Kddi Corporation

Kokusai Denshin Denwa Co., Ltd.

Kdd Corporation

Ddi Corp.

Kazunori Matsumoto

Kazuo Hashimoto

Yasuhiro Takishima

Duc Dung Nguyen

Shigeki Muramatsu

Kenji Matsuo

Fumiaki Sugaya

Gen Hattori

Chihiro Ono

Naomi Inoue

Toshiaki Uemukai

Keiichirou Hoashi

Kazushi Ikeda

Keiko Aoki

A method of selecting a factor for estimating characteristics of viewers who view a same content is disclosed. This method includes: storing a multi-layered hierarchal list of attribute words expressing viewer attributes, each of which is a candidate of the factor; extracting the attribute words from the same content; generating a first vector indicative of frequencies with which the attribute words occur in the same content; for a successively-selected one of the attribute words, successively generating a second vector indicative of frequencies with which the attribute words occur in ones of sets of learned-text information which contain the successively-selected attribute word; calculating a similarity score between the vectors on a per-attribute-word basis; and selecting one of the attribute words as the factor, which is associated with subordinate attribute words on a lower layer, based on a dispersion level of the similarity scores calculated for the subordinate attribute words.

Context-based automatic selection of factor for use in estimating characteristics of viewers viewing same content

The present invent ion provides a condensed SVM for high-speed learning using a large amount of training data. A first stage WS selector samples a plurality of training data from a training data DB, selects an optimal training vector xamong the plurality of training data, and outputs it to the WS manager. After the first stage finishes, a second stage WS selector extracts training data one by one from the training data DB and selects training data xsatisfying optimality and outputs it to the WS manager. An SVM optimizer extracts training data closest to the training data xselected by the first and second stage WS selectors from the WS being managed by the WS manager, and condenses the two first and second training data to one training data when the distance between these is smaller than a predetermined value.

Condensed SVM

An SV reduction method for multi-class SVMs is provided with which a number of SVs contained in the multi-class SVMs can be reduced without becoming trapped in a local minimum optimization solution and the reduction of the SVs can be performed at high precision and high speed. The method includes a step of selecting, from a plurality of initially present support vectors, support vector pairs z, z(i, j=1, 2, . . . , N); a step of preparing a single-variable objective function with a single global maximum and determining a maximum value k of the objective function; and a step of applying the maximum value k to the support vector pairs zand zto determine a temporary vector Ztemp[i] of small classification errors; and the support vector pairs z, zare represented by the temporary vector Ztemp[i].

SV reduction method for multi-class SVM

The apparatus for registering face identification features can eliminate time and effort for manually retrieving and preparing face images by being provided with a face image retrieving portionfor retrieving a face image of a person via a network using the person's name as a keyword, a face feature extracting portionfor extracting features, which greatly influence identification of a person, from the face images retrieved by the face image retrieving portion, and a celebrity name, face image and feature databasefor registering the face images retrieved by the face image retrieving portionand the face features extracted by the face feature extracting portionin a state where they are associated with the person's names.

Apparatus for registering face identification features, method for registering the same, program for registering the same, and recording medium

When generating animation content as summary content for a digital broadcast program, the timing for switching of animation images for display is controlled appropriately. Subtitle character string extraction means for extracting a subtitle character string from subtitle data contained in digital broadcast signals, still image extraction means for extracting one still image corresponding to the subtitle character string, and summary content generation means for generating summary content to display the extracted subtitle character strings together with the corresponding extracted still images, are provided; the summary content generation means decides the timing for switching display of the plurality of subtitle character strings and still images comprised by the summary content, based on the subtitle character strings.

Summary content generation device and computer program

A document filtering scheme capable of reducing the number of erroneously selected non-relevant documents without reducing the number of correctly selected relevant documents is disclosed. a non-relevant profile expressing non-relevant documents that are judged as having high similarities with respect to a relevant profile expressing a user's request but not matching with the user's request is utilized in addition to the relevant profile, such that each search target document for which the similarity with respect to the relevant profile is higher than a prescribed threshold for the relevant profile and the similarity with respect to the non-relevant profile is higher than a prescribed threshold for the non-relevant profile is removed as a non-relevant document.

Scheme for filtering documents on network using relevant and non-relevant profiles

An alarm propagation model of a transmission line is expressed by an equation. A fault portion on the transmission line is estimated by using the alarm propagation model. As a result, it is possible to estimate the fault portion by minimum observation time without lack of precision.

Method of fault diagnosis based on propagation model

A similar document retrieval method capable of realizing an improved retrieval performance. In this similar document retrieval method for retrieving similar documents of a reference document from a plurality of retrieval target documents, similarities of each one of the plurality of retrieval target documents with respect to the reference document are calculated by using each one of two or more similarity calculation methods separately, and the similar documents of the reference document are retrieved by carrying out a discrimination analysis with respect to each one of a plurality of similarities calculated by using each one of the two or more similarity calculation methods separately.

Apparatus for monitoring state of traffic in telecommunication network

A document retrieval apparatus is connected to a network, and includes a cluster database storing a cluster of node information elements linked for clustering the documents to a hierarchical tree structure based on degree of similarity in all of the documents. The apparatus can post to an end address in the node information element encountered on the way to follow links of the cluster by whenever one of the documents is updated. Also, the apparatus selects a specific number of documents, clusters those, assigns the remaining non-selected documents respectively to a leaf node to be similar to the documents in the cluster, and repeats recursively the above operations toward a direction of the leaf node of cluster.

Document retrieval apparatus

A system for filtering events which have occurred within a predetermined period of time includes a first generation unit for generating association rules between the events (parameters) which simultaneously occur within the predetermined period of time, a second generation unit for generating groups of events suited for the VAR model analysis based upon the association rules generated in the first generation unit, an analysis unit for analyzing groups of events generated in the second generation unit, the analysis of the event groups being executed by using the VAR model analysis, and a verifying unit for verifying reliability of the association rules generated in the first generation unit, the verification being executed by using the VAR model.

System for filtering events occurred within a predetermined period of

A causality detection apparatus for detecting causality of events based upon a database provided with log of the events includes a sorter module for generating an event list by sorting in time sequence the events, and a frequency analysis module for calculating, based upon the event list, the number of appearance of an event set pair &lt;X,Y&gt; of a first event set X(={e.sub.x1,e.sub.x2, . . . ,e.sub.xn, . . . }, e.sub.xn indicates an event and n is a natural number) and a second event set Y(={e.sub.y1,e.sub.y2, . . . ,e.sub.yn, . . . }, e.sub.yn indicates an event), the number of appearance of the first event set X and the number of appearance of the second event set Y. The event set pair &lt;X,Y&gt; has co-occurrence so that the second event set Y occurs within a predetermined time period (window size W) after a time of occurrence of the first event set X. The apparatus further includes a causality detection module for detecting, based upon the number of appearance of the event set pair &lt;X,Y&gt;, upon the number of appearance of the first event set X and upon the number of appearance of the second event set Y, significance of causality caused by the first event set X to occur the second event set Y within the window size W.

Growing community of inventors

Saitama, Japan

Kazunori Matsumoto