0

i am currently evaluating whether Watson Concept Insights can be used for the use case i am trying to implement. I have a large collection of documents of various styles, length, and content. I want to use Concept Insights to determine, for new documents, whether there are documents in the corpus that deal with similar concepts. As far as i can see, this is what Concept Insights was made for. The problem is that the documents i am dealing with are highly technical for a very specific field. I think that the Concepts derived from Wikipedia are not sufficient to tell different documents apart. So - can i add to that concept graph to make the search and comparison more useful to me?

Thank you.

Dominik Steenken
  • 1
  • 2
  • As it stands it is not possible to augment the concept graph that the Concept Insights service uses. Have you tried using a sample of your documents against Concept Insights to verify if there are limitations with your dataset? – chughts Jun 27 '16 at 10:43
  • Thank you for your answer! Assembling the corpus is, for us, a very work-intensive and time-consuming task that we are only going to undertake if and when there is a reasonable expectation that the result will be worthwhile. So we haven't done any experiments yet. We will, however, conduct some manual evaluations - would i be correct in assuming that, if i can manually find a concept on the english wikipedia, then Watson Concept Insights will reliably recognize it in the corpus? – Dominik Steenken Jun 28 '16 at 09:08
  • if there is a Wikipedia article then it will be recognized in the corpus – German Attanasio Jun 28 '16 at 17:08

1 Answers1

0

The comments for the questions above have answered my question. thanks.

Dominik Steenken
  • 1
  • 2