coreference resolution
In plain wordsWorking out which words in a text are talking about the same someone or something — for instance, that 'she', 'the senator' and 'Ms. Okafor' all mean one woman.
The task of working out which mentions in a text — names, noun phrases, pronouns — point to one and the same person, object or idea, and collecting each such set into a cluster called a coreference chain. Its subtask, anaphora resolution, links a pronoun or other referring expression to its antecedent. Resolution draws on grammatical agreement, syntactic constraints, salience and world knowledge: in 'the trophy didn't fit in the suitcase because it was too big', only commonsense reveals that 'it' is the trophy, the idea behind the Winograd schema challengeWinograd schema challenge, prop. n.: A test of machine common sense built from pairs of sentences in which changing one word changes what a pronoun refers to — something people resolve instantly using everyday knowledge.. Systems moved from rule-based methods and mention-pair classifiers to end-to-end neural models that score all candidate spans, and LLMs now resolve most everyday cases. Coreference matters for information extractioninformation extraction, n.: Automatically pulling structured facts out of ordinary text — who did what, when, where and for how much — and putting them into a table or database., summarization, question answeringquestion answering, n.: Systems that reply to people's questions with the right answer, whether by finding it in a passage they are given, searching a database or the web, or drawing on what they already know. and translation into languages with grammatical gender.
Without coreference resolution, the extractor credited the merger to 'it' instead of the acquiring company.
Origin
Anaphora resolution was studied in 1970s AI and linguistics, notably in Jerry Hobbs's 1978 pronoun-resolution algorithm. Coreference became a standard evaluation task at the Sixth Message Understanding Conference (MUC-6, 1995), and Kenton Lee and colleagues' end-to-end neural model (2017) set the modern approach.
Related
- information extractioninformation extraction, n.: Automatically pulling structured facts out of ordinary text — who did what, when, where and for how much — and putting them into a table or database.
- Winograd schema challengeWinograd schema challenge, prop. n.: A test of machine common sense built from pairs of sentences in which changing one word changes what a pronoun refers to — something people resolve instantly using everyday knowledge.
- named-entity recognitionnamed-entity recognition, n.: Software that spots names and similar expressions in text and labels each by type: in 'Maria flew from Lagos to Paris on Monday', it marks a person, two cities and a date.
- commonsense reasoningcommonsense reasoning, n.: The everyday background knowledge and reasoning people take for granted — dropped glasses break, hungry people eat, nobody is in two places at once — which machines long struggled to match.
- 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.
Sources
- Lee, K., He, L., Lewis, M. & Zettlemoyer, L. (2017). End-to-end Neural Coreference Resolution. EMNLP 2017.
- Grishman, R. & Sundheim, B. (1996). Message Understanding Conference-6: A Brief History. COLING 1996. aclanthology.org/C96-1079
Mentioned in
information extractioninformation extraction, n.: Automatically pulling structured facts out of ordinary text — who did what, when, where and for how much — and putting them into a table or database., 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.