0

I have a large set of KB Articles (software support), and I am looking for ways to extract answers based on a question asked against these KB articles. I am looking for guidance to understand the following -

  • Are there libraries available (maybe Python) that could allow me to get a potential answer or a set of ranked/scored answers based on a question posed?.

I had a look at IBMs Watson Service and that does a pretty good job, but these articles are not allowed to be interpreted through a cloud service.

PS: I know questions requesting an open-ended guidance are not usually part of StackOverflow's ecosystem, but the domain this question covers is too wide to get a good direction anywhere else.

Nitin Rastogi
  • 1,446
  • 16
  • 30

2 Answers2

0

Yes, its possible. My team at Bluetick Consultants has been working on this since 6 months and we have figured it out. We have tested it on dropbox KB. Our solution uses multiple python libraries, tensorflow and custom ML models.

Ankur Gupta
  • 28
  • 6
  • Great to hear that it works. Do you have it publicly documented?. If not, would you be able to share your toolset, broad steps you followed?. – Nitin Rastogi Jul 23 '18 at 17:57
  • Not yet public. We want to test it with other KBs as well. I may be able to help on specific problem than complete solution. – Ankur Gupta Jul 24 '18 at 07:43
  • Our use case is - We have KB articles for various internal IT Systems, and we are working on answering questions on these. – Nitin Rastogi Jul 24 '18 at 13:25
  • Cool. Please shoot an email to contact@bluetickconsultants.com and someone from my team will reach out to you. – Ankur Gupta Jul 25 '18 at 11:29
0

You can check cdQA python library which allows you to build custom Closed Domain Question Answering Systems without using the cloud: https://github.com/cdqa-suite/cdQA

erup
  • 183
  • 3
  • 12