Learning to Diversify Search Result via Subtopic Attention

Overview

This page provides the data and the code of the Document Sequence with Subtopic Attention (DSSA) model proposed in this paper:

                @inproceedings{Jiang:17SIGIR:DSSA,
                  author = {Jiang, Zhengbao and Wen, Ji-Rong and Dou, Zhicheng and Zhao, Wayne Xin and Nie, Jian-Yun and Yue, Ming},
                  title = {Learning to Diversify Search Results via Subtopic Attention},
                  booktitle = {Proceedings of the 40th SIGIR},
                  year = {2017},
                }
            

Paper

Please download the paper and the slides

Data

The data required to reproduce the experimental results can be downloaded from data_cv.tar.gz. This compressed file encompasses training and testing samples, relevance features, and embeddings of documents and queries.

Code

The code and usage are available at https://github.com/jzbjyb/DSSA. The README file clearly illustrates how to run the code to reproduce the experimental results. It also explains the format of the input files and some implementation details.