Explicit State Tracking with Semi-Supervisionfor Neural Dialogue Generation

Abstract

The task of dialogue generation aims to automatically provide responses given previous utterances. Tracking dialogue states is an important ingredient in dialogue generation for estimating users' intention. However, the expensive nature of state labeling and the weak interpretability make the dialogue state tracking a challenging problem for both task-oriented and non-task-oriented dialogue generation: For generating responses in task-oriented dialogues, state tracking is usually learned from manually annotated corpora, where the human annotation is expensive for training; for generating responses in non-task-oriented dialogues, most of existing work neglects the explicit state tracking due to the unlimited number of dialogue states.In this paper, we propose the semi-supervised explicit dialogue state tracker (SEDST) for neural dialogue generation. To this end, our approach has two core ingredients: CopyFlowNet and posterior regularization. Specifically, we propose an encoder-decoder architecture, named CopyFlowNet, to represent an explicit dialogue state with a probabilistic distribution over the vocabulary space. To optimize the training procedure, we apply a posterior regularization strategy to integrate indirect supervision. Extensive experiments conducted on both task-oriented and non-task-oriented dialogue corpora demonstrate the effectiveness of our proposed model. Moreover, we find that our proposed semi-supervised dialogue state tracker achieves a comparable performance as state-of-the-art supervised learning baselines in state tracking procedure.

Conversational Recommendation: Formulation, Methods, and Evaluation

“…As core problems of dialogue systems, natural language understanding and generation are crucial in CRSs. This direction focuses on how to understand a user's preferences and intentions from their utterances and generate fluent responses so as to deliver natural and effective dialogue actions (e.g., asking question and making recommendations) [8,9,11,20]. Multiple datasets and simulation environments have been released to help push the statee-of-the-art in this area For instance, Li et al [14] release a dataset comprising of more than 10,000 dialogues on movie recommendation and Chen et al [3] incorporate a knowledge graph to bridge dialogue understanding and generation with the recommendation component.…”

Improving Limited Labeled Dialogue State Tracking with Self-Supervision

“…Wu et al (2019b) introduced a self-supervised learning task, inconsistent order detection, to explicitly capture the flow of conversation in dialogues. Jin et al (2018) use unlabeled data to train probabilistic distributions over the vocabulary space as dialogue states for neural dialogue generation. Su et al (2020) provide both supervised and unsupervised learning algorithms to train language understanding and generation models in a dual learning setting.…”

Crsal

“…The dialogue state tracker (DST) is essentially responsible for extracting the latent linguistic features from the raw input texts. In a conversational recommender system, dynamically tracking dialogue belief states is the key for understanding users' demands and generating coherent, context-sensitive responses [32]. Conventional methods for dialogue state tracking mainly employ recurrent neural networks (RNN) and convolutional neural networks (CNN) [53] to extract n-gram features from delexicalized user utterances.…”