Exploring the context of recurrent neural network based conversational agents
This work addresses the challenge of context handling in conversational AI for researchers, but it is incremental as it builds on existing neural network methods with limited performance gains.
This paper tackled the problem of building non-goal driven conversational agents by comparing a Hierarchical Recurrent Encoder-Decoder with a simpler Encoder-Decoder model to handle conversation context. The result showed that the hierarchical model performed 35-40% worse in generating grammatically correct and meaningful responses, but it could extract relevant context information and group similar topics in context space.
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.