# Exploring the context of recurrent neural network based conversational agents

## TL;DR

This paper compares simple and hierarchical neural network models for conversational agents, revealing that while hierarchical models better capture context, they underperform in grammatical and meaningful responses, highlighting future research directions.

## Contribution

It introduces a hierarchical recurrent encoder-decoder architecture for conversational agents and analyzes its effectiveness in modeling conversation context compared to simpler models.

## Key findings

- Hierarchical model captures context but performs worse in grammar and relevance.
- Simple encoder-decoder outperforms hierarchical in response quality.
- Conversation context influences topic clustering in the embedding space.

## Abstract

Conversational agents have begun to rise both in the academic (in terms of research) and commercial (in terms of applications) world. This paper investigates the task of building a non-goal driven conversational agent, using neural network generative models and analyzes how the conversation context is handled. It compares a simpler Encoder-Decoder with a Hierarchical Recurrent Encoder-Decoder architecture, which includes an additional module to model the context of the conversation using previous utterances information. We found that the hierarchical model was able to extract relevant context information and include them in the generation of the output. However, it performed worse (35-40%) than the simple Encoder-Decoder model regarding both grammatically correct output and meaningful response. Despite these results, experiments demonstrate how conversations about similar topics appear close to each other in the context space due to the increased frequency of specific topic-related words, thus leaving promising directions for future research and how the context of a conversation can be exploited.