DRMM

This is an implementation of the paper "A Deep Relevance Matching Model for Ad-hoc Retrieval". CIKM 2016.

Guide to the code / usage instruction

this code is written in c++, and the dependence of Eigen is included. The abstract model structures are implemented in c++ class( see e.g. class NN4IR in file NN4IR.cpp ). you can start training the robust04 title dataset by simply run

sh build.sh

It will download the dataset and unpack it in the directory. All parameters passed by the config file which tells where the dataset lies. if you want to run other dataset or run other hyper-parameters, you can run

./NN4IR -config config_file

Dataset download

The whole datasets are listed in this site. The datasets are also placed at the pan.baidu.com (URL:https://pan.baidu.com/s/1rniqqeOU5ZKJwrlA0Kstkg password: jmca).

Requirements

• g++ 4.7 version or above( supporting c++11 )