Takafumi Nakanishi

We analysis data of a social networking service (SNS). For clarify the aspects. We used the "whisper services" within the "LINE PLAY" that is an avatar service from LINE Ltd. We found the distribution in the data that sharped close to a Poisson lognormal distribution. The potential of ease of connection is "TSUNAGARIYASUSA". Nodes have the potential of ease of connection each other. Nodes will link each other proportionate the potential, and the links distribute close to a Poisson lognormal distribution. We are expecting that this research will lead to the safety of minors.

<p>This paper presents an interconnection method for heterogeneous knowledge bases by utilizing relation information extracted from Wikipedia. Recently, the number of users who employ search engines for not only retrieving Web pages but also understanding or learning an arbitrary concept has been increasing. It is difficult to understand and learn an arbitrary concept by using most of the current search engines. To understand or learn an arbitrary concept thoroughly, it is necessary to easily determine the various relationships between the heterogeneous knowledge bases. We consider one of the method for interconnection of knowledge bases on heterogeneous fields is using Wikipedia resources.</p>

本稿では、任意の言葉を印象語として入力することにより、その印象に合致した音声表現を自動構成するための方式を提案する。本方式では、音相理論と呼ばれる理論を用いることにより、言葉の音声表現と印象語の相互の変換を実現する。さらに、本方式を意味の数学モデルに適用することにより、言葉の音声表現と任意の印象語、およびそれにより表現される任意のメディアデータを自由に連結することが可能となると考えられる。In this paper, we present an automatic composition method of sound expression of a word corresponding to the impressions of arbitrary words. We have proposed a method of automatic metadata extraction corresponding to the impression of the souds of arbitrary words by the research called &quot;Onso&quot;, and we propose an inverse operation of the automatic metadata extraction in this paper. This method enables a connection of any words and mediadata represented in impression words.

In this paper, we represent an automatic generation method for music data corresponding to impression of color information. This method can generate a music data by measuring relations between impression words extracted from color information and 8 categories of impression words given by a psychological research. This method can realize connections between heterogeneous media such as image and music. This method can realize communication places for compound media. When this method can realize, computers may be able to share and amplify Kansei with human each other.

In this paper, we present a method of automatic metadata extraction corresponding to the impression by the sound expression of arbitrary words. This method can extract impression words corresponding to the "Onso" from words or phrases. By this method, an arbitrary word given by a user can be transformed to a vector by impression words corresponding to "Onso". This method realizes the search function for media data corresponding to user's impression by using arbitrary words. In addition, this method can realize the connections between arbitrary words and heterogeneous media data.

In this paper, we present an integration method of retrival spaces. This method make it possible to integrate retrival spaces based on the relation between words. In order to realize search method utilizing a vector space model for a specific field, it is necessary to construct a retrival space for the field. The number of the vocabularies, which the retrival space constructed from a index of a document can express is restricted. The problem is solved by this method that uses other retrival spaces. It is thought that the retrival space constructed method by the index part of the documents containing this method is applicable to mediadata and document search for broad fields, such as a field of not only a scientific field but a hobby.

In order to realize associative search for a specific field by the mathematical model of meaning, it's necessary to construct a retrieval space. A construction method of retrieval space has been proposed that creates a data matrix by using a dictionary or a term dictionary. However, it's difficult to construct a retrieval space without them. This paper presents a new construction method of a retrieval space based on the relation between each word in documents. This method constructs a retrieval space by the relevance between each word calculated from word distributions in documents. By this method, a construction of a retrieval space for an associative search is performed automatically without any dictionary.

In order to realize associative search for a specific field by the mathematics model of a meaning, it is necessary to establish the retrieval space called metadata space for the specific field. A metadata space was established using a dictionary, a term dictionary, and special knowledge. However, when neither a dictionary nor a term dictionary existed in the target field, it was difficult to establish metadata space. This paper presents a new construction method of a metadata space based on the locality of words in documents. This method enables establishment of a metadata space which measures the relation between words easily without making any dictionary. The words related associative search for documents and mediadata of a specific field is realized by applying the metadata space to the mathematics model of a meaning. This paper shows the experimental results which applied the mathematics model of a meaning to the metadata space established using this method.

This paper presents an implementation method of associative heterogenous mediadata search for music data and image data. This method is realized by measuring correlation of the impression words extracted from media-lexicon transformation operator (automatic metadata extraction method) for each mediadata. The feature of this method is to bridge over heterogeneous mediadata which exist independently as different database resources.

This paper presents the system which extracts metadata automatically for voice media data. Specifically, the mounting system of Media-lexicon Transformation Operator ML is shown. First using research of Banse and Scherer, a sound is analyzed and 29 voice parameters are obtained. Next, correlation with a voice parameter and an impression word is asked for an impression word with weight. The obtained weight is metadata and the impression which man receives from a sound is shown.

This paper presents a composition method of the facial expressions by impression words. This method enables automatic composition of the facial expressions dynamically corresponding to the impression by using for Facial Action Coding System (FACS) which is the expression describing method most widely used in the research relevant to the facial expressions, measuring fundamental feeling (basic emotions) for the given arbitrary impression words, and searching for correlation of Action Unit (AU) which is the minimum unit of the basic emotions and expression operation. In this paper, we show an composition method of the facial expressions by impression words using a mathematical model of meaning which can be dynamically calculated according to the context and the situation that the semantic relation between languages or language and media data. We clarify the effectiveness of our method by showing experimental results.

This paper presents an implementation method of Kansei Operator for an automatic metadata extraction from image data. The Kansei Operator performs mapping with interpretation of human sensitivity for metadata extraction using logalismic function based on Fechner's law. The operator and a semantic associative search method enable semantic search according to the human Kansei for multimedia data. In this paper, we show an implementation method of automatic metadata extraction with Kansei Operator for image data. We clarify the effectiveness of our method by showing experimental results.

This paper presents an implementation method of heterogenous media to media search utilizing mathematical model of meaning. This method realizes connection of the heterogeneous mediadata mounted independently and can work these different-species mediadata systematically. In this paper, we show an implementation method of heterogeneous media to media search accordingly the media transformation like picture to facial expression. We clarify the effectiveness of our method by showing experimental results.