Yijun Li

Sentiment classification of Internet restaurant reviews written in Cantonese

Research highlights? Naive Bayes and SVM are used for Cantonese sentiment classification. ? Accuracy is influenced by interaction between classification models and features. ? Naive Bayes classifier achieves as well as or better accuracy than SVM. ? Character-based bigrams are better features than unigrams and trigrams in capturing Cantonese sentiment. Cantonese is an important dialect in some regions of Southern China. Local online users often represent their opinions and experiences on the web with written Cantonese. Although the information in those reviews is valuable to potential consumers and sellers, the huge amount of web reviews make it difficult to give an unbiased evaluation to a product and the Cantonese reviews are unintelligible for Mandarin Chinese speakers.In this paper, standard machine learning techniques naive Bayes and SVM are incorporated into the domain of online Cantonese-written restaurant reviews to automatically classify user reviews as positive or negative. The effects of feature presentations and feature sizes on classification performance are discussed. We find that accuracy is influenced by interaction between the classification models and the feature options. The naive Bayes classifier achieves as well as or better accuracy than SVM. Character-based bigrams are proved better features than unigrams and trigrams in capturing Cantonese sentiment orientation.

Sentiment Classification for Movie Reviews in Chinese by Improved Semantic Oriented Approach

Sentiment classification aims at mining reviews of customers for a certain product by automatic classifying the reviews into positive or negative opinions. With the fast developing of World Wide Web applications, sentiment classification would have huge opportunity to help people automatic analysis of customers’ opinions from the web information. Automatic opinion mining will benefit to both consumers and sellers. Up to now, it is still a complicated task with great challenge. There are mainly two types of approaches for sentiment classification, machine learning methods and semantic orientation methods. Though some pioneer researches explored the approaches for English movie review classification, few jobs have been done on sentiment classification for Chinese reviews. The improved semantic approach for sentiment classification on movie reviews written in Chinese was proposed in this paper. Data experiment shows the capability of this approach.

The Impact of Seller Reputation on the Performance of online sales: evidence from TaoBao buy-it-now (BIN) data

The understanding of the seller reputation and sales performance relationship is an important topic in C2C market. Previous research has mainly focused on the C2C market in the U.S. The findings of that research have not been validated in countries with different cultural backgrounds such as China. With the explosive growth of electronic markets in China, the study of C2C online sales becomes increasingly important. This paper attempts to understand seller reputation effects on sales price, number of sales, and total revenue, based on But-It-Now (BIN) data collected from TaoBao.com, the largest online auction site in China. Multiple regression analysis is performed to test the significance of the effects. This study revealed cross effects of negative ratings on the performance of a seller. By separating the cross effects of magnitude and reputation of negative ratings, this study demonstrated that positive ratings have a significant positive impact on the performance of sellers, while negative ratings have negative reputation effects and positive magnitude effects. This is the first study that has focused on understanding the Buy-It-Now pricing and seller reputation relationship. Since the BIN feature is the most popular way of selling in online C2C markets in China, we believe the findings of this research will provide insight to researchers performing cross-cultural comparisosn between China and other markets.

A two-stage clustering approach for multi-region segmentation

Previous research in multi-region segmentation has found that the customer segmentation derived based on the customer attributes from one region (i.e., city or country) cannot be directly adopted by another region. As a result, for a firm that operates in multiple regions, a market segmentation method that can integrate data from different regions to obtain a set of generalized segmentation rules can greatly enhance the competitiveness of the company. In this research, we applied self-organizing map (SOM) network, an unsupervised neural networks technique as both a dimension reduction and a clustering tool to market segmentation. A two-stage clustering approach, which first groups similar regions together then finds customer segmentation for each region-group, is proposed. Empirical data from one of the largest credit card issuing banks in China was collected. The data, that includes surveys of customer satisfaction attributes and credit card transaction history, is used to validate the proposed model. The results show that the two-stage clustering approach based on SOM for multi-region segmentation is an effective and efficient method compared to other approaches.

