The evaluation of conversational systems that chat with people remains an open-problem. Some studies have evaluated them by hand with ordinal scales like the Likert scale. One limitation with this approach is that we cannot use the previously evaluated values since the ordinal scales are not consistent across all of the evaluations. This makes it difficult to compare proposed and previous systems since we have to implement the previous systems and simultaneously evaluate them. We propose an automatic evaluation method for conversational systems that evaluates the sentences generated by systems on the basis of the similarities that are calculated with many reference sentences and their annotated evaluation values. Our proposed method's correlation coefficient with humans reached 0.514, and that of the human annotators was 0.783. Although there remains a gap between the estimated and the human-annotated values, the proposed method outperforms a baseline method that uses the BLEU scores as the evaluation values. We also show that we can gain a correlation coefficient of 0.499 with evaluating just 7% of all the data.

This paper proposes a multi-modal corpus that will consist of audio-visual data, transcribed speech, and annotated body movements of both the science communicators and the visitors at the National Museum of Emerging Science and Innovation (Miraikan) that will be active as they talk throughout the exhibition floor. This multi-modal corpus can be used in several areas such as interaction analysis, natural language processing and video processing. In this paper, the authors show how 1) spontaneous audio-visual data of conversations taking place in the exhibition floor are recorded, 2) utterances and annotated body movements of both science communicators and visitors are transcribed and/or annotated, and 3) how the multi-modal corpus will be used as a means of achieving each purpose in several fields. It is believed that researchers in numerous fields will enjoy improved opportunities for collaborative research by getting to know one another better through the use of this multi-modal corpus, which is expected to become publicly available in the spring of 2015.

The purpose of this study consists of the following two points: 1) to describe the way in which ``compliment'' is introduced and developed in the first meeting, 2) to reconsider the definition of compliments that previous studies have suggested as making the recipient to feel good. The analysis shows that compliments can be used not only as a means to make the recipient to feel good but also as a means to move the conversation forward by minimizing silences which are likely to occur in the first meeting.

The present study examined the topic introduction and its development in Japanese dyadic conversation between new acquaintances. In order to examine how the practice of topic introduction and its development might vary depending on the interlocutors' perceived age and gender, we examined the conversation data in which three base subjects were asked to interact with six different partners who differ in age and gender. The main results are as follows. 1) The older speaker introduces topics more frequently in general. 2) The older speaker introduces hearer-oriented topics in relatively higher rate than speaker-oriented topics. 3) The speaker and interlocutor collaboratively develop a topic regardless of age and gender relationship between speakers. 4) There are four sub-categories in `topic development,' that is, `delving,' `focusing,' `clarifying' and `co-constructing.' 5) `Delving' is observed most frequently within the four types of `topic development.' 6) Incomplete utterances in 'topic development' are used more frequently toward older interlocutors. These findings can be applied to the study for making more natural utterances and topic-developments, by the computer in a human computer interaction.

This paper address the way how persons establish the enviroment for telling with body's configurations in sitting conversation. We focus on the combination of the resting hands' position and analyses tow cases. i) In Case I, espacially participant, whose the position of the resting hands that is different from the other participants', became teller. ii) In Case I, espacially participant, whose the position of the resting hands that is different from the other participants', did not become teller. Through examining these cases, we sugget that participant use the combination of rest hand position as resouce for establishing the enviroment for telling.

I work on conversation analysis of interactions in kosodate hiroba in present-day Japan. Mothers take their very young children to kosodate hiroba, where they spend time watching them, sharing their problems with and getting advice from kosodate hiroba staff members and other mothers, and enjoy conversations. Based on long-term participant observation and interviews in two kosodate hiroba in Osaka, I analyze videotaped conversations among mothers and staff members and identify, in detail, several conversational ``devices'' and ``procedures,'' many of which concern mothers' ``troubles-talk'' about childcare. This study can contribute to the sociology of the family and to the study of child and family welfare/well-being, as well as to the rapidly developing field of conversation analysis regarding talk in institutional settings, in the sense that it investigates the actual interactional processes through which so-called childcare anxieties are reduced and social networks for childcare are constructed in kosodate hiroba.