Xingqi Wang

Grey System Theory based prediction for topic trend on Internet

Techniques extracting topics from dynamic Internet are relatively matured. However, people cannot accurately predict topic trend so far. Unfortunately, for prediction of topic trend, the availability of data is always very limited owing to the short life circle of topics, especially in such a highly efficient and fast-paced era. Based on Grey Verhulst Model, the paper presents an algorithm to predict topics trend. The principle of Grey Model for prediction application is analyzed and Grey Verhulst Model is established. In the meanwhile, real-world data from Youku (the largest video site in China and something like YouTube) is applied to test our presented algorithm. The average relative error of Grey Verhulst Model is less than 3%. The results show that Grey Verhulst Model has a higher prediction precision. The main contributions of this paper are as follows. First, we introduce Grey System Theory (GST) originated from system theory to the prediction of topics trend and to some extent, solve the problem with a high accuracy; second, to the best of our knowledge, it is the first attempt to employ GST in the field of topic trend prediction.

Shot Boundary Detection Method for News Video

It is very important to detect shot boundary accurately and quickly in a large number of news video data. Therefore, we proposed a new method with dual-detection model. The method is divided into two stages, i.e. pre-detection and re-detection. In the pre-detection stage, the uneven blocked differences based on the feature of human vision are presented and used in adaptive binary search to detect shot boundaries. In the re-detection round, Speeded Up Robust Features (SURF) method is applied to exclude false detections. The experimental results show that this method can detect abrupt boundaries of news video quickly and accurately.

A mixture language model for the classification of Chinese online reviews

In this essay, we propose an unsupervised topic and sentiment mixture model (LSS model) mainly on the basis of LDA model and combining with the sentiment factor, as an approach to sentiment classification of online reviews. There is an extra constraint that all words in a sentence are generated from one topic and one sentiment. Which means the hypothesis of the model is a word in a sentence is of one meaning only. LSS model is totally unsupervised and it needs no labelled corpora or any other sentiment seed words. Experiments show that LSS model performs better than JST model and ASUM model. The F-1 value of sentiment classification is 8% higher than ASUM model and 12.5% than JST model.

A New Method For Text Location in News Video Based on Ant Colony Algorithm

Abstract—Text in video is a very compact and accurate clue for video indexing and summarization. The paper presents a new method for text location in news video with ant colony algorithm. Three features of characters are extracted as a basis for the formation of heuristic function. In order to balance the weight of the three features, three functions are introduced to transform them. The ants will be randomly put on sub-blocks of video frames for searching text areas. Therefore, ants would leave pheromone in each sub-block. After the ant colony algorithm is finished, it produces a pheromone matrix. By binarizing the pheromone matrix, the text blocks can be located. The result has proved that the binarization method proposed in this paper is more accurate than otsu’s method. At last, to reduce the false detection rate, the different directions of edge intensity ratio of text areas are computed, as the real text areas’ edge intensity ratio is much smaller than the false one.