In the MNIST GuessNumber dataset, each sample consists of an image (left), a set of sequential questions with answers (right), and a target digit. The goal of this game is to find out the target digit through a multi-round question-answering.
This repository is based on Python3 and generates the MNIST-GuessNumber dataset.
The code was developed by Rui Zhao (Siemens AG & Ludwig Maximilian University of Munich).
The creation of MNIST-GuessNumber dataset is inspired by MNIST-Dialog dataset.
MNIST-GuessNumber dataset is a lightweight goal-oriented visual dialog testbed.
It is for quick test of Reinforcement Learning (RL) algorithms in dialog settings.
We designed this MNIST guess-number game and used it in our paper "Efficient Dialog Policy Learning via Positive Memory Retention".
The paper is published in 2018 IEEE Spoken Language Technology (SLT), link: https://ieeexplore.ieee.org/document/8639617.
The preprint version of the paper is avaliable at: https://arxiv.org/abs/1810.01371
This work was also presented at NIPS 2018 Visually Grounded Interaction and Language (ViGIL) Workshop.
python generate_dataset.py
Then you can find the generated images and dialogs in the data folder :)
Citation of the paper:
@inproceedings{zhao2018efficient,
title={Efficient Dialog Policy Learning via Positive Memory Retention},
author={Zhao, Rui and Tresp, Volker},
booktitle={2018 IEEE Spoken Language Technology Workshop (SLT)},
pages={823--830},
year={2018},
organization={IEEE}
}
MIT