search

ratishsp / data2text-plan-py

Code for AAAI 2019 paper on Data-to-Text Generation with Content Selection and Planning
163 stars 46 forks source link

how do you generate the file "roto_train-beam5_gens.h5-tuples.txt" #27

Closed yangshuodelove closed 3 years ago

yangshuodelove commented 3 years ago

Hi, Thanks for the excellent paper and code. (1) What is the purpose of the file "roto_train-beam5_gens.h5-tuples.txt"? (2) How do you generate it? (3) If I use other datasets, how to generate files like it?

Thank you very much.

yangshuodelove commented 3 years ago

I found it was too hard for me to understand the code step by step without the background knowledge of OpenNMT. Should I understand the code of the model part first or learn OpenNMT in advance? Could you give me some advice?

ratishsp commented 3 years ago

Hi @yangshuodelove Thanks for the feedback! (1) What is the purpose of the file "roto_train-beam5_gens.h5-tuples.txt"? -The file is needed to provide training supervision to the content planner. (2) How do you generate it? -We generate it using an Information Extraction approach by mapping summary with the input table, identifying matching entities and values, and predicting the type of record. (3) If I use other datasets, how to generate files like it? -You may use similar IE or string matching techniques to generate such supervision.

ratishsp commented 3 years ago

I found it was too hard for me to understand the code step by step without the background knowledge of OpenNMT. Should I understand the code of the model part first or learn OpenNMT in advance? Could you give me some advice?

Yes, it is advantageous to have background of OpenNMT in advance. The code of this repo is based on a much older (0.1) version of OpenNMT. The link is https://github.com/OpenNMT/OpenNMT-py/tree/v0.1

yangshuodelove commented 3 years ago

Thank you very much. @ratishsp