There are more opportunities to interact with dialogue agents by the spread of smartphones and smart speakers. However, the dialogue agents in these devices lack the ability to entertain people. Dialogues with these agents are not as enjoyable as talking with a close friend. In order for dialogue agents to provide enjoyment to people, the system needs to recognize and generate humor from the context of the dialogue. At the moment, it is not clear whether humor that includes contextual information during dialogue is funny or how contextual information affects the funny. Therefore, to construct a standard database for evaluating the fun of puns considering the dialogue context, we collected dialogues containing puns on Twitter. Using the collected dialogues, multiple people evaluated the fun. Moreover, we analyzed the characteristics of a dialogue context that affects the fun of puns.
In this research, we propose a method to automatically generate sentences expressing stock price trends using stock prices and earnings reports as part of the automatic generation of Nikkei 225 market summary articles. The sentences expressing stock price trends generated by our method are as follows: “Retail, wholesale, and construction-related industries are rising. Machinery, steel, and electrical equipment-related sectors fall.” Our method estimates investment themes, such as stock market trends, businesses, products, and social backgrounds, using stock prices and earnings reports. Specifically, our method extracts groups of companies that have experienced large fluctuations in stock prices, and then refines the groups of companies by clustering based on keywords extracted from earnings reports of each company. Next, our method estimates investment themes from the keywords in the narrowed-down group of companies, and generates sentences expressing stock price trends based on the investment themes.
Local assembly minutes record all the statements made in assemblies. Studies have been conducted to analyze local issues and to clarify the efforts and political stance of the assembly members based on the statements in the minutes. Previous studies have used word-based methods such as TF-IDF, which cannot consider multi-word phrases or contexts. In this paper, we construct a BERT-based classifier that predicts the speaker of each statement in the minutes, and calculate the contribution of each token to the prediction by SHAP. We then extract characteristic expressions of political interest from the statements in the minutes on a clause-by-clause basis according to the SHAP values, and analyze the experimental results. By taking into account the dependency relations between the clauses, the context of the extracted expressions can be presented. The analysis results show that our method can extract more characteristic expressions that reveal the political interests of the speaker than TF-IDF. In addition, we confirmed that this method can extract the speaker’s unique phrases and habits of saying, which are difficult to extract with TF-IDF.
The individual physical appearances of robots are considered significant, similar to the way that those of humans are. We investigated whether users prefer robots with male or female physical appearances for use in daily communication situations depending on occupation and whether gender equality orientation are related to this preference. One thousand adult men and women aged 20–60 participated in the questionnaire survey. The results of our study showed that in most situations depending on occupation, “neither” was selected mostly. Furthermore, we examined the relationship between gender preference and gender equality orientation, and results showed that the tendency is higher for people who were not specific about their gender preferences. It is concluded that there is no need to introduce a robot that specifies its gender. Robots with a gender-neutral appearance might be more appropriate for applications requiring complex human–robot interaction and help avoid reproducing a gender bias.
Although e-Learning has advantages such as individualized and flexible program construction, education quality assurance, its passive and monotony interaction makes it difficult to sustain learning motivation. Recently, social robots taking over entire job functions are attracting a great deal of attention. Most educational robots have been developed for each learning system, intended for infant or elementary education. In this paper, we propose a motivating robot for higher education applicable to existing e-Learning systems. The efficacy of the proposed robot was verified through two-months demonstration experiment.
In the development of the fully automated vehicle, the vehicle has to be controlled to reduce the fear of autonomous driving for passengers. However, the driving situation when passengers feel a sense of fear and strangeness is not clearly explained while the vehicle runs autonomously without manual operation. Therefore, we investigate the driving situation when passengers feel a sense of fear and strangeness based on the timing and magnitude of the intervention operation by using a VR simulator that replicates fully automated driving. In this paper, experiments that the automated vehicle avoids dangerous situations are conducted. A subject can operate intervention when he/she feels fear or strangeness. We analyze the relationship between the intervention operation and manual driving operation. According to the experimental results, we discuss the operating characteristics of each person and the features of the fear of an automated driving.
In this study, we discuss the acceptability of interactive anthropomorphic agents from the perspective of a comparison between Japan and the United States. Specifically, we conduct “survey 1: questionnaire survey on acceptability of interaction with robots participating in society” and “survey 2: impression evaluation of verbal consideration assuming interaction with interactive anthropomorphic agents.” In survey 1, without specifying a specific anthropomorphic agent, we evaluate the negative attitude and uneasy impression toward robots participating in society as a whole. In survey 2, we asked the experimental participants to watch a video in which they were supposed to interact with an interactive anthropomorphic agent, and evaluated their impressions. The results of these surveys indicated that American participants tended to have more negative attitudes and anxious impressions toward anthropomorphic agents as a whole than Japanese participants, but on the other hand, it was possible to increase the acceptance of American participants in situations in which they interacted with anthropomorphic agents in a pseudo-dialogue. On the other hand, it was suggested that it is possible to increase the receptivity of American participants to anthropomorphic agents in situations in which they interact with them.