search

PhilippChr / CONVEX

Code for our CIKM 2019 paper. As far as we know, CONVEX is the first unsupervised method for conversational question answering over knowledge graphs. A demo and our benchmark (and more) can be found at
https://convex.mpi-inf.mpg.de/
MIT License
28 stars 9 forks source link

The existentials question #9

Closed 2213500360 closed 3 years ago

2213500360 commented 3 years ago

Hello,After reading your paper and coding, I have a question. There are many existentials questions in your dataset, such as "Is the rain song and immigrant song there?",so How did you get the answer to this kind of questions? I found the following code in your coding. ranked_answers = [{'answer': "yes", 'answer_score': 1.0, 'rank': 1}, {'answer': "no", 'answer_score': 0.5, 'rank': 2}] I didn't understand it. Could you explain it？thinks.

PhilippChr commented 3 years ago

Hello, thanks for your interest in our work! In our methodology, we do not have an explicit mechanism for handling existential questions. Therefore, we identify existential questions using keywords, and return "Yes" as the default answer. Still, existential questions are valid questions in a conversation, and occurred naturally in our user study. We hope that in future research, methods for dealing with existential questions are developed, which would ultimately improve the performance by quite some margin.

2213500360 commented 3 years ago

Thank you for your explanation, but I have another question. How do you solve the quantity questions like "how many"?

PhilippChr commented 3 years ago

CONVEX uses Wikidata as information source, and Wikidata has such information for many entities. E.g. the Game of Thrones (https://www.wikidata.org/wiki/Q23572) entity has the information: Game of Thrones - number of seasons - 8

So such questions are treated in the standard way outlined in the paper.

2213500360 commented 3 years ago

Thank you for answer!