For all tokens starting with idx_ , please refer to the vocabulary file for the corresponding word.

For all train/valid/test files, format is same, with the difference in the answer pool size :

<Domain><TAB><QUESTION><TAB><Groundtruth><TAB><Pool>

For InsuranceQA.question.anslabel.*:

<Domain><TAB><QUESTION><TAB><Groundtruth>

For InsuranceQA.label2answer.*

<Answer Label><TAB><Answer Text>

InsuranceQA.question.anslabel.token.1000.pool.solr.test

Include ground truth: 1744 no ground truth: 256 recall: 0.872

InsuranceQA.question.anslabel.token.1000.pool.solr.train

Include ground truth: 11069 no ground truth: 1820 recall: 0.858794320739

InsuranceQA.question.anslabel.token.1000.pool.solr.valid

Include ground truth: 1721 no ground truth: 279 recall: 0.8605

InsuranceQA.question.anslabel.token.100.pool.solr.test

Include ground truth: 1323 no ground truth: 677 recall: 0.6615

InsuranceQA.question.anslabel.token.100.pool.solr.train

Include ground truth: 8311 no ground truth: 4578 recall: 0.644813406781

InsuranceQA.question.anslabel.token.100.pool.solr.valid

Include ground truth: 1278 no ground truth: 722 recall: 0.639

InsuranceQA.question.anslabel.token.1500.pool.solr.test

Include ground truth: 1799 no ground truth: 201 recall: 0.8995

InsuranceQA.question.anslabel.token.1500.pool.solr.train

Include ground truth: 11428 no ground truth: 1461 recall: 0.886647528901

InsuranceQA.question.anslabel.token.1500.pool.solr.valid

Include ground truth: 1776 no ground truth: 224 recall: 0.888

InsuranceQA.question.anslabel.token.500.pool.solr.test

Include ground truth: 1625 no ground truth: 375 recall: 0.8125

InsuranceQA.question.anslabel.token.500.pool.solr.train

Include ground truth: 10391 no ground truth: 2498 recall: 0.806191325937

InsuranceQA.question.anslabel.token.500.pool.solr.valid

Include ground truth: 1592 no ground truth: 408 recall: 0.796