

editor = "Diane Cook and Jian Pei and Wei Wang and Osmar Za{\"i}ane and Xindong Wu",

booktitle = "2011 IEEE 11th International Conference on Data Mining (ICDM 2011)",

publisher = "IEEE Computer Society",

address = "United States",

note = "Unknown host publication ; Conference date: 01-01-1800",

}

Analysis of Textual Variation by Latent Tree Structures. / Roos, Teemu Teppo; Zou, Yuan.
2011 IEEE 11th International Conference on Data Mining (ICDM 2011). ed. / Diane Cook; Jian Pei; Wei Wang; Osmar Zaïane; Xindong Wu. IEEE Computer Society, 2011. p. 567-576 .

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

TY - GEN

T1 - Analysis of Textual Variation by Latent Tree Structures

AU - Roos, Teemu Teppo

AU - Zou, Yuan

PY - 2011/12/11

Y1 - 2011/12/11

N2 - We introduce Semstem, a new method for thereconstruction of so called stemmatic trees, i.e., trees encodingthe copying relationships among a set of textual variants.Our method is based on a structural expectation-maximization(structural EM) algorithm. It is the first computer-basedmethod able to estimate general latent tree structures, unlikeearlier methods that are usually restricted to bifurcating treeswhere all the extant texts are placed in the leaf nodes. Wepresent experiments on two well known benchmark datasets, showing that the new method outperforms current stateof-the-art both in terms of a numerical score as well asinterpretability.

AB - We introduce Semstem, a new method for thereconstruction of so called stemmatic trees, i.e., trees encodingthe copying relationships among a set of textual variants.Our method is based on a structural expectation-maximization(structural EM) algorithm. It is the first computer-basedmethod able to estimate general latent tree structures, unlikeearlier methods that are usually restricted to bifurcating treeswhere all the extant texts are placed in the leaf nodes. Wepresent experiments on two well known benchmark datasets, showing that the new method outperforms current stateof-the-art both in terms of a numerical score as well asinterpretability.

KW - 113 Computer and information sciences

U2 - 10.1109/ICDM.2011.24

DO - 10.1109/ICDM.2011.24

M3 - Conference contribution

SN - 9781457720758

SP - 567

EP - 576

BT - 2011 IEEE 11th International Conference on Data Mining (ICDM 2011)

A2 - Cook, Diane

A2 - Pei, Jian

A2 - Wang, Wei

A2 - Zaïane, Osmar

A2 - Wu, Xindong

PB - IEEE Computer Society

T2 - Unknown host publication

Y2 - 1 January 1800

ER -

Analysis of Textual Variation by Latent Tree Structures

Abstract

Fields of Science

Access to Document

Projects

Stemmatologian algoritmiset menetelmät

Cite this