Simple first stage retrieval #38
Description
A basic program to perform first stage retrieval. Efficiency is a non-goal. First stage efficiency can be found in software like PISA.
The win here would be to use any of the scoring functions that are available in the toolkit as a first stage ranker. For example, once BM25F is implemented a first stage run can be produced with it.
Main features:
- Options to set queries and index, top-k results
- Select the ranking feature to use (e.g. QL, BM25, etc)
- Dumps a run file in TREC format
- Choice of ranking feature may determine which parts of the index are needed to load
- Index will need to store docno, docid mapping