摘要 :
We present a dialogue control method called the "dual-cost method", by which a spoken dialogue system carries out an efficient dialogue within the confines of the data stored as the system's knowledge in its database. Due to speec...
展开
We present a dialogue control method called the "dual-cost method", by which a spoken dialogue system carries out an efficient dialogue within the confines of the data stored as the system's knowledge in its database. Due to speech recognition errors, a spoken dialogue system has to carry out a "confirmation dialogue" to clarify a user request. Previous methods have a problem of invoking unnecessary confirmations, since they attempt to confirm the whole contents of a user request even though the request is beyond the system's knowledge. To resolve this problem, we introduce the notions of confirmation cost and information transfer cost. The confirmation cost is the length of a confirmation dialogue and depends on the speech recognition rate. The information transfer cost is the length of a system response after the confirmation dialogue and depends on the system's knowledge. The dual-cost method controls a dialogue based on the minimization of these two costs and can avoid unnecessary exchanges, which are inevitable in conventional methods.
收起
摘要 :
This paper is an attempt to further the conversation about the possibilities of dialogue with technology that Wegerif and Major (AI Soc, https://doi.org/10.1007/s00146-018-0828-6, 2018) have initiated. In their paper Wegerif and M...
展开
This paper is an attempt to further the conversation about the possibilities of dialogue with technology that Wegerif and Major (AI Soc, https://doi.org/10.1007/s00146-018-0828-6, 2018) have initiated. In their paper Wegerif and Major (AI Soc, https://doi.org/10.1007/s00146-018-0828-6, 2018) have argued that "constructive dialogue with technology is possible, even essential, and that this takes the form of opening a dialogic space" and they also "argue against Buber that dialogic spaces do not all take the same form, but that they take a multitude of forms depending, to a large extent, on the mediating technology". Reflecting on the paper by Wegerif and Major (AI Soc, https://doi.org/10.1007/s00146-018-0828-6, 2018), the present paper attempts to highlight certain issues and possibilities for such a dialogue. In the first section of the paper, Wegerif and Major's (AI Soc, https://doi.org/10.1007/s00146-018-0828-6, 2018) discussion of technology-mediated dialogue has been reflected upon. This is followed by a reflection on the possibility and challenges of engaging in a dialogue with technology as 'other' by referring to the case of a robot 'Sophia' getting citizenship in the Kingdom of Saudi Arabia. The third section discusses certain concerns regarding the interpretation of Buber's idea of dialogue by Wegerif and Major (AI Soc, https://doi.org/10.1007/s00146-018-0828-6, 2018). The conclusive section highlights some key aspects that need to be considered while reflecting on the possibility of dialogue with technology.
收起
摘要 :
Formal dialogue systems model rule-based interaction between agents and as such have multiple applications in multi-agent systems and AI more generally. Their conceptual roots are in formal theories of natural argumentation, of wh...
展开
Formal dialogue systems model rule-based interaction between agents and as such have multiple applications in multi-agent systems and AI more generally. Their conceptual roots are in formal theories of natural argumentation, of which Hamblin’s formal systems of argumentation in Hamblin (Fallacies. Methuen, London, 1970, Theoria 37:130–135, 1971) are some of the earliest examples. Hamblin cites the medieval theory of obligationes as inspiration for his development of formal argumentation. In an obligatio, two agents, the Opponent and the Respondent, engage in an alternating-move dialogue, where the Respondent’s actions are governed by certain rules, and the goal of the dialogue is establishing the consistency of a proposition. We implement obligationes in the formal dialogue system framework of Prakken (Knowl Eng Rev 21(2):163–188, 2006) using Dynamic Epistemic Logic (van Ditmarsch etal. in Dynamic epistemic logic, Synthese Library Series. Springer, Berlin, 2007). The result is a new type of inter-agent dialogue, for consistency-checking, and analyzing obligationes in this way also sheds light on interpretational and historical questions concerning their use and purpose in medieval academia.
收起
摘要 :
The goal of dialogue management in a spoken dialogue system is to take actions based on observations and inferred beliefs. To ensure that the actions optimize the performance or robustness of the system, researchers have turned to...
