search

texttron / tevatron

Tevatron - A flexible toolkit for neural retrieval research and development.
http://tevatron.ai
Apache License 2.0
435 stars 87 forks source link

train retriever #116

Open chenzhongwu opened 2 months ago

chenzhongwu commented 2 months ago

Hi! Should I write the code for generating the training datasets for the retriever by myself? What if my own datasets are hard to choose positive and negative passages? Thanks!

MXueguang commented 2 months ago

yes, for each query, the training requires at least a positive passages (human judged or heuristically selected.)