Benchmark for Complex Answer Retrieval

Qrels and runs are available here.

 

Safe-environment used for Table 1 is here. Each line is structured like this:

query-id \t query \t paragraph-id \t paragraph \n

 

Candidate-set used for Table 3 is here. Each line is structured like this:

query-id \t query \t paragraph-id \t paragraph \n

 

When you use this data, remember to cite:

Nanni F., Mitra B., Magnusson M., Dietz L. Benchmark for Complex Answer Retrieval. ICTIR 2017.