Word Sense Disambiguation: A Unified Evaluation Framework and Empirical Comparison
TL
TL;DR
A unified evaluation framework is developed and the results show that supervised systems clearly outperform knowledge-based models in Word Sense Disambiguation, and a linear classifier trained on conventional local features still proves to be a hard baseline to beat.
Abstract
Word Sense Disambiguation is a long-standing task in Natural Language Processing, lying at the core of human language understanding. However, the evaluation of automatic systems has been problematic, mainly due to the lack of a reliable evaluation framework. In this paper we develop a unified evaluation framework and analyze the performance of various Word Sense Disambiguation systems in a fair setup. The results show that supervised systems clearly outperform knowledge-based models. Among the supervised systems, a linear classifier trained on conventional local features still proves to be a hard baseline to beat. Nonetheless, recent approaches exploiting neural networks on unlabeled corpora achieve promising results, surpassing this hard baseline in most test sets.
References92 items
1
A statistical interpretation of term specificity and its application in retrieval
2
Who is Killed by Police: Introducing Supervised Attention for Hierarchical LSTMs
3
Event Detection via Gated Multilingual Attention Mechanism
4
Graph Convolutional Networks With Argument-Aware Pooling for Event Detection
5