Given a user query and candidate passages corresponding to each, the task is to mark the most relevant passage which contains the answer to the user query.
The problem is in the domain of Machine Reading Comprehension (MRC). MRC is an area of research in AI which is about answering a question, given a passage of text containing the answer. As search engines evolve to respond to speech inputs and as usage ambient devices like speakers grow in the society etc. returning 10 blue links to a search query is not always desirable. For instance, try searching “what is cricket ball made of” in Bing.com and a direct answer pops up. There is a lot of progress that still needs to be made in this field, and this is one of the exciting areas of research.
I am looking for some state-of-the-art approaches for this problem. Thanks in advance. This is a part of Microsoft's AI challenge 2018 (https://microsoft.com/en-in/campaign/artificial-intelligence/microsoft-ai-challenge-india-2018.aspx)
Nguyen, T., Rosenberg, M., Song, X., Gao, J., Tiwary, S., Majumder, R., & Deng, L. (2016). MS MARCO: A human generated machine reading comprehension dataset. arXiv preprint arXiv:1611.09268.
Hu, M., Peng, Y., Huang, Z., Qiu, X., Wei, F., & Zhou, M. (2017). Reinforced Mnemonic Reader for Machine Reading Comprehension. arXiv preprint arXiv:1705.02798.
Sun, Y., Guo, D., Tang, D., Duan, N., Yan, Z., Feng, X., & Qin, B. (2018). Knowledge Based Machine Reading Comprehension. arXiv preprint arXiv:1809.04267.
Tan, C., Wei, F., Zhou, Q., Yang, N., Lv, W., & Zhou, M. (2018, August). I Know There Is No Answer: Modeling Answer Validation for Machine Reading Comprehension. In CCF International Conference on Natural Language Processing and Chinese Computing (pp. 85-97). Springer, Cham.
Thanks Sobhan. But it would be more helpful if you could tell who will you approach this problem.
Dear Sayak Paul ,
Look the link, maybe useful.
https://hanxiao.github.io/2018/09/09/Dual-Ask-Answer-Network-for-Machine-Reading-Comprehension/
http://tcci.ccf.org.cn/summit/2017/dlinfo/003.pdf
Regards, Shafagat
Hello,
The best parameter in comprehension is the keywords that appears in a sentence. Your system or technique should contain a search engine that classifies sentences by main keywords, next an NLP algorithm could help you analyse the sentence.
Example, if you have a paragraph containing a senetence like. A farmer is eating an apple.
And the question is, what is the farmer eating?
1. Your engine should retreave all the sentences containing keywords, farmer, eating.
2. It should analyse the sentences and locate specifically the apple, to be your answer.
Best regards,
you need a semantic network. If you can find a path in between the keywords you will have a representation of the sense in the text. A generation of answers is now possible.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence