What you seem to be asking for is POS (parts of speech) analysis. You can use OpenNLP to do that for you, likely outside of Solr. OpenNLP will identify nouns, verbs, etc in your sentences. The question is, can you identify certain of those types to be filtered out from your queries?
A simple bit of Java code using OpenNLP should answer that for you. Upayavira On Mon, May 4, 2015, at 05:52 PM, bbarani wrote: > Hi, > > Note: I have very basic knowledge on NLP.. > > I am working on an answer engine prototype where when the user enters a > keyword and searches for it we show them the answer corresponding to that > keyword (rather than displaying multiple documents that match the > keyword) > > For Ex: > > When user searches for 'activate phone', we have answerTags tagged in the > SOLR documents along with answer field (that will be displayed as answer > for > this keyword). > > <arr name="answerTags"> > <str>activate phone</str> > <str>activation</str> > <str>activations</str> > <str>activate</str> > </arr> > > <str name="answers"> > This is the answer > </str> > > This works fine when user searches for the exact keyword tagged in the > 'answerTag' field. > > Now I am trying to figure out a way to match keywords based on position > of > speech too. > > Example: > > I want > > 'how to activate phone' to match 'activate phone' in answerTags field > 'how to activate' to match 'activate' in answerTags field > > I dont want to add all possible combinations of search keywords in > answerTags... rather match is based on position of speech and other NLP > techniques. > > I am trying to figure out a way to standardize the keywords (may be using > NLP) and map it to predefined keywords (may be rule based NLP?). I am not > sure how to proceed with these kinds of searches.. Any insight is > appreciated. > > > > > > > > > -- > View this message in context: > http://lucene.472066.n3.nabble.com/Answer-engine-NLP-related-question-tp4203730.html > Sent from the Solr - User mailing list archive at Nabble.com.
