Clause anaphora resolution of Japanese demonstrative determiner by ontology extension using distributed expression of words. - In: 13th International Conference on Signal-Image Technology and Internet-Based Systems, (2017), S. 129-136
An anaphora is a type of expression whose reference depends upon another referential element. In order to become aware of situations or human intentions by natural language understanding, it is important to estimate correctly what anaphoric expressions indicate. In particular, an anaphora resolution for Japanese demonstrative determiners is critical to be aware of situation, context or a causal relationship in dialogue. Ontology-based similarity has been proposed as a way to identify referential elements. However, a method using ontology alone cannot resolve all anaphoric relations. This paper proposes an anaphora analysis method that uses semantic similarity between different parts-of-speech and conceptual dependency structure. This method specializes in resolving clause anaphoric. The evaluation of this method showed around 2% accuracy for resolving Japanese demonstrative determiner in balanced corpus.
https://doi.org/10.1109/SITIS.2017.31
Design of counseling robot for production by 3D printer. - In: 13th International Conference on Signal-Image Technology and Internet-Based Systems, (2017), S. 56-62
Nowadays, a lot of IT personnel have psychological distress. Meanwhile, counselors to help them are lack in number. To solve the problem, we proposed a counseling agent (CA) called CRECA (context respectful counseling agent). CRECA listens to clients and promotes their reflection context respectfully, namely in a context preserving way. This agent can be enhanced using a body language called "unazuki" in Japanese, a kind of "nodding" to greatly promote dialogue, often accompanying "un-un" (meaning "exactly") of Japanese onomatopoeia. This body language is expected to significantly help represent empathy or entire approval. In this paper, the agent is integrated with such a "unazuki" or "dialog promotion nodding" robot to continue the conversation naturally or context respectfully towards clients further reflection. To realize such "unazuki", the robot nods twice at each end of dialog sentence input by clients. Here, we introduce our newly developed robot that behaves human-like by an appropriate nodding behavior. The main motivation for developing a more human-like robot was the extension of application fields from IT workers' counselling to people, who suffer from more social problems such as financial debt, or anxiety of victory or defeat. For such applications, it is often very important that the agent behaves as much as possible human-like. Finally, we present the experimental evaluation results that proves such nodding is effective in counseling.
https://doi.org/10.1109/SITIS.2017.20
Flick: Japanese input method editor using N-gram and recurrent neural network language model based predictive text input. - In: 13th International Conference on Signal-Image Technology and Internet-Based Systems, (2017), S. 50-55
Smartphone is prevalent among many people. Smartphone is used not only by personal use but also by business. However, inputting Japanese text to smartphone requires longer time than PC. For this reason, predictive input, which is suggesting next words, is important to type word efficiently. On the other hands, Recurrent Neural Networks (RNNs) are very powerful sequence models. Thus, we developed the input method editor (IME), which is using n-gram and a recurrent neural networks language model based predictive text input. This IME is aimed at decreasing actions of inputting text. The evaluation experiments show our method outperforms conventional Japanese IME in terms of amount of time.
https://doi.org/10.1109/SITIS.2017.19
Extending the SVM integration with case based restarting GA to predict solar flare. - In: 2017 IEEE International Conference on Systems, Man, and Cybernetics (SMC), ISBN 978-1-5386-1645-1, (2017), S. 1994-1999
The world and human life can be had serious impact by Unusual high solar flare. In particular, serious problems may be caused by destroy satellites, damage electric power plants due to it. Therefore, it is desirable to predict sun flare peaks. However, such prediction is difficult and often inaccurate. In previous work, we used a Support Vector Machine (SVM) as a machine learning technique to predict the sun flare intensity based on data of the past. Later, we extended this SVM by a Case Based Genetic Algorithm, because such prediction turned out to be a combinatorial challenge of patterns. Since GAs are known for a tendency to fall into local optima with no further improvement, we developed a strategy of increasing the mutation rate and/or introduced restarts, whenever such stagnation occurred. The focus of the present paper is a further extension by a technology called UFCORIN along with experimental work to estimate optimal parameters.
https://doi.org/10.1109/SMC.2017.8122911
Evaluation of a classification method for MR image segmentation. - In: 2017 IEEE International Conference on Systems, Man, and Cybernetics (SMC), ISBN 978-1-5386-1645-1, (2017), S. 1581-1586
The paper introduces a proposal for an automated magnetic resonance (MR) image segmentation called Case-Based Genetic Algorithm Location-Dependent Image Classification (CBGA-LDIC) and presents its evaluation results. This method finds an appropriate cell set towards efficient image segmentation. It uses location-dependent image classification (LDIC), which is integrated by genetic algorithm (GA) combined with case based reasoning (CB). LDIC is a local heuristic, which defines multiple location-dependent classifiers. Each classifier is trained by Gaussian mixture model. CBGA-LDIC decomposes the whole image into some cells, makes a set of cells, and then trains classifiers. The method is applied to knee bones, beca0use these bone formations are similar in their location. Therefore, good combinations of cells are useful and stored in case bases. To show, that this method produces better results that other ones and to find optimal parameters, some experiments have been performed and their results are presented in this paper.
