Abstract In this paper an emotional humanoid robot based on Latent Semantic Analysis is presented. The robot is capable of interacting and entertain human users through the exhibition of spontaneous and non-repetitive emotional behaviours. The Latent Semantic Analysis (LSA) paradigm, used to encode the semantics of words through a statistical analysis of a large corpus of text, is employed to build... In this paper an emotional humanoid robot based on Latent Semantic Analysis is presented. The robot is capable of interacting and entertain human users through the exhibition of spontaneous and non-repetitive emotional behaviours. The Latent Semantic Analysis (LSA) paradigm, used to encode the semantics of words through a statistical analysis of a large corpus of text, is employed to build an emotional conceptual space that simulates the emotional associative capabilities of human beings, through “Latent Semantic Behaviours”. The LSA paradigm integrates traditional knowledge representation and intuitive capabilities provided by geometric and sub-symbolic information modelling. The effectiveness of the proposed approach has been validated by implementing the system with the humanoid robot from Aldebaran, Nao.12. Read More Read Less

Abstract Visually grounded human-robot interaction is recognized to be an essential ingredient of socially intelligent robots, and the integration of vision and language increasingly attracts attention of researchers in diverse fields. However, most systems lack the capability to adapt and expand themselves beyond the preprogrammed set of communicative behaviors. Their linguistic capabilities are still far from being satisfactory which make them... Visually grounded human-robot interaction is recognized to be an essential ingredient of socially intelligent robots, and the integration of vision and language increasingly attracts attention of researchers in diverse fields. However, most systems lack the capability to adapt and expand themselves beyond the preprogrammed set of communicative behaviors. Their linguistic capabilities are still far from being satisfactory which make them unsuitable for real-world applications. In this paper we will present a system in which a robotic agent can learn a grounded language model by actively interacting with a human user. The model is grounded in the sense that meaning of the words is linked to a concrete sensorimotor experience of the agent, and linguistic rules are automatically extracted from the interaction data. The system has been tested on the NAO humanoid robot and it has been used to understand and generate appropriate natural language descriptions of real objects. The system is also capable of conducting a verbal interaction with a human partner in potentially ambiguous situations. Read More Read Less

Abstract The issues of humor and emotions are strongly related to each other. Their role is recently being appreciated by HCI researchers, and numerous scientific ventures are launched to investigate this subject from various perspectives. Among others, research projects exist focusing on implementing humor generators into conversational systems, in order to facilitate their interaction with humans. In such research, evaluation experiments... The issues of humor and emotions are strongly related to each other. Their role is recently being appreciated by HCI researchers, and numerous scientific ventures are launched to investigate this subject from various perspectives. Among others, research projects exist focusing on implementing humor generators into conversational systems, in order to facilitate their interaction with humans. In such research, evaluation experiments are often focused, if not limited to, examining users’ emotive reactions towards systems (both during and after the interactions). However, in HHI (Human- Human Interaction),humor is often used, be it intentionally or subconsciously, in reaction to interlocutor’s emotions, in order to change them from negative to positive, as, for instance, anxiety into joy, or at least reduce the degree of negativity . This was showed in numerous research, mostly in the field of psychology. Thus, it can be said that in HHI, both emotions preceding and following humorous acts are taken into consideration, while most research in HCI focus only on the latter, thus representing what we call a “twostage approach” to humor and emotions. In this paper we first describe current state of the art in the field of research on humor and emotions, propose to distinguish two types of approach to relation between these two issues, next we compare existing research on this matter in HHI (psychology) and HCI, and describe our research progress. In our project, launched to construct a humor-equipped conversational system, we developed a system, which 1) analyses users’ emotions during the interaction, 2) uses humor as a reaction to users’ particular emotional states, and 3) analyses their reactions to it. Thus, the system represents what we defined as a “multi-stage approach” to humor and emotions, taking into consideration users’ emotive states both before and after humorous stimuli. We present the results of experiments conducted to evaluate our system’s performance, and propose an idea of a user-adapting humor sense model, in which the humor-emotion chain can be extended even further. Read More Read Less

Abstract This paper proposes a political question generation method that includes Japanese noun phrases of the form N1 no N2. We focus on generating a yes-no question whether a user is interested in the utterance of a councilor. An example of a political question is as follows: nisankatanso(N1) no haisyutu(N2) nituite kuwasiku sir- itai desu ka? (Do you want to know... This paper proposes a political question generation method that includes Japanese noun phrases of the form N1 no N2. We focus on generating a yes-no question whether a user is interested in the utterance of a councilor. An example of a political question is as follows: nisankatanso(N1) no haisyutu(N2) nituite kuwasiku sir- itai desu ka? (Do you want to know more about emission levels of carbon dioxide?) In this paper, N1 and N2, mean a noun or a com- pound noun. The form N1 no N2 which includes the Japanese post- positional no has a much broader usage. These expressions of the form N1 it no N2 cannot often use a question expression for asking the interesting such as simin no minasan (All of citizens). Therefore, our method checks if an expression is whether a user can understand a question expression. Moreover, our method expands the expression N1 when a question expression of the form N1 no N2 is ambiguous. The experiment results yielded good. Read More Read Less

