word sense disambiguation
In plain wordsWorking out which meaning of a word is intended in context — whether 'bank' means a riverside or a place that keeps money.
The task of deciding which of an ambiguous word's meanings, as listed in a sense inventory such as a dictionary or WordNet, is intended in a particular context. Approaches include knowledge-based methods that compare context words with dictionary definitions, supervised classifiers trained on sense-annotated corpora, semi-supervised bootstrapping that exploits regularities such as one sense per discourse, and neural models whose contextual embeddings separate senses implicitly. Word sense disambiguation was long considered a central obstacle for machine translationmachine translation, n.: Software that renders writing or speech in one language into another language, as online translation services and phone translation apps do. and information retrievalinformation retrieval, n.: The study and engineering of search: how to organize huge piles of documents so that the few that matter to someone's question can be found and ranked quickly. AI systems rely on it to look things up., where choosing the wrong sense changes the output. It is now rarely a separate processing stage, since contextual models such as BERTBERT, prop. n.: An influential 2018 Google language model that learned to understand text by filling in hidden words. It made search engines and many language tools much better at grasping meaning in context. and LLMs resolve most ambiguity in passing, but it remains a probe of lexical knowledge.
Contextual embeddings do word sense disambiguation implicitly: 'bass' near 'guitar' and 'bass' near 'lake' land far apart.
Origin
Warren Weaver's 1949 translation memorandum proposed resolving ambiguity from neighboring words, and Yehoshua Bar-Hillel argued in 1960 that fully automatic high-quality translation was unattainable partly because of such ambiguity. Michael Lesk's dictionary-overlap algorithm (1986) and David Yarowsky's bootstrapping method (1995) are landmark techniques.
Related
- natural language understandingnatural language understanding, n.: The part of language AI concerned with working out what a piece of text means — who did what to whom, what the writer wants and what follows from it — rather than with producing text.
- machine translationmachine translation, n.: Software that renders writing or speech in one language into another language, as online translation services and phone translation apps do.
- word embeddingword embedding, n.: A list of numbers standing for a word, learned so that words used in similar ways get similar lists. It lets a computer measure that 'cat' is closer to 'dog' than to 'car'.
- BERTBERT, prop. n.: An influential 2018 Google language model that learned to understand text by filling in hidden words. It made search engines and many language tools much better at grasping meaning in context.
- distributional hypothesisdistributional hypothesis, n.: The idea that words found in similar company — surrounded by similar words — tend to be similar in meaning, the insight that lets computers learn something about word meaning from large amounts of text alone.
- computational linguisticscomputational linguistics, n.: The science of human language studied with computers: using computational models to understand how language works, and building the language technology that grows out of that understanding.
Sources
- Navigli, R. (2009). Word Sense Disambiguation: A Survey. ACM Computing Surveys 41(2), Article 10.
- Yarowsky, D. (1995). Unsupervised Word Sense Disambiguation Rivaling Supervised Methods. ACL 1995.
Mentioned in
WordNetWordNet, prop. n.: A hand-made map of English vocabulary in which words sharing a meaning are bundled together and the bundles are linked — recording, say, that a dog is a kind of animal and a wheel is part of a car.