IBM Watson™ Discovery Service Ideas

We've moved...

You'll be redirected shortly, we've moved to our new idea portal: https://ibm-watson.ideas.aha.io

Ability to do semantic search

Currently first level search in WDS is based on term overlap which can be improved with relevancy training. The term overlap can be improved with query expansion.

Now for a query on "car", customer also wants to retrieve documents on "Vehicle", "Motor" etc. This is general domain knowledge and some customers believe that WDS will handle this. Currently this can be handled only with custom query expansion.

We need an option where this general domain synonyms are already utilized in WDS to provide a semantic search. Customers understand this can be created with effort, but than there is nothing "smart" that WDS brings to the table.

We need an option (may be optional via configuration) where WDS does semantic search using general domain knowledge. If this does not give good results, customers can have a configuration option and choose only terms overlap search.

  • Lalit Agarwalla
  • Apr 20 2018
  • Attach files
  • Admin
    Phil Anderson commented
    May 9, 2018 05:12

    We already do this today: https://apps.na.collabserv.com/blogs/152f58a2-3bb3-4992-86a7-c56ad4bbd21c/entry/Smarter_document_retrieval_with_Semantic_Match?lang=en_us

  • Lalit Agarwalla commented
    May 9, 2018 06:55

     Hi Phil, The feature requested is different from what already exists. In the existing semantic match, it semantic information is used only during scoring the results, not during search.

    In the requested feature, the semantic search to be used during initial search. During scoring & ranking, it is already done as you have mentioned. The requested feature will help in improved recall. This can be configurable feature.