Evaluating Distributed Representations for Multi-Level Lexical Semantics: A Research Proposal

TL;DR

This research proposal aims to evaluate how well neural network-based distributed representations encode different levels of lexical semantics, bridging computational models and linguistic theory.

Contribution

It formalizes three levels of lexical semantics and proposes a framework for evaluating language models across these levels using multilingual datasets.

Findings

Formalization of local, global, and mixed lexical semantic levels

Development of multilingual datasets for semantic evaluation

Framework for assessing neural network representations of lexical meaning

Abstract

Modern neural networks (NNs), trained on extensive raw sentence data, construct distributed representations by compressing individual words into dense, continuous, high-dimensional vectors. These representations are expected to capture multi-level lexical meaning. In this thesis, our objective is to examine the efficacy of distributed representations from NNs in encoding lexical meaning. Initially, we identify and formalize three levels of lexical semantics: \textit{local}, \textit{global}, and \textit{mixed} levels. Then, for each level, we evaluate language models by collecting or constructing multilingual datasets, leveraging various language models, and employing linguistic analysis theories. This thesis builds a bridge between computational models and lexical semantics, aiming to complement each other.