We have developed an active listening system for a conversation robot, specifically for reminiscing. The aim of the system is to contribute to the prevention of dementia in elderly persons and to reduce loneliness in seniors living alone. Based on the speech recognition results from a user’s utterance, the proposed system produces back-channel feedback, repeats the user’s utterance and asks information about predicates that were not included in the original utterance. Moreover, the system produces an appropriate empathic response by estimating the user’s emotion from their utterances. One of the features of our system is that it can determine an appropriate response even if the speech recognition results contain some errors. Our results show that the conversations of 45.5% of the subjects (n = 110) with this robot continued for more than two minutes on the topic “memorable trip”. The system response was deemed correct for about 77% of user utterances. Based on the results of a questionnaire, positive evaluations of the system were given by the elderly subjects.
In this paper, we present a novel picture-book search system Pitarie, which can find a picture book that matches a child’s interests and language developmental stage. By reading the appropriate picture book to children, positive effects such as faster language development and enhanced emotional education are expected. Pitarie searches are based on two new natural language processing technologies particularly designed for picture books: morphological analysis and text readability estimation for sentences written mainly in Hiragana script. In this paper, we introduce Pitarie with a focus on such novel technologies and their level of quality. Finally, we report the results of the questionnaire for the entire system. Books that were selected based on recommendations by Pitarie had an average rating of 4.44–4.54 on a 5-point evaluation scale from both children’s interest and language developmental stage viewpoints.
This paper proposes a GUI support tool for bilingual dictionary compilation and translation, called “Bilingual KWIC.” Bilingual KWIC acquires bilingual expressions from a parallel corpus and displays the result in KWIC format. Displaying in KWIC format enables users to easily correct errors of word alignment and compare two or more types of equivalents. Since Bilingual KWIC does not use morphological information and uses only character-level information, it can deal with any language pairs and can acquire translation equivalents for any input other than words. In this paper, we introduce Bilingual KWIC and describe its features and development process.
Diversity is an important indicator for improving user experience in recommender systems. Previous research indicate that people prefer diverse recommended item lists. However, few studies have experimented with online user experience of recommender systems owing to lack of clarity regarding the effects of diversity of recommender systems on user experience. This paper reports the online experience of diversity of web service recommender systems. We analyzed the recommender system without diversity for user activity in web services. As a result, the second half of the recommended list is underwhelming. We have constructed a diverse recommender system by decreasing user features, and have compared our system to the existing system for user activity in web services. Consequently, our system has succeeded in improving the weekly retention and active rates. Therefore, the number of clicks on the recommended list have increased.
We propose an Frequently Asked Question (FAQ) search method that uses a document classifier for classifying a natural language query to a corresponding FAQ. The document classifier classifies a query with words that occur in the query. However, since FAQs have little redundancy, using FAQs as training data for the document classifier is not sufficient for classifying queries that have the similar meaning but different surface expressions. To tackle this problem, our method generates training data automatically from FAQs and corresponding histories and trains the document classifier with them. Furthermore, with the automatically generated training data, our method learns a ranking model that uses classification results of the document classifier. Experimental results on a company FAQs and corresponding histories showed that our method outperformed pseudo-relevance feedback and query expansion model that uses word alignment model in statistical machine translation.
In sociology, occupation and industry variables are as important as sexual and age variables. For the purpose of statistical processing, answers collected from open-ended questions in social surveys need to be converted into code, which requires considerable time and effort and often results in inconsistencies in large scale surveys. This work deals with occupation and industry coding. In this work, we develop an automatic system using hand-crafted rules and Support Vector Machines. Our system can assign three candidate codes to an answer and estimates the confidence level of the primary predicted code for each national/international standard code sets. The system has now been released through the website of the Center for Social Research and Data Archives. The user can get the required coding result by uploading the data file in a specific format.
April 03, 2017 There had been a system trouble from April 1, 2017, 13:24 to April 2, 2017, 16:07(JST) (April 1, 2017, 04:24 to April 2, 2017, 07:07(UTC)) .The service has been back to normal.We apologize for any inconvenience this may cause you.
May 18, 2016 We have released “J-STAGE BETA site”.
May 01, 2015 Please note the "spoofing mail" that pretends to be J-STAGE.