The source codes of the paper "Complex Factoid Question Answering with a Free-Text Knowledge Graph (WWW 2020)".
Run python setup.py develop to install required dependencies for DELFT.
For each experimented dataset, we provide processed graph input here, after downloading, unzip it and put into data folder (the tokenized version is for glove setting, while the other is for bert setting). We also provide trained model here, unzip the downloaded model and put into experiments folder. Also, the original datasets are here.
For delft glove, downloading pre-trained glove embeddings are required, the link is here.
The experiments include DELFT-Bert and DELFT-Glove , with different embeddings (Glove or Bert), each experiment has a seperate folder (see readme.md on each folder).
If you have questions, suggestions and bug reports, please email chenz@cs.umd.edu.