2
$\begingroup$

What is the appropriate method to find n-grams/sub-phrases/parts-of-sequences that are referring to a specific topic or belong to a certain category?

For instance:

Imagine a topic of "transfer of ownership" and some example sentences:

  • A >change of ownership< occurs when a title is transferred from one person or entity to another
  • Making a >change in business ownership< is a lengthy and complex process
  • In this respect, the >transfer of legal ownership< is not the relevant feature for determining the treatment of repo-like operations

What I am looking for is a method to identify sentences and ideally parts of sentences which refer to a specific theme / topic.

What NLP methods are appropriate?

$\endgroup$
0
$\begingroup$

What is the appropriate method to find n-grams/sub-phrases/parts-of-sequences that are referring to a specific topic or belong to a certain category?

  • An important question to solve this problem would be: what is the range of input topics? Is the topic selected among a predefined closed list? Can it be any search query?
  • A similar question might be asked about the target documents and/or terms: can they be processed so that any candidate term is extracted in advance, and the task only consists in identifying the right terms for the particular topic?

Assuming the most open variant of the question (i.e. nothing is available beforehand), I think that one would need:

  • a terminology extraction system which extracts any candidate term from the text (preferably specific to the data to be processed).
  • a third-party resource in order to calculate a semantic representation (typically a vector) of any possible term including the input topic query, so that the topic can be matched/compared against any term.
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.