https://doi.org/10.1109/SMC.2017.8122840
Context respectful counseling agent integrated with robot nodding for dialog promotion. - In: 2017 IEEE International Conference on Systems, Man, and Cybernetics (SMC), ISBN 978-1-5386-1645-1, (2017), S. 1540-1545
Nowadays, a lot of IT personnel have psychological distress. Meanwhile, counselors to help them are lack in number. To solve the problem, we proposed a counseling agent (CA) called CRECA (context respectful counseling agent). CRECA listens to clients and promotes their reflection context respectfully namely in a context preserving way. This agent can be enhanced using a body language called "unazuki" in Japanese, a kind of "nodding" to greatly promote dialogue, often accompanying "un-un" (meaning "exactly") of Japanese onomatopoeia. This body language is expected to significantly help represent empathy or entire approval. In this paper, the agent is integrated with such a "unazuki" or "dialog promotion nodding" robot to continue the conversation naturally or context respectfully towards clients' further reflection. To realize such "unazuki", the robot nods twice at each end of dialog sentence input by clients. The experimental evaluation proves such nodding is effective in counseling.
https://doi.org/10.1109/SMC.2017.8122833
A stochastic approach to character growth in automated narrative generation. - In: Proceedings of the Thirtieth International Florida Artificial Intelligence Research Society Conference, ISBN 978-1-57735-787-2, (2017), S. 152-157
Goal aware context respectful counseling agent. - In: 12th International Conference on Signal Image Technology & Internet-Based Systems : SITIS 2016 : 28 November-1 December 2016 : Naples, Italy : proceedings, ISBN 978-1-5090-5698-9, (2016), S. 252-257
A lot of IT workers suffer from stress in doing their work and there are a few counselors to help them. To cope with this problem, an enhanced context respectful counseling agent is proposed. This counseling agent extracts emotional words from clients' utterances throughout their dialogue to detect their changes and provides clients with such changes as dialogue summary. If no change is detected, it replies with clients' utterances' paraphrases followed by context-respectful prompts to dig/narrow problems. Especially to continue the conversation towards clients' further reflection, our newly enhanced agent restricts or relates the topics with eventual and emotional words in dialogue sentences. Namely, it asks still context-respectful but self-disclosure type questions to understand the distressed client more correctly or more sincerely. For supporting the clients' awareness of a method to achieve the goal or wish lingualized in conversation, the newly proposed counseling agent interacts by context-respectfully asking mental difficulties to achieve wish or goal. This promotes clients' reflection to obtain a broader view and thus conversation time for problem solving can be shorten. This introduces clients' more speedy awareness and enables the decrease of clients' mental and physical burden. Owing to this enhancement, clients keeping and deepening reflection on themselves without boring in conversation, reach more problem clarification and self-awareness, which enables them to solve their more difficult problems.
https://doi.org/10.1109/SITIS.2016.48
An efficient classification method for knee MR image segmentation. - In: 12th International Conference on Signal Image Technology & Internet-Based Systems : SITIS 2016 : 28 November-1 December 2016 : Naples, Italy : proceedings, ISBN 978-1-5090-5698-9, (2016), S. 36-41
Aiming at application to automated recognition of knee bone magnetic resonance (MR) images, an evolutional classification method called CBGA-LDIC is proposed. CBGA-LDIC finds an appropriate cell set towards efficient image segmentation. This method uses location-dependent image classification (LDIC), which is integrated by genetic algorithm (GA) combined with case based reasoning (CB). LDIC introduces a new but local heuristics for image segmentation, and defines multiple classifiers dependent on location. Each classifier is trained by Gaussian mixture model. CBGA-LDIC decomposes the whole image into some cells, makes a set of cells, and then trains classifiers. Since the knee bones and/or their formations are similar in their location, good combinations of cells seem useful for other clients and are stored in case bases. Thus this method is expected to produce the better results when good combinations of cells are selected from cases as initial individuals of GA, especially through its repetition on restarting GA. This is verified by some experimentations shown in this paper.
https://doi.org/10.1109/SITIS.2016.15
Enhanced IoT-aware online shopping system. - In: 12th International Conference on Signal Image Technology & Internet-Based Systems : SITIS 2016 : 28 November-1 December 2016 : Naples, Italy : proceedings, ISBN 978-1-5090-5698-9, (2016), S. 31-35
In online business, it is important to construct sale web pages offering attractive services for popular products in order to improve page access as well as purchase rates. Moreover, online shop owners need to hold various types of sales frequently throughout the year to keep customers coming back. Also, online shopping systems have to adapt to its circumstances such as customers' needs and the surrounding economic situations. To cope with this, a self-organized IoT aware system is proposed. Here, awareness is achieved by monitoring / analyzing the data of user's behavior such as gazing, listening, smelling etc. The focus here is the analysis of such as users' eye gaze, listening /smelling activity etc. The aim is to derive insights about products the user is interested in or not and to adapt the system accordingly. This is achieved by presenting product information including sales associates' presence, their voice, and product's smell or even other products with similar attributes, by hiding products with less interesting attributes and by building a user's interest model through integrating all the positive preferences.
https://doi.org/10.1109/SITIS.2016.14