Boosting Tricks for Word Mover's Distance

Konstantinos Skianis (1) , Fragkiskos D. Malliaros (2) , Nikolaos Tziortziotis (3) , Michalis Vazirgiannis (4)

1 BLUAI
2 OPIS - OPtimisation Imagerie et Santé
3 Jellyfish
4 X-DEP-INFO - Département d'informatique de l'École polytechnique

Word embeddings have opened a new path in creating novel approaches for addressing traditional problems in the natural language processing (NLP) domain. However, using word embeddings to compare text documents remains a relatively unexplored topic-with Word Mover's Distance (WMD) being the prominent tool used so far. In this paper, we present a variety of tools that can further improve the computation of distances between documents based on WMD. We demonstrate that, alternative stopwords, cross document-topic comparison, deep contextualized word vectors and convex metric learning, constitute powerful tools that can boost WMD.

Mots clés: Word mover's distance; Word embeddings; Text classification

Domaines: Traitement du texte et du document; Recherche d'information [cs.IR]; Apprentissage [cs.LG]; Intelligence artificielle [cs.AI]

Konstantinos Skianis, Fragkiskos D. Malliaros, Nikolaos Tziortziotis, Michalis Vazirgiannis. Boosting Tricks for Word Mover's Distance. ICANN 2020 - 29th International Conference on Artificial Neural Networks, Sep 2020, Bratislava, Slovakia. ⟨hal-03088924⟩