This repo contains the tensorflow implementation of the paper “Question Answering on Knowledge Bases and Text using Universal Schema and Memory Networks”.

Dependencies

TensorFlow <= 0.12

Training

I have set up training with most default params on a very small dataset so that it is easier to get started. Just running the script should work.

/bin/bash run.sh ./config.sh

Data

The processed data (train/dev/test split) is stored in data_formatted/ directory. To download the KB files used for the project run,

sh get_data.sh

After downloading the data, you will have to change the appropriate entries in the config.sh file (kb_file and text_kb_file).

Pretrained embeddings

The embeddings used for initializing the network can be downloaded from here

Model outputs

We are also releasing the output predictions of our model for comparison. Find them in the model_outputs directory.

Trained Model

We are also sharing our pretrained model. Get it here. The following will load the model and get the answers from the dev set. Please change the config appropriately.

sh run.sh ./test_from_saved_model.sh