terryqj0107/gecor

Source code and dataset for the paper "GECOR: An End-to-End Generative Ellipsis and Co-reference Resolution Model for Task-Oriented Dialogue" (Jun Quan, Deyi Xiong, Bonnie Webber and Changjian Hu. EMNLP, 2019)

gecor's Issues

Repetitive word generation in the user utterances output

Hi,

When I am trying validation for small questions in the GECOR model and I am getting repeated words in the final output. I will provide couple of examples:

1:
user : What is the price of pizza?
rewritten question : What is the price of price of price of price of pizza?

2:
user : What is the capital India?
It is taking it as :
user : What is the capital ? even though India is in the vocab file I have given.
rewritten question : what is the capital of capital of name_SLOT ?

This is happening for most of the question I am trying out.

Running using other dataset

Is it possible to load model trained on CamRest676 and generate output for other datasets? For example CoQA dataset.

Missing vocab

Hi,
I cannot find CamRest676_annotated_complete.pkl in the repository. Thanks!