Coreference resolution is the process of finding the source of an apparent reference. For example, if you say “the girl who lives next door”, you are referring to someone specific. If you also say “the pretty girl who plays piano”, then this second reference could refer to any number of people, depending on which houses they live in and how many girls play piano in those neighborhoods. Finding out who the second reference refers to is known as referential resolution. This article will explain different methods used to resolve references in natural language processing (NLP), and give brief examples of each method.
Synonym resolution
Synonym resolution is the process of finding synonyms for words. For example, if you say “the pretty girl next door”, you are referring to “the girl who lives next door to you”. Therefore, the word “pretty” is a synonym for “girl”. You may also use synonyms for other attributes of the word, such as “the pretty girl who lives next door to you”, where “pretty” means “the attribute of being a girl”. Because synonyms are substitutions for one another, synonym resolution can be used to find the source of an apparent reference. Synonym resolution is mostly a heuristic process. It is not guaranteed to work in all cases. In particular, a word may have multiple synonyms (overlapping senses) and it may be impossible to choose a synonym that works for all cases. If a synonym is not found, the word’s meaning can be used as a default synonym. Depending on the language, the default synonym may be different in each language. In languages with polyvalent words, such as some Indo-European languages, choosing an appropriate default synonym may be very important.
Hypernym Resolution
Hypernym resolution is the process of finding subtypes of a word or phrase. For example, if you say “the girl who plays the piano”, you are referring to “the girl who plays the piano in your neighborhood”. Therefore, the word “piano” is a hypernym of the word “girl”. You may also use hypernyms for other attributes of the word, such as “the girl who plays a piano”, where “piano” means “the attribute of being a piano”. Depending on the language, the default synonym for the hypernym may be different in each language. In languages with polyvalent words, such as some Indo-European languages, choosing an appropriate default synonym may be very important.
Semantic Role Analogy Resolution
Semantic Role Analogy Resolution is the process of finding analogies of words. For example, if you say “the girl who lives next door”, you are referring to “the girl who lives next door to you”. Therefore, the word “next door” is an analogy of the word “girl”. You may also use analogies for other attributes of the word, such as “the girl who lives next door to you”, where “next door” means “the attribute of being next door to you”. Depending on the language, the default synonym for the analogy may be different in each language. In languages with polyvalent words, such as some Indo-European languages, choosing an appropriate default synonym may be very important.
Entity Matching Resolution
Entity matching resolution is the process of finding entities in the text and matching them to the entities in the training data. For example, if you say “the girl who plays the piano”, you are referring to “the girl who plays the piano in your neighborhood”. Therefore, the word “piano” is a matching entity. You may also use matching entities for other attributes of the word, such as “the girl who plays a piano”, where “piano” means “the attribute of being a piano”. Depending on the language, the default synonym for the matching entity may be different in each language. In languages with polyvalent words, such as some Indo-European languages, choosing an appropriate default synonym may be very important.
Co-reference resolution
Co-reference resolution is the process of finding words that refer to the same entity as the one in the apparent reference. For example, if you say “the girl who lives next door”, you are referring to “the girl who lives next door to you”. Therefore, the word “next door” is a co-reference of the word “you”. You may also use co-references for other attributes of the word, such as “the girl who lives next door to you”, where “next door” means “the attribute of being next door to you”. Depending on the language, the default synonym for the co-reference may be different in each language. In languages with polyvalent words, such as some Indo-European languages, choosing an appropriate default synonym may be very important.
Summary
The resolution of reference is the process of finding the source of an apparent reference. Synonym resolution is mostly a heuristic process that works in many cases, but may not work in all cases. Hypernym resolution is the process of finding subtypes of a word or phrase. Semantic Role Analogy Resolution is the process of finding analogies of words. Entity matching resolution is the process of finding entities in the text and matching them to the entities in the training data. Co-reference resolution is the process of finding words that refer to the same entity as the one in the apparent reference.