Yu Zhengtao

Question Classification Based on Incremental Modified Bayes

How to use the incremental training corpus to improve the question classification accuracy rate in the process of question classification based on statistic learning. A question classification method based on the incremental modified Bayes was presented in this paper. The method used the modified Bayes and combined the incremental learning to correct the parameter by the incremental training set stage by stage, and established the question classification model based on the incremental modified Bayes. A question classification experiment was done in the domain of Yunnan tourism, the experimental results showed that the presented method evidently excelled than the modified Bayes method in the accuracy rate and the training time, the average accuracy rate was improved 3.3 percentage points than the accuracy rate of the modified Bayes method; the average training time was improved 39.1 percentage points than the training time efficiency of the modified Bayes method.How to use the incremental training corpus to improve the question classification accuracy rate in the process of question classification based on statistic learning. A question classification method based on the incremental modified Bayes was presented in this paper. The method used the modified Bayes and combined the incremental learning to correct the parameter by the incremental training set stage by stage, and established the question classification model based on the incremental modified Bayes. A question classification experiment was done in the domain of Yunnan tourism, the experimental results showed that the presented method evidently excelled than the modified Bayes method in the accuracy rate and the training time, the average accuracy rate was improved 3.3 percentage points than the accuracy rate of the modified Bayes method; the average training time was improved 39.1 percentage points than the training time efficiency of the modified Bayes method.

Implementing a quantitative-based methodology for project risk assessment DSS

Project risk assessment is a critical activity adopted in project risk management process to prevent risks and to enhance the success rate of projects. But so far it is a big challenge for project managers and experts to combine their expertise with intelligent technology to evaluate project risks due to insufficient risk related data. Based on this, a novel attempt to integrate analytic hierarchy process (AHP) and support vector regression (SVR) is proposed to build the assessment decision model. A prototype system called project risk assessment decision support system (PRADSS) which consists of risk index system, AHP evaluation model, intelligent evaluation system, knowledge base and man-machine interaction is brought forward to reduce negative risk factors and assist in decision making. Software project risk assessment is conducted to show the efficiency and feasibility of the prototype system.

Word Sense Disambiguation Based on Bayes Model and Information Gain

Word sense disambiguation has always been a key problem in natural language processing. In the paper, we use the method of information gain to calculate the weight of different positions context, which affect to ambiguous words. And take this as the foundation. We select the ahead and back six position¿s context of ambiguous words to construct the feature vectors. The feature vectors are endued with different value of weight in Bayesian model. Thus, the Bayesian model is improved. We use the sense of the HowNet to describe the meaning of ambiguous words. The average accuracy rate of the experiments of 10 Chinese ambiguous words was 95.72% in close test and the average accuracy rate was 85.71% in open test. The results showed that the method was proposed in this paper were very effective.

A new algorithm based on word co-occurrence and its application in domain concept extraction

This paper puts forward an algorithm named CFE (Co-occurrence Frequency Emphasized) for new concept selection which based on word co-occurrence, it has been applied to domain concept extraction. This algorithm grasps the central idea of word co-occurrence rigidly. In the process of concept extracting, the co-occurrence frequency is taken as the basis of new concept selection, and the impact of absolute frequency to domain concept selection is taken into account. It is able to guarantee each candidate word will fairly get the opportunity of being selected as new concept. According to the data of the tests, a good result has been achieved.

The effects of domain knowledge relations on domain text classification

The text classification usually uses the statistical method to select characteristic. When it is carried out in different domains, the special interior knowledge relationships between domains will not be considered. In this paper, a new text classification model is proposed, which is based on the domain knowledge relations. This model adopts the support vector machine study algorithm, combine statistic samples and domain terminology to make up classification feature space, and calculate the similarity between domain conceptions, so that classification characteristic is entrusted with certain weight, realizing domain text classification. The new model has been made use of to carry out a text classification experiment about YunNan travel domain and non-travel domain. The result shows that domain knowledge has great effects on domain text classification and the accuracy of classification has been improved by 4 percentage compared with the improved TFIDF method.

Vietnamese Part of Speech Tagging Based on Multi-category Words Disambiguation Model

POS tagging is a fundamental work in Natural Language Processing, which determines the subsequent processing quality, and the ambiguity of multi-category words directly affects the accuracy of Vietnamese POS tagging. At present, the POS tagging of English and Chinese has achieved better results, but the accuracy of Vietnamese POS tagging is still to be improved. For address this problem, this paper proposes a novel method of Vietnamese POS tagging based on multi-category words disambiguation model and Part of Speech dictionary, the multi-category words dictionary and the non-multi-category words dictionary are generated from the Vietnamese dictionary, which are used to build POS tagging corpus. 396,946 multi-category words have been extracted from the corpus, by using statistical method, the maximum entropy disambiguation model of Vietnamese part of speech is constructed, based on it, the multi-category words and the non-multi-category words are tagged. Experimental results show that the method proposed in the paper is higher than the existing model, which is proved that the method is feasible and effective.