Search

• Proposed an encoder–decoder for joint dialog act segmentation and recognition. • Used an attentional mechanism to integrate contextual information in dialogs. • Showed that exploiting context improves recognition accuracy. • The proposed model outperforms conventional models significantly. A dialog act represents the communicative function of an utterance in a conversation, and thus provides informative cues for understanding, managing, and generating dialog. While most spoken dialog systems process user input and system output at the turn level, a single turn can consist of multiple dialog acts in human conversations. Therefore, segmenting turn-level tokens into a meaningful dialog act unit is just as important as recognizing the dialog act. Towards joint segmentation and recognition of dialog acts, we propose an encoder–decoder model featuring joint coding and incorporate contextual information by means of an attentional mechanism. The proposed encoder–decoder outperforms other models in segmentation, and the application of attentions significantly reduces recognition error rates. By combining the encoder–decoder model with contextual attention, we achieve state-of-the-art performance in the joint evaluation of dialog act segmentation and recognition. [ABSTRACT FROM AUTHOR]

Viewing dialogue management as a reinforcement learning task enables a system to learn to act optimally by maximising a reward function. This reward function is designed to induce the system behaviour required for the target application and for goal-oriented applications, this usually means fulfilling the user’s goal as efficiently as possible. However, in real-world spoken dialogue system applications, the reward is hard to measure because the user’s goal is frequently known only to the user. Of course, the system can ask the user if the goal has been satisfied but this can be intrusive. Furthermore, in practice, the accuracy of the user’s response has been found to be highly variable. This paper presents two approaches to tackling this problem. Firstly, a recurrent neural network is utilised as a task success predictor which is pre-trained from off-line data to estimate task success during subsequent on-line dialogue policy learning. Secondly, an on-line learning framework is described whereby a dialogue policy is jointly trained alongside a reward function modelled as a Gaussian process with active learning. This Gaussian process operates on a fixed dimension embedding which encodes each varying length dialogue. This dialogue embedding is generated in both a supervised and unsupervised fashion using different variants of a recurrent neural network. The experimental results demonstrate the effectiveness of both off-line and on-line methods. These methods enable practical on-line training of dialogue policies in real-world applications. [ABSTRACT FROM AUTHOR]

Dialogue act recognition is an important component of a large number of natural language processing pipelines. Many research works have been carried out in this area, but relatively few investigate deep neural networks and word embeddings. This is surprising, given that both of these techniques have proven exceptionally good in most other language-related domains. We propose in this work a new deep neural network that explores recurrent models to capture word sequences within sentences, and further study the impact of pretrained word embeddings. We validate this model on three languages: English, French and Czech. The performance of the proposed approach is consistent across these languages and it is comparable to the state-of-the-art results in English. More importantly, we confirm that deep neural networks indeed outperform a Maximum Entropy classifier, which was expected. However, and this is more surprising, we also found that standard word2vec embeddings do not seem to bring valuable information for this task and the proposed model, whatever the size of the training corpus is. We thus further analyse the resulting embeddings and conclude that a possible explanation may be related to the mismatch between the type of lexical-semantic information captured by the word2vec embeddings, and the kind of relations between words that is the most useful for the dialogue act recognition task. [ABSTRACT FROM AUTHOR]

This article introduces a new methodology to enhance an existing traditional Spoken Dialogue System (SDS) with optimal turn-taking capabilities in order to increase dialogue efficiency. A new approach for transforming the traditional dialogue architecture into an incremental one at a low cost is presented: a new turn-taking decision module called the Scheduler is inserted between the Client and the Service. It is responsible for handling turn-taking decisions. Then, a User Simulator which is able to interact with the system using this new architecture has been implemented and used to train a new Reinforcement Learning turn-taking strategy. Compared to a non-incremental and a handcrafted incremental baselines, it is shown to perform better in simulation and in a real live experiment. [ABSTRACT FROM AUTHOR]

Task-oriented dialogue systems need to take appropriate actions not only for clear user requests but also for ambiguous and vague ones. In this study, "ambiguous" denotes that although users have potential requests, they failed to clearly define and verbalize their content and conditions which can be associated with system actions. For such ambiguous requests, taking reflective actions is one plausible choice for such systems. In our study, "reflective" denotes taking actions that satisfy user requests before the users themselves clarify their demands. We constructed such a reflective dialogue agent by collecting a corpus that includes pairs of ambiguous user requests and corresponding reflective system actions on sightseeing navigation with a smartphone. Since annotating every possible combination of user requests and system actions is impossible, this study built a corpus where one reflective action is annotated to one ambiguous user request. To train an action selection model on such incomplete training data in which only one action is associated with a request, we applied the positive/unlabeled (PU) learning method, which assumes that only part of the data is labeled with positive examples. In addition, we enhanced the action selection by extracting and distilling knowledge that corresponds to causality from the training data using a causality detection model. The experimental results show that both the PU learning method and the causality detection model improved the performances of the reflective action selection compared to the conventional positive/negative (PN) learning method. • We built a reflective action selection system given ambiguous user requests. • We collected a dialogue corpus that bridges ambiguous requests to reflective actions. • We explored whether several actions can be reflective for one request on test data. • We applied the PU learning method to deal with partially annotated training data. • We distilled causality knowledge to utilize in the reflective action selection. [ABSTRACT FROM AUTHOR]

