In a chat-oriented dialogue system, if the topic of the system's utterance does not attract the interest of the user, it is desirable that the system switches the topic. In this paper, for the system to estimate the appropriate timing of switching the topic, we propose a method for estimating users' desire to continue conversation on current topics. Our proposed method improved the estimation accuracy by 13.7 point from the baseline that constantly estimates that the user has no desire to continue conversation on the current topic.

This paper proposes a spoken dialogue system that can deliver information effectively using utterance plans created prior to conversation. The effectiveness is achieved with the system's ability to present the core content and further provide the details upon user's responses. To enable this interactive capability, the system generates two kinds of plans in advance: the primary plan for delivering the core content of an article, and the subsidiary plans that organize detailed information that could answer user's possible inquiries. This paper presents how the system makes the plans, and discusses the effectiveness by showing the example dialogue between the system and a user.

Consider the body image that does not depend on the integration of the senses from the developmental disorder research.

An analysis of conversations in business plan consultation sessions was conducted in terms of the concern-alignment model of consensus-building dialogues. we argue that the concernalignment model is effective in capturing the unfolding of consensus-building processes at the level of expression and exchange of underlying purposes and preferences of conversation participants. A set of high-level dialogue acts for proposals and concerns were introduced, and dialogue structures in consultation conversations were described as joint exploration in the space of possible concerns to seek optimal business plan through consultation sessions.

There are many scenes where decision making is necessary for without intervening by the facilitators in daily talks like discussions in office meetings and in the school activities. Although a lot of business books and tehnical books about facilitation are published, it is not found concretely and substantially what and how a facilitator perform in a decision-making scene. We recorded video and audio data how the experienced facilitators build consensus as parties concerned of the decision-making. We analyzed a consensus building process for the recorded data of the talks of the facilitators and acquire some informations.

We present a multimodal analysis of storytelling performance in group conversation as evaluated by external observers. A new multimodal data corpus, including the performance score of participants, is collected through group storytelling task. We extract multimodal features regarding explanators and listener from a manual description of spoken dialog and from various nonverbal patterns. We also extract multimodal co-occurrence features, such as utterance of explanator overlapped with listener's back channel. In the experiment, we modeled the relationship between the performance indices and the multimodal features using machine learning techniques. Experimental results show that the highest accuracy is 82% for the total storytelling performance (sum of score of indices) obtained with a combination of verbal and nonverbal features in a binary classification task.

In this paper, we report the activity of a preparatory project for building a large-scale corpus of conversational Japanese (NINJAL collaborative research project, 2014/7/1-2015/8/31). The aims of this project are to establish i) a corpus design for collecting various kinds of everyday conversations in a balanced manner, ii) methodology of recording naturally occurring conversations, and iii) a transcription system suitable for efficiently transcribing natural conversations. We first describe the survey of everyday conversational behavior, which was conducted last year for the corpus design. We then discuss how to record naturally occurring conversations, focusing on technological and ethical issues.