Entailment & Synonym Extraction

Language is complex, and there are any number of ways of saying the same thing. TextRazor can help make your application robust to the subtelties of language by extracting relevant word entailments using the document's context.

When building search or other word-matching systems a common approach to is to use synonym lists to expand the words you're looking for.

However, since words can be used in different senses, using a dictionary or a resource like WordNet results in too many false positives. The word "charged" includes synonyms such as "accuse", "bill" and "assault", most of which are only valid in certain contexts.

TextRazor offers a Contextual Entailment/Synonym extractor. For each word or phrase in the text we algorithmically identify a list of potential entailments (words "implied" by your content), by taking into account the source word's meaning. When matching your text we use the context each word comes with to determine how confident we are in a specific entailment being valid in a given context.

Entailment extraction can be used to build powerful semantic search applications, or to help make your pattern extraction rules more robust to different expressions of the same underlying meaning.

Language Support

Contextual entailment extraction is currently only available in English.

API Calls

To retrieve entailments from your text simply add the "entailment" extractor to your request.

Read more in our Python Client or REST Documentation.

You can try out the Entailment Extraction system with your own documents through our Online Demo.