coreferring

Coreferring, or coreference, is the linguistic relation in which two or more expressions refer to the same entity within a discourse or in the real world. Coreferential expressions include pronouns (he, they), proper names (Alice), and noun phrases (the winner of the race). This relation often spans multiple sentences and is essential for coherent interpretation, making coreference resolution a central task in linguistics and natural language processing.

Coreference types include anaphora, where a later expression refers back to an earlier antecedent (Alice arrived and she sat down; 'she' corefers with 'Alice'). Cataphora, where a pronoun or determiner refers forward to a later noun phrase (Before he spoke, John entered the room; 'he' refers to John). Exophora describes references to entities outside the text, such as when pointing to someone present in the surroundings. Coreference can involve more than two expressions referring to the same entity.

Coreference resolution is a common NLP task that identifies all coreferential mentions in a text. It supports information extraction, question answering, summarization, and machine translation. Modern approaches combine linguistic features with statistical or neural models and rely on annotated corpora. Datasets such as OntoNotes and the CoNLL shared tasks have driven progress in English and other languages. Challenges include ambiguity, long-distance references, discourse structure, and the need for world knowledge to distinguish plausible antecedents.