Implementation of deep text matching algorithms, such as DSSM...(to be continued).
- Training set: In the training file, each line contains four parts, and word id of each part are concatenated by comma:
- Id of query
- Query
- Positive doc
- Negative docs: number of negative doc is defined in config file. And each negative sample is concatenated by semicolon