Text Similarity in Vector Space Models: A Comparative Study

17 Pages Posted: 26 Oct 2018

See all articles by Omid Shahmirzadi

Omid Shahmirzadi

Ecole Polytechnique Fédérale de Lausanne

Adam Lugowski

Patent Research Foundation

Kenneth A. Younge

Ecole Polytechnique Fédérale de Lausanne

Date Written: September 15, 2018

Abstract

Automatic measurement of semantic text similarity is an important task in natural language processing. In this paper, we evaluate the performance of different vector space models to perform this task. We address the real-world problem of modeling patent-to-patent similarity and compare TFIDF (and related extensions), topic models (e.g., latent semantic indexing), and neural models (e.g., paragraph vectors).

Contrary to expectations, the added computational cost of text embedding methods is justified only when:

1) the target text is condensed; and

2) the similarity comparison is trivial.

Otherwise, TFIDF performs surprisingly well in other cases: in particular for longer and more technical texts or for making finer-grained distinctions between nearest neighbors. Unexpectedly, extensions to the TFIDF method, such as adding noun phrases or calculating term weights incrementally, were not helpful in our context.

Keywords: text similarity, vector space model, text embedding, patent, big data

Suggested Citation

Shahmirzadi, Omid and Lugowski, Adam and Younge, Kenneth A., Text Similarity in Vector Space Models: A Comparative Study (September 15, 2018). Available at SSRN: https://ssrn.com/abstract=3259971 or http://dx.doi.org/10.2139/ssrn.3259971

Omid Shahmirzadi

Ecole Polytechnique Fédérale de Lausanne ( email )

Station 5
Odyssea 1.04
1015 Lausanne, CH-1015
Switzerland

Adam Lugowski

Patent Research Foundation ( email )

Seattle, WA
United States

Kenneth A. Younge (Contact Author)

Ecole Polytechnique Fédérale de Lausanne ( email )

Station 5
1015 Lausanne
Switzerland

Register to save articles to
your library

Register

Paper statistics

Downloads
18
Abstract Views
128
PlumX Metrics