Author:
Wei-An Chen, Jihg-Hong Lin, and Shyh-Kang Jeng
Abstract:
In this research project, we propose a model, the Harmony Graph, to decompose music into a social-network-like structure according to its harmonies. The whole Harmony Graph network represents the harmony progressions in music. The Harmony Graph is utilized to visualize, distinguish, and generate music for four prepared corpora using social network techniques. We experimented on different characteristics in social network analysis, and we found significant differences among the Harmony Graphs of the four corpora. A new measure called Agglomeration is created to characterize the agglomerating phenomenon that cannot be described sufficiently by existing measures. A corpus-based music composition method is also proposed in this research. By performing random-walk in a Harmony Graph, we generated new music that differs from yet reflects the style of music pieces in the corpus. With the link prediction technique, we also generated music more pleasant aurally than simply using random walks.
Keywords:
Social Network Analysis, Corpus Visualization, Corpus-Based Generation
Author:
Min-Hsiang Li, Shih-Hung Wu, Yi-Ching Zeng, Ping-che Yang, and Tsun Ku
Abstract:
The character sets used in China and Taiwan are both Chinese, but they are divided into simplified and traditional Chinese characters. There are large amount of information exchange between China and Taiwan through books and Internet. To provide readers a convenient reading environment, the character conversion between simplified and traditional Chinese is necessary. The conversion between simplified and traditional Chinese characters has two problems: one-to-many ambiguity and term usage problems. Since there are many traditional Chinese characters that have only one corresponding simplified character, when converting simplified Chinese into traditional Chinese, the system will face the one-to-many ambiguity. Also, there are many terms that have different usages between the two Chinese societies. This paper focus on designing an extensible conversion system, that can take the advantage of community knowledge by accumulating lookup tables through Wikipedia to tackle the term usage problem and can integrate language model to disambiguate the one-to-many ambiguity. The system can reduce the cost of proofreading of character conversion for books, e-books, or online publications. The extensible architecture makes it easy to improve the system with new training data.
Keywords:
Chinese Character Conversion, Language Model, Wikipedia, Lookup Table
Author:
Chuan-Jie Lin and Pin-Hsien Chao
Abstract:
This paper focuses on tourism-related opinion mining, including tourism-related opinion detection and tourist-attraction target identification. The experimental data are blog articles labeled as being in the domestic tourism category in a blogspace. Annotators were asked to annotate the opinion polarity and the opinion target for every sentence. Different strategies and features have been proposed to identify opinion targets, including tourist attraction keywords, coreferential expressions, tourism-related opinion words, and a 2-level classifier. We used machine learning methods to train classifiers for tourism-related opinion mining. A retraining mechanism is proposed to obtain the system decisions of preceding sentences. The precision and recall scores of tourism-related opinion detection were 55.98% and 59.30%, respectively, and the scores of tourist attraction target identification among known tourism-related opinionated sentences were 90.06% and 89.91%, respectively. The overall precision and recall scores were 51.30% and 54.21%, respectively.
Keywords:
Tourism-Related Opinion Mining, Tourist Attraction Target Identification, Opinion Analysis
Author:
Heng Lu, Zhen-Hua Ling, Li-Rong Dai, and Ren-Hua Wang
Abstract:
This paper presents a decision tree pruning method for the model clustering of HMM-based parametric speech synthesis by cross-validation (CV) under the minimum generation error (MGE) criterion. Decision-tree-based model clustering is an important component in the training process of an HMM based speech synthesis system. Conventionally, the maximum likelihood (ML) criterion is employed to choose the optimal contextual question from the question set for each tree node split and the minimum description length (MDL) principle is introduced as the stopping criterion to prevent building overly large tree models. Nevertheless, the MDL criterion is derived based on an asymptotic assumption and is problematic in theory when the size of the training data set is not large enough. Besides, inconsistency exists between the MDL criterion and the aim of speech synthesis. Therefore, a minimum cross generation error (MCGE) based decision tree pruning method for HMM-based speech synthesis is proposed in this paper. The initial decision tree is trained by MDL clustering with a factor estimated using the MCGE criterion by cross-validation. Then the decision tree size is tuned by backing-off or splitting each leaf node iteratively to minimize a cross generation error, which is defined to present the sum of generation errors calculated for all training sentences using cross-validation. Objective and subjective evaluation results show that the proposed method outperforms the conventional MDL-based model clustering method significantly.
Keywords:
Speech Synthesis, Hidden Markov Model, Decision Tree Pruning, Cross-validation, Minimum Generation Error