Social media analytics for radical opinion mining in hate group web forums

Web forums are frequently used as platforms for the exchange of information and opinions as well as propaganda dissemination. But online content can be misused when the information being distributed, such as radical opinions, is unsolicited or inappropriate. This study introduces a technique that combines machine learning and semantic-oriented approaches to identify radical opinions in hate group Web forums. Four types of text features (syntactic, stylistic, content-specific, and lexicon features) are extracted as text classification predictors, and three classification techniques (SVM, Naïve Bayes, and Adaboost) are implemented. Postings from two hate group Web forums are collected and the preliminary results are encouraging. In addition, cross-validation indicates the proposed technique is stable and extendible to timeframes beyond that of the training data. The proposed technique can also be an effective tool for other sentiment classification problems.

Improving aspect extraction by augmenting a frequency-based method with web-based similarity measures

Abstract Online review mining has been used to help manufacturers and service providers improve their products and services, and to provide valuable support for consumer decision making. Product aspect extraction is fundamental to online review mining. This research is aimed to improve the performance of aspect extraction from online consumer reviews. To this end, we augment a frequency-based extraction method with PMI-IR, which utilizes web search in measuring the semantic similarity between aspect candidates and target entities. In addition, we extend RCut, an algorithm originally developed for text classification, to learn the threshold for selecting candidate aspects. Experiment results with Chinese online reviews show that our proposed method not only outperforms the state of the art frequency-based method for aspect extraction but also generalizes across different product domains and various data sizes.

Sentiment classification of online Cantonese reviews by supervised machine learning approaches

Cantonese is an important Chinese dialect spoken in some regions of Southern China. Local online users often represent their opinions and experiences with written Cantonese on the web. With two supervised machine learning approaches, this paper conducts a series of experiments to explore appropriate methods for automatic sentiment classification in the very noisy domain of online Cantonese-written reviews. Findings indicate that the support vector machine classifier based on a Mandarin Chinese word segmentation tool performs surprisingly well. The accuracy, precision and recall respectively for positive and negative reviews all reach above 85% when the training corpus contains 5,000 or more reviews.

The universal expression of periodical average publication delay at steady state

The steady state solution of differential equations of periodical publication process is deduced, and based on this, the indicator of periodical publication delay, which reflects the degree of information ageing in editorial board of a periodical, is established. The indicator is proved to be the sum of two items: the pure publication delay, which reflects the editing rapidity of a periodical, and the ratio of deposited contribution quantity to the publishing quantity in one year, which reflects the waiting period of adopted papers deposited in editorial board. As a demonstration, the delay indicators of seven periodicals are calculated. Finally, the application of this indicator is discussed.

A new nonmonotone trust-region method of conic model for solving unconstrained optimization

In this paper, we present a new nonmonotone trust-region method of conic model for solving unconstrained optimization problems. Both the local and global convergence properties are analyzed under reasonable assumptions. Numerical experiments are conducted to compare this method with some existed ones which indicate that the new method is efficient.

The influence of publication delays on three ISI indicators

SummaryBased on the transform function model of the observed citing process, the analytical expression of the age distribution of citations is deduced, and it is theoretically proved that the peak value of the citation distribution curve would fall and shift backward along with increasing the average publication delay and the peak age has a direct proportion relation with the pure delay and would be prolonged along with increasing the delay or decreasing the aging rate. The influence of the average publication delay on three ISI indicators impact factor, immediacy index and cited half-life are studied; in one subject discipline, the bigger the delay, the lower the three indicators of journals. Using the sensitivity theory, sensitivity formulae of the three indicators to publication delay parameters are deduced and it is found that responses of these indicators to changes of publication delays are different according to different time constant of the aging process; The faster the aging rate of a discipline literature is, the worse the influence of publication delays on the indicators of journals in the discipline.