Proceedings of the Workshop on Linguistic Distances, page 7,
Sydney, July 2006. c©2006 Association for Computational Linguistics
Semantic Similarity: What for?
Ido Dagan
Bar Ilan University
dagan@macs.biu.ac.il
Abstract
Linguistic similarity has been a promi-
nent notion and tool in computational lin-
guistics and related areas, as elaborated
nicely in the announcement of this work-
shop. Yet, what exactly counts as “sim-
ilarity”, or when two linguistic concepts
should be regarded as similar, often re-
mains rather vague and ill posed, which is
in fact quite typical for unsupervised no-
tions. This talk will focus on similarity
at the semantic level, and will explore the
perspective that different notions of simi-
larity may be defined relative to concrete
modeling goals. In particular, I will refer
to the two major goals in semantic mod-
eling: predicting likelihood of occurrence,
which is the typical goal in disambigua-
tion and language modeling, and recogniz-
ing target meanings, which is the typical
semantic goal in text understanding appli-
cations such as question answering, infor-
mation extraction, summarization and in-
formation retrieval. We will discuss each
goal and present corresponding semantic
similarity approaches.
