coreferences

Coreference is a linguistic concept in which two or more expressions in a discourse refer to the same entity. Expressions that can participate in coreference include pronouns (she, they), proper names (Alice, Paris), and definite noun phrases (the president, the city). For example: "Alice arrived late because she missed the bus." Here the pronoun she refers to Alice, establishing a coreference link between the two mentions.

Coreference can be intra-sentential (within a single sentence) or inter-sentential (across sentences) and can extend over multiple occurrences in a text. Anaphora is a common type of coreference where a later expression points back to an earlier antecedent. Cataphora is the reverse, where an expression points forward to a later reference. When several mentions refer to the same entity, they form a coreference chain or cluster that links all related expressions together.

In computational linguistics, coreference resolution is the task of identifying which mentions refer to the same entity. This involves detecting mentions, selecting antecedents, and clustering related mentions. Approaches range from rule-based to statistical and neural models, often using cues from syntax, discourse structure, world knowledge, and entity typing. Cross-document coreference extends the problem to identifying references across multiple texts.

Coreference underpins understanding of discourse and supports applications such as information extraction, machine translation, summarization, and question answering. Annotated datasets and evaluation metrics (precision, recall, F1) are used to measure the accuracy of linking decisions. Ongoing challenges include ambiguity, long-range dependencies, and cross-linguistic variation in discourse patterns.