usagedisambiguation

Usage disambiguation is the process of determining which particular use or sense of a word or phrase is intended in a given context when multiple interpretations are possible. It encompasses not only lexical senses but also variations in usage, such as register, domain, collocation, and pragmatic force. The goal is to map a form to a specific meaning or way of using it within a message, sentence, or discourse.

Usage disambiguation is broader than traditional word sense disambiguation. While WSD typically focuses on selecting the correct sense from a fixed set of meanings, usage disambiguation also considers factors like formality, technical versus everyday language, and contextual expectations. For example, the word lead can refer to guiding someone, the metal, or a pronunciation shift; the intended reading depends on surrounding words, topic, and typical collocations. Similarly, the verb read can denote present or past tense, influencing interpretation beyond a single sense.

Methods for usage disambiguation include rule-based analyses that encode linguistic knowledge and statistical or machine learning approaches that learn from annotated data. Modern systems often employ contextual representations from neural models, such as contextual embeddings, to capture subtle cues in syntax, semantics, and discourse. Evaluation uses labeled corpora and task-specific metrics, including accuracy in selecting the intended sense or usage.

Applications span natural language processing and information retrieval, including voice assistants, search query understanding, machine translation, and lexicography. By clarifying intended meaning, usage disambiguation improves comprehension, reduces ambiguity, and supports more accurate language tools and resources.