Structural semantic interconnections: a knowledge-based approach to word sense disambiguation.

Area of Science:

• Natural Language Processing
• Artificial Intelligence
• Computational Linguistics

Background:

• Word Sense Disambiguation (WSD) is a challenging AI problem with significant implications for web applications.
• Traditional knowledge-based and modern machine learning/statistical methods for WSD have inherent limitations.
• The growing availability of large-scale lexical resources presents new opportunities for knowledge-based WSD.

Purpose of the Study:

• To introduce a novel knowledge-based method for Word Sense Disambiguation called Structural Semantic Interconnections (SSI).
• To address the limitations of existing WSD approaches by integrating diverse lexical resources.
• To demonstrate the applicability of SSI across various semantic disambiguation tasks.

Main Methods:

• Developed the Structural Semantic Interconnections (SSI) method to create structural sense specifications for words in context.
• Integrated multiple lexical resources, partly manually and partly automatically, to generate sense specifications.
• Utilized a grammar G to define relations between sense specifications and select the most plausible word sense.

Main Results:

• Applied the SSI algorithm to diverse disambiguation tasks, including ontology population and text/glossary disambiguation.
• Conducted evaluation experiments across specific domains (tourism, computer networks) and standard test sets.
• Demonstrated the effectiveness of SSI in addressing semantic disambiguation challenges.

Conclusions:

• The SSI method provides a robust knowledge-based alternative for Word Sense Disambiguation.
• SSI effectively leverages integrated lexical resources to improve WSD accuracy.
• The approach shows promise for enhancing various web-based applications reliant on accurate word meaning identification.