Replacing OOV Words For Dependency Parsing With Distributional Semantics
Prasanth Kolachina, Martin Johannes Riedl, Chris Biemann · DSpace repository (University of Tartu) · 2017
Lexical information is an important feature in syntactic processing like part-ofspeech (POS) tagging and dependency parsing.However, there is no such information available for out-of-vocabulary (OOV) words, which causes many classification errors.We propose to replace OOV words with in-vocabulary words that are semantically similar according to distributional similar words computed from a large background corpus, as well as morphologically similar according to common suffixes.We show performance differences both for count-based and dense neural vector-based semantic models.Further, we discuss the interplay of POS and lexical information for dependency parsing and provide a detailed analysis and a discussion of results: while we observe significant improvements for count-based methods, neural vectors do not increase the overall accuracy.