Reputation: 145
I have ingested a pdf document about 60 pages and the requirement is to return only the paragraph that matches the given query. Currently, the complete text from document is being returned. I tried out various samples and wondering if this is possible.
Upvotes: 1
Views: 623
Reputation: 21
There are a few approaches to try. You may want to try chunking the documents into smaller segments.
Alternatively there is a new feature that may help you called Passage Retrieval and Relevancy Training. There is a YouTube video released on May 18th 2017 that discusses this.
https://www.youtube.com/watch?v=8BiuQKPQZJk
Good luck!
Upvotes: 1
Reputation: 18
This question is a duplicate of How to get exact answers instead of the whole document using Watson Discovery?
One of the answers:
You are correct that today Discovery service returns entire documents for a matched query, ranked by relevance to the query.
Upvotes: 0