Lexical Semantics and Word-Sense Disambiguation
The computational study of word meaning: resolving which sense a word carries in context, measuring semantic similarity, and modeling relations among word meanings.
Definition
Lexical semantics in computational linguistics is the representation, comparison, and disambiguation of word meanings by machine.
Scope
Covers the meaning of individual words and their relations — sense inventories, word-sense disambiguation, semantic similarity and relatedness, and distributional models of lexical meaning. It addresses both knowledge-based methods using resources like WordNet and corpus-based and neural methods. Compositional sentence meaning is covered in a sibling topic.
Core questions
- How is a word's sense determined from its context?
- How is semantic similarity between words quantified?
- How do knowledge-based and corpus-based methods to lexical meaning differ?
- How are polysemy and metaphor handled computationally?
Key concepts
- word sense
- polysemy
- sense inventory
- word-sense disambiguation
- semantic similarity
- distributional semantics
- vector-space model
- lexical relation
Key theories
- Word-sense disambiguation
- Selecting the contextually appropriate sense of an ambiguous word from a sense inventory using surrounding context, knowledge resources, or learned classifiers.
- Distributional lexical meaning
- Representing word meaning by co-occurrence statistics and association measures, so that semantically related words have similar contextual profiles.
History
Word-sense disambiguation has been a long-standing challenge, surveyed comprehensively by Navigli in 2009. Distributional approaches, rooted in Harris's hypothesis and advanced by Church and Hanks's association measures, gradually supplied the data-driven similarity models that now dominate lexical semantics.
Debates
- Discrete senses versus continuous meaning
- Whether word meaning is best modeled as a fixed inventory of discrete senses or as a continuous space, a tension sharpened by contextual embeddings that blur sense boundaries.
Key figures
- Roberto Navigli
- Kenneth Church
- Hinrich Schütze
- Zellig Harris
Related topics
Seminal works
- navigli2009
- church1989
Frequently asked questions
- Why is word-sense disambiguation difficult?
- Many words have several senses, and choosing the right one often requires broad world knowledge and subtle contextual cues that are hard to encode, which is why it remained a benchmark challenge for decades.