Tadahiko Kumamoto

Proposal of impression mining from news articles

Each word in a language not only has its own explicit meaning, but also can convey various impressions. In this paper, we focus on the impressions people get from news articles, and propose a method for determining impressions of these news articles. Our proposed method consists of two main parts. One part involves building an “impression dictionary” that describes the relationships among words and impressions. The other part of the method involves determining impressions of input news articles using the impression dictionary. The impressions of a news article are represented as scale values in user-specified impression scales, like “sad – glad” and “angry – pleased”. Each scale value is a real number between 0 and 1, and is calculated from the words (common nouns, action nouns, verbs, adjectives, and katakana characters) extracted from an input news article using the impression dictionary.

Design and evaluation of a music retrieval scheme that adapts to the user's impressions

We have developed a scheme for music retrieval that adapts to the users impressions of the musical pieces. First, we conducted impression-estimation experiments in which 100 subjects gave their impression of 80 musical pieces, and then, using a clustering method, we classified the 100 subjects into 20 groups based on the results. Next, we created a user model for each group consisting of formulas for numerically expressing the impressions and a set of vectors calculated using the formulas. We then developed a procedure for identifying the most suitable model for an unidentified user. Testing of the models and procedure in an existing impression-based music-retrieval system demonstrated the effectiveness of the proposed scheme.

Design and Implementation of Natural Language Interface for Impression-Based Music-Retrieval Systems

In order to realize a user-friendly impression-based music-retrieval system for input of impressions, it is essential to develop a natural language interface (NLI) that enables users to input impressions without consciously controlling their vocabulary. An NLI requires interpretation rules that represent correspondences between words and the query vectors to be generated from these words. In this paper, we propose a low-cost process where the impression words to be interpreted are estimated through a small number of impression words that have already been related to query vectors, and interpretation rules are made from the results of these estimates. We design an NLI that can interpret 164 impression words, and incorporate it into an existing impression-based music-retrieval system. Results obtained from experiments with 35 subjects prove that inexpensive interpretation rules can be made without decreasing system performance.

Multidimensional sentiment calculation method for Twitter based on emoticons

Purpose – The purpose of this paper is to propose a method of calculating the sentiment value of a tweet based on the emoticon role. Design/methodology/approach – Classification of emoticon roles as four types showing “emphasis”, “assuagement”, “conversion” and “addition”, with roles determined based on the respective relations to sentiment of sentences and emoticons. Findings – Clustering of users of four types based on emoticon sentiment. Originality/value – Formalization, using regression analysis, of the relation of sentiment between sentences and emoticons in all roles.

A Natural Language Dialogue System for Impression-based Music Retrieval

Abstract —Impression-based music retrieval is the best way tofind pieces of music that suit the preferences, senses, or mentalstates of users. A natural language interface (NLI) is more usefuland effective than a graphical user interface for impression-basedmusic retrieval since an NLI interprets users’ spontaneous inputsentences to represent musical impressions and generates queryvectors for music retrieval. Existing impression-based musicretrieval systems, however, have no dialogue capabilities formodifying the most recently used query vector. We evaluateda natural language dialogue system we developed that deals notonly with 164 impression words but also with 14 comparativeexpressions, such as “a little more” and “more and more,” and,if necessary, modifies the most recently used query vector througha dialogue. We also evaluated performance using 35 participantsto determine the effectiveness of our dialogue system. Index Terms —Music retrieval, impression-based, naturallangauge dialogue.

A query by musical impression system using N-gram based features

Impression-based music retrieval is not suitable for retrieving specific musical pieces from the huge volume of a music database, but enables users to find musical pieces that suit their preferences, feelings, or mental states. We have therefore been developing an impression-based music retrieval system that enables this. In this paper, we define the form of vectors that numerically express impressions of musical pieces, and propose a method of generating such a vector from a musical piece. The most significant attribute of this method is that it uses n-gram statistics of information on pitch, strength, and length of every tone in that musical piece as features extracted from it. We also present the results of evaluating the performance of the system

User preference modeling based on interest and impressions for news portal site systems

We have developed an application called My Portal Viewer (MPV) [1] that effectively integrates many articles collected from multiple news sites and presents these integrations through a familiar interface such as a page the user has often visited. MPV dynamically determines keywords of interest that a user might potentially be interested in based on the history of the articles the user has read and creates categories based on these interest words. MPV and many other similar integration systems, however, cause problems where users cannot find only their interest articles in each category because they are only ranked by frequency and the cooccurrence of keywords. We propose a new method of selecting further articles from each category using a users impressions of articles. The improved MPV, called MPV Plus, selects and recommends more desirable articles using the method we propose. This paper presents the design concept and process flow of MPV Plus and reports on its effectiveness as evaluated in experiments.

u-PaV: Automatic Transformation of Web Content into TV-like Video Content for Ubiquitous Environment

We propose a system that automatically transforms web content into TV-like video content for ubiquitous environments. We call this system the ubiquitous/universal passive viewer (u-PaV). The u-PaV consists mainly of audio and visual components. The audio component uses synthesized speech to read out titles and lines extracted from the target web content. Simultaneously, the visual component of the u- PaV presents the titles and lines to a user display through a ticker. Keywords and images extracted from the web content are animated on the display. A suitable background color is determined based on the overall impression of the content. The u-PaV synchronizes the ticker, animation, and speech. We introduce the u-PaV and explain how the keywords are extracted and how the impression value of the web content is determined. A test with 50 users showed that the u-PaV is easier to use and understand than browsing web content alone.

Automated content transformation with adjustment for visual presentation related to terminal types

In the ubiquitous environment, mobile terminals like personal digital assistant or cellular phone are used to access the Web. Although the presentation of Web content depends on the terminal type, the Web interface of mobile terminals is designed almost the same as that of personal computers. However, comparing mobile and desktop terminals, some difference in device characteristics, such as screen resolution or sound faithfulness, affect the presentation on the terminal. Moreover, when changing the access style between interactive and passive, the presentation should be adjusted for the terminal type because the lacking information is usually not perceived after transformation. To adjust the play style of Web content to the terminal type, we developed a method for visual presentation. First, using keyword analysis, we represent the emotional aspect of a Web page by adjusting the background color of the screen. Second, we emphasize the subject of the content by the logotype of the keyword. Third, we segment the picture of the content into pieces and present one after another using visual effects. These procedures can be applied to any type of terminal. We implemented the procedures in a ubiquitous/universal passive viewer (u-PaV) system that transforms Web content into passive viewing content that is especially useful for mobile terminals. Extracting images and text from Web content, u-PaV generates a Macromedia Flash file, which is a popular multimedia format.

Visualizing Temporal Changes in Impressions from Tweets

Twitter is effective for connecting with celebrities, on-screen talent, and strangers, as well as with friends and acquaintances, and is a social networking service utilized by several generations. On Twitter, many users will offer constructive suggestions, while a small number of users will sometimes enact mental abuse. Some users may post lively tweets, while other users may express anger in their tweets. By carefully reading a larger number of an individual users tweets, it is possible to judge the impressions from tweets received from an individual users typical posts. Therefore, this paper proposes a web application system for visualizing Twitter users based on temporal changes in the impressions from the tweets posted by users on Twitter. When system users input the account name of a Twitter user and a period for analysis, the proposed system collects the users tweets posted during the period, quantifies the impressions from each tweet, and generates pic and line charts illustrating the impression values of the tweets and their temporal changes.