Skip to content

Latest commit

 

History

History
8 lines (6 loc) · 566 Bytes

README.md

File metadata and controls

8 lines (6 loc) · 566 Bytes

WikiQA-CNN

Implementation of the Adaboost algorithm for letter recognition, which is a multi classification problem. Given the problem q and the set of sentences A, identify all the answers to q in A and sort all the sentences. Put the answer before the non-answer. The answers are ranked random, and the non-answers' ranks are also free.

Based on CNN, I implemented a QA system, and the MRR and MAP on dev_set is over 0.64.

You need to prepare the needed files and put them in the "./data/trec" folder. Please refer to the files in the "./data/wiki" folder .