Basque Word Sense Disambiguation: Did Systems Perform in the Upperbound?

Title: Basque Word Sense Disambiguation: Did Systems Perform in the Upperbound?

Research Question: Can machine learning systems accurately disambiguate the meanings of Basque words?

Methodology: The researchers created a hand-tagged corpus of 40 Basque words (nouns, verbs, and adjectives) by selecting examples from the Egunkaria newspaper and Euskal Hiztegia dictionary. They used these examples to train and evaluate four competing systems. The systems used various machine learning algorithms to disambiguate the word senses.

Results: The systems achieved high precision rates, exceeding the most frequent baseline and performing above the expected upperbound. This suggests that the inter-tagger agreement is not a real upperbound for the Basque Word Sense Disambiguation task.

Implications: The results indicate that machine learning systems can accurately disambiguate the meanings of Basque words. This has implications for the field of natural language processing, as it demonstrates the potential of these systems to handle agglutinative languages like Basque. The study also provides valuable insights into the design of lexical-sample tasks for other languages.

Link to Article: https://arxiv.org/abs/0204029v1 Authors: arXiv ID: 0204029v1