Exploiting link structure for web page genre identification

Abstract

As the World Wide Web grows at an unprecedented pace, web page genre identification has recently attracted increasing attention because of its importance in web search. A common approach for genre identification is to utilize textual features that can be extracted directly from the web page itself, i.e., On-Page features. The extracted features are subsequently given to a machine learning algorithm that will perform classification. However, these approaches may not be e↵ective when the web page contains limited textual information (e.g., full of images). In this paper, we tackle the genre identification of web pages in such situation. We propose a framework that not only uses On-Page features, but also takes into account information in neighboring pages, i.e., the pages that are connected to the original page by backward and forward links. We first introduce a graph-based model called GenreSim which selects an appropriate set of neighboring pages. We then construct a multiple classifier combination module that utilizes information from the selected neighboring pages and On-Page features to improve genre identification performance. The experiments are conducted on well-known corpora, and the favorable results indicate that our proposed framework is e↵ective, particularly in identifying web pages with limited textual information.

Web page classification based on heterogeneous features and a combination of multiple classifiers

“…The 7-web-genres dataset (Zhu et al, 2016) has a total of 1400 HTML pages in seven categories, i.e., blog, eshop, FAQ, online newspaper front page, listing, personal home page, and search page. They are functions of web pages.…”

“…Based on the work of computing the frequencies of different HTML tags (Zhu et al, 2016), we capture the structural features of web pages by constructing vectors according to the tree-like structure of HTML tags. In web pages, HTML tags are arranged in a tree-like structure.…”

“…The classification results of dataset B are sorted as dataset B * according to the maximum predicted scores from high to low. For example, for the 7-web-genres dataset (Zhu et al, 2016), Table 1 lists 10 results with the low predicted score for dataset B * , and Table 2 lists 10 results with the high predicted score for dataset B * . Numbers 1-7 represent seven categories, and the real categories are the known labels in the dataset.…”

“…The classification performance can be improved by combining multiple classifiers, such as voting, bagging, and boosting (Baskin et al, 2017). Zhu et al (2016) used a decision matrix to construct a model with multiple SVM classifiers to classify web pages, but the combination of different types of high-performance classifiers could be better. Elsalmy et al (2017) enhanced the predictive power of web page classification models by stacking, but stacking is complicated.…”

Phi-Delta-Diagrams: Software Implementation of a Visual Tool for Assessing Classifier and Feature Performance

“…This project, which has been for two decades the largest publicly available Web directory, catalogs a huge number of web pages by means of a suitable taxonomy, each node containing web pages related to a specific topic. Categorizing web pages is an essential activity to improve user experience [22], particularly when classes are topics [23,24] and when the page at hand must be labeled as relevant or not [25]. In this scenario, both dataset has been generated from a pair of ODP categories, whose samples have been preprocessed for extracting the corresponding textual content.…”