Hi my solution uses multi value text fields for storing data objects. It works best with relational search using natural language. For example, "car", "automobile", "vehicle" may denote the same class, but they are not equivalent under certain contexts. Multiple value attributes could help the search engine better interpret different expression by user referring to the same concept.
On Sunday, July 10, 2016, Puneet Pawaia <puneet.paw...@gmail.com> wrote: > Hi Yangrui > We are testing the Rank and Retrieve as well as the NLP interface. However > this is being done by another team and so I would not be able to comment > further on it. > I would like to know what kind of Solr field you are using for storing the > output from your classes. And also what function output you are putting > into that field. > Thanks > Puneet > > On 10 Jul 2016 00:17, "Yangrui Guo" <guoyang...@gmail.com <javascript:;>> > wrote: > > > Hi Puneet, > > > > I only use Watson's text to speech as user interface, because a lot of > > people think NLP is the same as voice recognition. If you don't need > voice > > recognition you could remove Watson from it. Stanford has better > dependency > > parsing and can be used offline. However it seems you are using Watson's > > retrieve and rank API, which is based on Solr, am I correct? > > > > Yangrui > > > > On Saturday, July 9, 2016, Puneet Pawaia <puneet.paw...@gmail.com > <javascript:;>> wrote: > > > > > Hi Yangrui, > > > > > > I have been looking at your code for squery. > > > Unfortunately, I am not very conversant with SolrJ. I seem to be > missing > > > how and what data is added to the Solr index. > > > Also, I see some references to IBM Watson in your code. Are you using > IBM > > > Watson? If yes, then why use the Stanford NLP if you can use the Watson > > > NLP? > > > > > > Regards > > > Puneet > > > > > > > > > On Sat, Jul 9, 2016 at 11:37 AM, Puneet Pawaia < > puneet.paw...@gmail.com <javascript:;> > > > <javascript:;>> > > > wrote: > > > > > > > Hi Alessandro > > > > > > > > I am looking at being able to answer questions like "Can a > non-compete > > > > clause in an employment agreement be enforced after the expiry of the > > > > agreement?" > > > > We are doing some testing with IBM Watson and with a sample test > data, > > we > > > > are able to get relevant replies to the above question. Since IBM > > Watson > > > > uses Solr at its backend, I was wondering if we can get the same > > working > > > at > > > > the Solr level without having to use Watson. > > > > > > > > Regards > > > > Puneet > > > > > > > > On Sat, Jul 9, 2016 at 11:34 AM, Puneet Pawaia < > > puneet.paw...@gmail.com <javascript:;> > > > <javascript:;>> > > > > wrote: > > > > > > > >> Hi Alessandro > > > >> > > > >> I am looking at being able to answer questions like "Can a > non-compete > > > >> clause in an employment agreement be enforced after the expiry of > the > > > >> agreement?" > > > >> > > > >> On Sat, Jul 9, 2016 at 4:34 AM, Alessandro Benedetti < > > > >> abenede...@apache.org <javascript:;> <javascript:;>> wrote: > > > >> > > > >>> Hi Puneet, > > > >>> your requirement : > > > >>> "I would like users to be able to write queries in natural language > > > >>> rather > > > >>> than keyword based search." > > > >>> > > > >>> Is really really vague :( > > > >>> Can you try to help us with some specific example, starting of > course > > > >>> from > > > >>> the simplest use cases you have initially in mind ? > > > >>> > > > >>> Moving from keyword based search to natural language is a really > > > complex > > > >>> task. > > > >>> Proceeding step by step can help you. > > > >>> > > > >>> Do you want for example to set up a Q&A basic system ? > > > >>> In that case you should take care of query rewriting. > > > >>> You need basically to identify your base requirement and then > build a > > > >>> specific parser for that. > > > >>> You can use triple stores and knowledge bases to enrich both your > > query > > > >>> and > > > >>> your index, but let's start from the basis, what is your simplest > > > >>> requirement ? > > > >>> > > > >>> On Fri, Jul 8, 2016 at 1:56 PM, Jay Urbain <jay.urb...@gmail.com > <javascript:;> > > > <javascript:;>> wrote: > > > >>> > > > >>> > I've added multivalued fields within my SOLR schema for indexing > > > >>> entities > > > >>> > extracted using NLP methods applied to the text I'm indexing, > along > > > >>> with > > > >>> > fields for other discrete data extracted from relational > databases. > > > >>> > > > > >>> > A Java application reads data out of multiple relational > databases, > > > >>> uses > > > >>> > NLP on the text and indexes each document (de-normalized) using > > > SOLRJ. > > > >>> > > > > >>> > I initially tried doing this with content handlers, but found it > > much > > > >>> > easier to just write a Java application. > > > >>> > > > > >>> > SOLRJ Java API reference: > > > >>> > https://cwiki.apache.org/confluence/display/solr/Using+SolrJ > > > >>> > > > > >>> > Stanford NLP: > > > >>> > http://stanfordnlp.github.io/CoreNLP/ > > > >>> > > > > >>> > Best, > > > >>> > Jay > > > >>> > > > > >>> > > > > >>> > On Thu, Jul 7, 2016 at 9:52 PM, Puneet Pawaia < > > > puneet.paw...@gmail.com <javascript:;> <javascript:;> > > > >>> > > > > >>> > wrote: > > > >>> > > > > >>> > > Hi Jay > > > >>> > > Any place I can learn more on this method of integration? > > > >>> > > Thanks > > > >>> > > Puneet > > > >>> > > > > > >>> > > On 8 Jul 2016 02:58, "Jay Urbain" <jay.urb...@gmail.com > <javascript:;> > > > <javascript:;>> wrote: > > > >>> > > > > > >>> > > > I use Stanford NLP and cTakes (based on OpenNLP) while > indexing > > > >>> with a > > > >>> > > > SOLRJ application. > > > >>> > > > > > > >>> > > > Best, > > > >>> > > > Jay > > > >>> > > > > > > >>> > > > On Thu, Jul 7, 2016 at 12:09 PM, Puneet Pawaia < > > > >>> > puneet.paw...@gmail.com <javascript:;> <javascript:;>> > > > >>> > > > wrote: > > > >>> > > > > > > >>> > > > > Hi > > > >>> > > > > > > > >>> > > > > I am currently using Solr 5.5.x to test but can upgrade to > > Solr > > > >>> 6.x > > > >>> > if > > > >>> > > > > required. > > > >>> > > > > I am working on a POC for natural language query using > Solr. > > > >>> Should I > > > >>> > > use > > > >>> > > > > the Stanford libraries or are there any other libraries > > having > > > >>> > > > integration > > > >>> > > > > with Solr already available. > > > >>> > > > > Any direction in how to do this would be most appreciated. > > How > > > >>> > should I > > > >>> > > > > process the query to give relevant results. > > > >>> > > > > > > > >>> > > > > Regards > > > >>> > > > > Puneet > > > >>> > > > > > > > >>> > > > > > > >>> > > > > > >>> > > > > >>> > > > >>> > > > >>> > > > >>> -- > > > >>> -------------------------- > > > >>> > > > >>> Benedetti Alessandro > > > >>> Visiting card : http://about.me/alessandro_benedetti > > > >>> > > > >>> "Tyger, tyger burning bright > > > >>> In the forests of the night, > > > >>> What immortal hand or eye > > > >>> Could frame thy fearful symmetry?" > > > >>> > > > >>> William Blake - Songs of Experience -1794 England > > > >>> > > > >> > > > >> > > > > > > > > > >
