PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Analysis of Textual Variation by Latent Tree Structures
Teemu Roos and Yuan Zou
In: The 2011 ICDM IEEE International Conference on Data Mining, Vancouver, Canada(2011).


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 state- of-the-art both in terms of a numerical score as well as interpretability.

PDF - Requires Adobe Acrobat Reader or other PDF viewer.
EPrint Type:Conference or Workshop Item (Talk)
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Computational, Information-Theoretic Learning with Statistics
Learning/Statistics & Optimisation
Natural Language Processing
Theory & Algorithms
Information Retrieval & Textual Information Access
ID Code:9129
Deposited By:Petri Myllymäki
Deposited On:21 February 2012