search

clintpgeorge / ediscovery

Active Topic Learning based Legal eDiscovery
6 stars 1 forks source link

Identifying responsive and non-responsive documents #4

Open clintpgeorge opened 11 years ago

clintpgeorge commented 11 years ago

Currently, we just use a threshold (e.g. 100) on the ranked documents (based on cosine distance or KL) to decide whether a set of documents are responsive or not. We need to improve this algorithm.