Abstract One of the burning problems lately in Japan has been cyber-bullying, or slandering and bullying people online. The problem has been especially noticed on unofficial Web sites of Japanese schools. Volunteers consisting of school personnel and PTA (Parent-Teacher Association) members have started Online Patrol to spot malicious contents within Web forums and blogs. In practise, Online Patrol assumes reading through... One of the burning problems lately in Japan has been cyber-bullying, or slandering and bullying people online. The problem has been especially noticed on unofficial Web sites of Japanese schools. Volunteers consisting of school personnel and PTA (Parent-Teacher Association) members have started Online Patrol to spot malicious contents within Web forums and blogs. In practise, Online Patrol assumes reading through the whole Web contents, which is a task difficult to perform manually. With this paper we introduce a research intended to help PTA members perform Online Patrol more efficiently. We aim to develop a set of tools that can automatically detect malicious entries and report them to PTA members. First, we collected cyber-bullying data from unofficial school Web sites. Then we performed analysis of this data in two ways. Firstly, we analysed the entries with a multifaceted affect analysis system in order to find distinctive features for cyber-bullying and apply them to a machine learning classifier. Secondly, we applied a SVM based machine learning method to train a classifier for detection of cyber-bullying. The system was able to classify cyber-bullying entries with 88.2% of balanced F-score. Read More Read Less

Abstract We begin this paper by putting forward the topic of human conscience as a metaphysical experience. We present our ongoing research on moral reasoning categories and make first attempts to verify their usefulness in creating an agent with a dynamic algorithm for moral reasoning. Our approach assumes creating such an agent basing on two factors, the idea of wisdom of... We begin this paper by putting forward the topic of human conscience as a metaphysical experience. We present our ongoing research on moral reasoning categories and make first attempts to verify their usefulness in creating an agent with a dynamic algorithm for moral reasoning. Our approach assumes creating such an agent basing on two factors, the idea of wisdom of web-crowd and emotion-buttressed reasoning. We present a novel approach to the idea of ethics. Instead of the usual non-cognitive one we propose a model with computational structure and discuss applicability of this approach. Finally, we present some of the first results of a preliminary experiment performed to prove our approach. Read More Read Less

Abstract Affect detection from open-ended text-based dialogue and contextual profiles is challenging but essential for the building of effective intelligent user interfaces. In this paper, we report updated developments of an affect detection model from text, including affect detection from two particular types of metaphorical affective expressions (food and cooking metaphors) and affect detection based on context. We use Markov chains... Affect detection from open-ended text-based dialogue and contextual profiles is challenging but essential for the building of effective intelligent user interfaces. In this paper, we report updated developments of an affect detection model from text, including affect detection from two particular types of metaphorical affective expressions (food and cooking metaphors) and affect detection based on context. We use Markov chains and dynamic algorithm to simulate the improvisational mood of a particular character for contextual affect analysis and explore an alternative approach for emotional context modelling in a comparatively complex story context. The overall affect detection model has been embedded in an intelligent conversational AI agent interacting with human users under loose scenarios. Evaluation for the updated affect detection component is also provided. Our work contributes to the journal themes on human-agent interaction, affective computing and dialogue based systems. Read More Read Less

Abstract We have created an automatic humor generation system for Japanese that generates two man comedy routines or humorous responses to free text input. Evaluating humor is rather difficult since humor is subjective and many factors influence the perceived funniness. We evaluated our system in several ways. First, a traditional evaluation with evaluators ranking comedy performances from 1 (boring) to 5... We have created an automatic humor generation system for Japanese that generates two man comedy routines or humorous responses to free text input. Evaluating humor is rather difficult since humor is subjective and many factors influence the perceived funniness. We evaluated our system in several ways. First, a traditional evaluation with evaluators ranking comedy performances from 1 (boring) to 5 (funny) gave a result of 3.3. To complement this evaluation and to see if 3.3 is good enough for real world usage we entered a comedy performance created using our system in a funny robot competition with a ¥500,000 prize.We did not win the ¥500,000, but we made it to the final and could use the audience reactions during the live performance for evaluation. We also did sentiment analysis of blog postings covering the competition. These evaluations, while informal, indicate that our system performs competitively compared to the other contestants, that were all handmade systems. That the system could compete against human made contributions shows that a score of 3.3 can be “good enough” for real world applications. We believe that evaluations like these can be useful despite being very informal, since they measure what we want to know: do people expecting something funny think the system is funny? Read More Read Less