As knowledge-grounded dialogue systems are attracting intense research interest, technology that facilitates reference to various types of external knowledge as dialogue-system data is also developing. A knowledge-grounded dialogue system must be capable of (1) accurately interpret the conversation content, (2) determining whether external knowledge should be referenced for the current turn, (3) identifying the external knowledge to be referenced, and (4) generating a response. If the referenced external knowledge is in the form of a question–answer pair, this pair is closely related to the domain to which the external knowledge belongs and the entity to which it pertains. This study proposes a knowledge-grounded dialogue system that predicts the domain and entity associated with a question–answer pair referenced by the dialogue system; the system then leverages this information to effectively implement the above three external-knowledge-based capabilities. As the dialogue data associated with external knowledge is diverse, adaptation to the slots and entities of different dialogue datasets is challenging. In this study, the Triple Copy (TripPy) model, which is one of the leading benchmark models for dialogue-state tracking with the Multi-domain Wizard-of-Oz (MultiWOZ) dataset, is further developed to adapt to DSTC10 data for the external knowledge-based dialogue system; hence, dialogue content is effectively interpreted. The developed knowledge-grounded dialogue system incorporates knowledge-seeking turn detection, knowledge selection, and knowledge-grounded response generation models. The model, in DSTC10 dataset, achieves 15.49%p, 12.54%p, and 36.45%p improvements over the baseline, the state-of-the-art model for the previous version of the data (DSTC9 dataset), in terms of the F1 score, Recall@1 score, and BLEU-1 score, respectively. Moreover, in a dialogue-state tracking task for DSTC10 dataset, a 15.41%p improvement in joint goal accuracy score is achieved compared to the TripPy model. • A goal-oriented dialogue system using external knowledge to respond to a user is proposed. • Adapting a dialogue-state tracking model to targeting datasets to well understand a conversation. • The advanced knowledge-grounded dialogue system using the extracted entities and tracked domain. [ABSTRACT FROM AUTHOR]

Referring to visually perceivable objects is a very common occurrence in everyday language use. In order to produce expressions that refer, the speaker needs to be able to pick out visual properties that the referred object has and determine the words that name those properties, such that the expression can direct a listener's attention to the intended object. The speaker can aid the listener by looking in the direction of the object and by providing a pointing gesture to indicate it. In order to resolve the reference, the listener has a difficult job to do: simultaneously use all of the linguistic and non-linguistic information; the words of the referring expression that denote properties of the object, such as its colour or shape, need to already be known, and the non-linguistic gaze direction and pointing gesture of the speaker need to be incorporated. Crucially, the listener does not wait until the end of the referring expression before she begins to resolve it; rather, she is interpreting it as it unfolds. A model that resolves referring expressions as the listener must be able to do all of these things. In this paper, we present such a generative model of reference resolution. We explain our model and show empirically through a series of experiments that the model can work incrementally (i.e., word for word) as referring expressions unfold, can incorporate multimodal information such as gaze and pointing gestures in two ways, can learn a grounded meaning of words in the referring expression, can incorporate contextual (i.e., saliency) information, and is robust to noisy input such as automatic speech recognition transcriptions, as well as uncertainty in the representation of the candidate objects. [ABSTRACT FROM AUTHOR]

Abstract: Reinforcement techniques have been successfully used to maximise the expected cumulative reward of statistical dialogue systems. Typically, reinforcement learning is used to estimate the parameters of a dialogue policy which selects the system''s responses based on the inferred dialogue state. However, the inference of the dialogue state itself depends on a dialogue model which describes the expected behaviour of a user when interacting with the system. Ideally the parameters of this dialogue model should be also optimised to maximise the expected cumulative reward. This article presents two novel reinforcement algorithms for learning the parameters of a dialogue model. First, the Natural Belief Critic algorithm is designed to optimise the model parameters while the policy is kept fixed. This algorithm is suitable, for example, in systems using a handcrafted policy, perhaps prescribed by other design considerations. Second, the Natural Actor and Belief Critic algorithm jointly optimises both the model and the policy parameters. The algorithms are evaluated on a statistical dialogue system modelled as a Partially Observable Markov Decision Process in a tourist information domain. The evaluation is performed with a user simulator and with real users. The experiments indicate that model parameters estimated to maximise the expected reward function provide improved performance compared to the baseline handcrafted parameters. [Copyright &y& Elsevier]

• A novel diversity-promoting joint training network for open-domain dialogue generation. • We introduce future information during the decoding stage, thus the generation of each word can leverage complete information of the target utterance. • We devise a maximum entropy regularizer to alleviate the over-estimation of high-frequency words. • Our network introduces more linguistic information from target utterances to increase diversity, and captures direct semantic information to better constrain relevance simultaneously. • The method can be applied to any sequence-to-sequence architecture. The sequence-to-sequence (Seq2Seq) model generates target words iteratively given the previously observed words during the decoding process, which results in the loss of the holistic semantics in the target response and the complete semantic relationship between responses and dialogue histories. In this paper, we propose a generic diversity-promoting joint network, called Holistic Semantic Constraint Joint Network (HSCJN), enhancing the global sentence information, and then regularizing the objective function with penalizing the low entropy output during the training stage. Our network introduces more target information to improve diversity and captures direct semantic information to better constrain relevance simultaneously. Moreover, the proposed method can be easily applied to any Seq2Seq structure. Extensive experiments on several dialogue corpuses show that our method effectively improves both semantic consistency and diversity of generated responses, and achieves better performance than other competitive methods. [ABSTRACT FROM AUTHOR]