An End-to-End Goal-Oriented Dialog System with a Generative Natural Language Response Generation

TL;DR

This paper enhances goal-oriented dialog systems by incorporating positional encodings for better understanding and using a feedforward neural network for flexible, efficient response generation, improving accuracy and resource usage.

Contribution

It introduces positional encodings and a feedforward neural network for end-to-end dialog systems, overcoming previous limitations of fixed response candidates and ignoring word order.

Findings

Improved accuracy on Dialog bAbI Tasks

Reduced computation time and space consumption

Flexible response generation without fixed candidate set

Abstract

Recently advancements in deep learning allowed the development of end-to-end trained goal-oriented dialog systems. Although these systems already achieve good performance, some simplifications limit their usage in real-life scenarios. In this work, we address two of these limitations: ignoring positional information and a fixed number of possible response candidates. We propose to use positional encodings in the input to model the word order of the user utterances. Furthermore, by using a feedforward neural network, we are able to generate the output word by word and are no longer restricted to a fixed number of possible response candidates. Using the positional encoding, we were able to achieve better accuracies in the Dialog bAbI Tasks and using the feedforward neural network for generating the response, we were able to save computation time and space consumption.