Data Mining - Concepts and Techniques.

We will review some of the major results in random graphs and some of the more challenging open problems. We will cover algorithmic and structural questions. We will touch on newer models, including those related to the WWW.

Random graphs

“Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks”の学習報告

Social Network Analysis

Emotion and Adaptation

We present an approach to construct a translation dictionary used in multilingual natural language processing such as machine translation. We use a learning and extracting method of bilingual word pairs from parallel corpora using the maximum entropy modeling. We are doing the experiment in Japanese and Chinese parallel corpora.

http://ieeexplore.ieee.org/iel5/8325/26388/01176054.pdf

To define the feature function in extracting Japanese-Chinese bilingual word pairs using maximum entropy modeling

Natural language has attracted importance to some researchers working on Emotion recognizing. Our study focuses on the semi-automatic acquisition technique to obtain the emotion information with the constructed emotion thesaurus we have made. In this paper, we made two models, one is the emotion thesaurus, the other is the emotion analyzer. Firstly we construct Chinese language emotion thesaurus, the word in which thesaurus has its own classifying of the corresponding emotion information. Secondly, using the thesaurus we set up a model to recognize the object sentence including functions of lexical analysis syntax analysis, emotion sensing and emotion computing. The following investigations are presented to show that the concept of "image value" can be used by the constructed system to obtain the emotion information from the sentence based on the constructed emotion thesaurus.

Construction and evaluation of Chinese emotion classification model

Image description is not sufficient in traditional facial expression recognition (FER) methods, therefore this paper proposes a FER method based on multi-scale vector triangle. It combines vector triangle pattern with image pyramid to extract facial expression features. Firstly, construct a facial image pyramid to produce images in different scales. Secondly, divide each image into blocks, and extract vector triangle features of each sub-image. Then, use histogram to statistical characteristics, and calculate Euclidean distance between the histograms. Finally, fusion weighted eigenvalues and come to the recognition results. Multi-scale vector triangle pattern can not only avoid the loss of information in image asymmetric regions, but also reflect image features in different scales. It can describe images more adequately. In order to verify the effectiveness of the algorithm, this paper uses the Japanese Female Facial Expression (JAFFE) database to do the experiments and compare the results with Complete Local Binary Patterns (CLBP), Gabor wavelet, Active Appearance Models (AAM) and so on. Experimental results indicate that this method has higher recognition rate and better real-time effect.

Facial expression recognition based on multi-scale vector triangle

In this paper, we describe a novel acoustic model adaptation technique which generates “speaker-independent” HMM for the target environment. Recently, personal digital assistants like cellular phones are shifting to IP terminals. The encoding-decoding process utilized for transmitting over IP networks deteriorates the quality of speech data. This deterioration causes degradation in speech recognition performance. Acoustic model adaptations can improve recognition performance. However, the conventional adaptation methods usually require a large amount of adaptation data. The proposed method uses HMM-based speech synthesis to generate adaptation data from the acoustic model of HMM-based speech recognizer, and consequently does not require any speech data for adaptation. Experimental results on G.723.1 coded speech recognition show that the proposed method improves speech recognition performance. A relative word error rate reduction of approximately 12% was observed.

Acoustic model adaptation for coded speech using synthetic speech.

Fuzzy density is an important part of fuzzy integral, which is used to describe the reliability of classifiers in the process of fusion. Most of the fuzzy density assignment methods are based on the training priori knowledge of the classifier and ignore the difference of the testing samples themselves. To better describe the real-time reliability of the classifier in the fusion process, the dispersion of the classifier is calculated according to the decision information which outputted by the classifier. Then the divisibility of the classifier is obtained through the information entropy of the dispersion. Finally, the divisibility and the priori knowledge are combined to get the fuzzy density which can be dynamically adjusted. Experiments on JAFFE and CK databases show that, compared with traditional fuzzy integral methods, the proposed method can effectively improve the decision performance of fuzzy integral and reduce the interference of unreliable output information to decision. And it is an effective multi-classifier fusion method.

Fuji Ren

Papers

To define the feature function in extracting Japanese-Chinese bilingual word pairs using maximum entropy modeling

Construction and evaluation of Chinese emotion classification model

Facial expression recognition based on multi-scale vector triangle

Acoustic model adaptation for coded speech using synthetic speech.

A Fusion Decision Method Based on the Dynamic Fuzzy Density Assignment