disambigeerimine

Disambigeerimine, or disambiguation, is the process of resolving ambiguity in language, data, or signals by identifying the intended meaning, sense, or referent among multiple candidates. It aims to map an ambiguous input to a single, appropriate interpretation in a given context.

In linguistics, disambiguation often refers to word sense disambiguation, where a polysemous word is assigned the correct sense from a set of possible senses. In information retrieval, knowledge representation, and natural language processing, it also covers entity disambiguation or entity linking, where mentions in text are connected to the correct real-world entity such as a person, place, or organization.

Approaches vary. Knowledge-based methods use lexical databases and semantic networks to compare context with known senses. Supervised learning employs labeled corpora to train classifiers. Unsupervised and semi-supervised methods seek patterns without extensive annotation. Neural models and contextual embeddings can infer meaning from surrounding words.

Common resources include WordNet, BabelNet, Wikidata, and Wikipedia disambiguation pages. Evaluation typically reports accuracy, precision, recall, and F1 on standard datasets such as SemCor or domain-specific corpora.

Applications include improving search results, question answering, machine translation, and text summarization, as well as aiding information extraction and content recommendation. Example: the word bank can refer to a financial institution or a riverbank; disambiguation uses contextual clues to identify the intended sense.

Challenges include cross-language and domain transfer, sense granularity, and scarcity of labeled data for certain languages or domains. Ongoing work combines symbolic knowledge with neural methods to improve robustness and explainability. Related concepts include word sense disambiguation and entity linking.