展开
The goal of dialogue management in a spoken dialogue system is to take actions based on observations and inferred beliefs. To ensure that the actions optimize the performance or robustness of the system, researchers have turned to reinforcement learning methods to learn policies for action selection. To derive an optimal policy from data, the dynamics of the system is often represented as a Markov Decision Process (MDP), which assumes that the state of the dialogue depends only on the previous state and action. In this article, we investigate whether constraining the state space by the Markov assumption, especially when the structure of the state space may be unknown, truly affords the highest reward. In simulation experiments conducted in the context of a dialogue system for interacting with a speech-enabled web browser, models under the Markov assumption did not perform as well as an alternative model which classifies the total reward with accumulating features. We discuss the implications of the study as well as its limitations.
收起
摘要 :
Recently, research on open domain dialogue systems have attracted extensive interests of academic and industrial researchers. The goal of an open domain dialogue system is to imitate humans in conversations. Previous works on sing...
展开
Recently, research on open domain dialogue systems have attracted extensive interests of academic and industrial researchers. The goal of an open domain dialogue system is to imitate humans in conversations. Previous works on single turn conversation generation have greatly promoted the research of open domain dialogue systems. However, understanding multiple single turn conversations is not equal to the understanding of multi turn dialogue due to the coherent and context dependent properties of human dialogue. Therefore, in open domain multi turn dialogue generation, it is essential to modeling the contextual semantics of the dialogue history rather than only according to the last utterance. Previous research had verified the effectiveness of the hierarchical recurrent encoder-decoder framework on open domain multi turn dialogue generation. However, using an RNN-based model to hierarchically encoding the utterances to obtain the representation of dialogue history still face the problem of a vanishing gradient. To address this issue, in this article, we proposed a static and dynamic attention-based approach to model the dialogue history and then generate open domain multi turn dialogue responses. Experimental results on the Ubuntu and Opensubtitles datasets verify the effectiveness of the proposed static and dynamic attention-based approach on automatic and human evaluation metrics in various experimental settings. Meanwhile, we also empirically verify the performance of combining the static and dynamic attentions on open domain multi turn dialogue generation.
收起
摘要 :
Human-machine interfaces for spoken language require a model of dialogue structure that captures the variability and unpredictability within dialogues of a given type as well as the variation between dialogue types. We propose to ...
展开
Human-machine interfaces for spoken language require a model of dialogue structure that captures the variability and unpredictability within dialogues of a given type as well as the variation between dialogue types. We propose to use co-operating grammars as such a model. This proposal is illustrated by a small example to demonstrate its adequacy and to show how a general method for modelling dialogues, that is, a metamodel, can be established.
收起
摘要 :
We propose a novel practical dialogue management system that satisfies the requirements of robust dialogue management, efficient domain knowledge construction, and flexible architecture for maintenance and extensibility. The propo...
展开
We propose a novel practical dialogue management system that satisfies the requirements of robust dialogue management, efficient domain knowledge construction, and flexible architecture for maintenance and extensibility. The proposed system uses a corpus-based framework and a dynamic dialogue transition network model, which work in a cooperative and complementary manner. The former supports automatic generation of domain knowledge from an annotated corpus, whereas the latter manages dialogue flows robustly. The system can also automatically carry out user-intention analyses and response generation since it retrieves the most similar utterance and its response pair by estimating similarity between the input utterance and corpus utterances. Therefore, the system can control a new domain dialogue by updating the corpus. In our experiments on two different corpora, the system achieved F_(0.5)-scores of 91% and 90% in terms of user intention recognition with a task completion rate of 95% and 91%.
收起
摘要 :
This special issue of the Journal is concerned with speech and language processing issues in the overall environment of end-to-end dialogue systems, and in particular with the sorts of techniques deployed in the COMPANIONS project...
展开
This special issue of the Journal is concerned with speech and language processing issues in the overall environment of end-to-end dialogue systems, and in particular with the sorts of techniques deployed in the COMPANIONS project which most of the contributors to this issue are associated in one way or another. The aim of the COMPANIONS project was to produce multimodal dialogue agent demonstrators within four years, and the papers in this volume that originate in that project are, in effect, two year prototypes, submitted to evaluations but designed principally as platforms (separately or by a new fusion of components) for further research on the deployment of emotion modelling and of machine learning (ML) techniques of a variety of forms. As will be described, there is already some reportable ML activity in these two-year prototypes.
收起