AMUSED: A Multi-Stream Vector Representation Method for Use in Natural Dialogue

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 750–758 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 750 AMUSED: A Multi-Stream Vector Representation Method for Use in Natural Dialogue Gaurav Kumar*1, Rishabh Joshi*2, Jaspreet Singh*3, Promod Yenigalla1 1Samsung Research Institute, Bangalore, 2Carnegie Mellon University, 3Stony Brook University 1{gaurav.k1, promod.y}@samsung.com, 2rjoshi2@andrew.cmu.edu, 3jaspreet.ahluwalia@stonybrook.edu Abstract The problem of building a coherent and non-monotonous conversational agent with proper discourse and coverage is still an area of open research. Current architectures only take care of semantic and contextual information for a given query and fail to completely account for syntactic and external knowledge which are crucial for generating responses in a chit-chat system. To overcome this problem, we propose an end to end multi-stream deep learning architecture which learns unified embeddings for query-response pairs by leveraging contextual information from memory networks and syntactic information by incorporating Graph Convolution Networks (GCN) over their dependency parse. A stream of this network also utilizes transfer learning by pre-training a bidirectional transformer to extract semantic representation for each input sentence and incorporates external knowledge through the neighborhood of the entities from a Knowledge Base (KB). We benchmark these embeddings on next dialogue prediction task and significantly improve upon the existing techniques. Furthermore, we use AMUSED to represent query and responses along with its context to develop a retrieval based conversational agent which has been validated by expert linguists to have comprehensive engagement with humans. Keywords: Natural Language Processing, Dialogue Systems, Knowledge Graphs, GCN, Memory Networks 1. Introduction With significant advancements in Automatic speech recog- nition systems (Hinton et al., 2012; Kumar et al., 2018) and the field of natural language processing, conversational agents have become an important part of the current research. It finds its usage in multiple domains ranging from self-driving cars (Chen et al., 2017b) to social robots and virtual assistants (Chen et al., 2017a). Conversational agents can be broadly classified into two categories: a task-oriented chatbot and a chit-chat based system, respec- tively. The former works towards completion of a certain goal and are specifically designed for domain-specific needs such as restaurant reservations (Wen et al., 2017), movie recommendation (Dhingra et al., 2017), flight ticket booking systems (Wei et al., 2018) among many others. The latter is more of a personal companion and engages in human-computer interaction for entertainment or emotional companionship. An ideal chit chat system should be able to perform non-monotonous interesting conversation with context and coherence. Current chit chat systems are either generative (Vinyals and Le, 2015) or retrieval based in nature. The generative ones tend to generate natural language sentences as responses and enjoy scalability to multiple domains without much change in the network. Even though easier to train, they suffer from error-prone responses (Zhang et al., 2018b). Retrieval based methods select the best response from a given set of answers, which makes them error-free. But, since the responses come from a specific dataset, they might suffer from distribution bias during the …

Content truncated to 500 words