| | |
| | |
Stat |
Members: 3667 Articles: 2'599'751 Articles rated: 2609
07 February 2025 |
|
| | | |
|
Article overview
| |
|
Judgment Language Matters: Multilingual Vector Space Models for Judgment Language Aware Lexical Semantics | Ira Leviant
; Roi Reichart
; | Date: |
1 Aug 2015 | Abstract: | It is a common practice in the vector space model (VSM) literature to
evaluate the models’ ability to predict human similarity scores for a set of
word pairs. However, existing evaluation sets, even those used to evaluate
multilingual VSMs, consist of English words only. In this paper we show that
this practice may have significant undesired effects on VSM evaluation. By
translating the popular wordsim353 evaluation set to three languages and
training state-of-the-art VSMs on corpora of the corresponding languages as
well as on English, we show that: (a) The judgment language in which word pairs
are presented to human evaluators, all fluent in that language, has a
substantial impact on their produced similarity scores; (b) Given the judgment
language of an evaluation set, this judgment language is a good choice for the
VSM training corpus language; and (c) Monolingual VSMs can be combined into
multilingual VSMs that can predict human similarity scores for a variety of
judgment languages better than any monolingual model. Our results highlight the
impact of the judgment language on the human generated similarity scores and
point on the importance of constructing judgment language aware VSMs. | Source: | arXiv, 1508.0106 | Services: | Forum | Review | PDF | Favorites |
|
|
No review found.
Did you like this article?
Note: answers to reviews or questions about the article must be posted in the forum section.
Authors are not allowed to review their own article. They can use the forum section.
|
| |
|
|
|