19 research outputs found
Combining classifiers for flexible genre categorization of web pages
No full textComputer Science Department, College of Computer and Information Sciences, King Saud UniversityCollege of Computer and Information Sciences
Computer Science Department
P. O. Box 51178, Riyadh 11543, KSU.With the increase of the number of web pages, it is very difficult to find wanted information easily and quickly out of thousands of web pages retrieved by a search engine. To solve this problem, many researches propose to classify documents according to their genre, which is another criteria to classify documents different from the topic. Most of these works assign a document to only one genre. In this paper we propose a new flexible approach for document genre categorization. Flexibility means that our approach assigns a document to all predefined genres with different weights. The proposed approach is based on the combination of two homogenous classifiers: contextual and structural classifiers. The contextual classifier uses the URL, while the structural classifier uses the document structure. Both contextual and structural classifiers are centroid-based classifiers. Experimentations provide a micro-averaged break-even point (BEP) more than 85%, which is better than those obtained by other categorization approaches
Multi-Label Genre Classification of Web Pages Using an Adaptive Centroid-Based Classifier
No full text
Una combinación basada en operadores OWA para la Clasificación de Género Multi-etiqueta de páginas web
No full textThis paper presents a new method for genre identification that combines homogeneous classifiers using OWA (Ordered Weighted Averaging) operators. Our method uses character n-grams extracted from different information sources such as URL, title, headings and anchors. To deal with the complexity of web pages, we applied MLKNN as a multi-label classifier, in which a web page can be affected by more than one genre. Experiments conducted using a known multi-label corpus show that our method achieves good results.En este trabajo se presenta un nuevo método para la identificación de género que combina clasificadores homogéneos utilizando OWA (promedio ponderado) Pedimos operadores. Nuestro método utiliza caracteres n-gramas extraídos de diferentes fuentes de información, tales como URL, título, encabezados y anclajes. Para hacer frente a la complejidad de las páginas web, se aplicó MLKNN como un clasificador multi-etiqueta, en el que una página web puede verse afectada por más de un género. Los experimentos llevados a cabo usando un conocido corpus multi-etiqueta muestran que nuestro método logra buenos resultados
MLICC: A Multi-Label and Incremental Centroid-Based Classification of Web Pages by Genre
No full textAn Improved Centroid-Based Approach for Multi-label Classification of Web Pages by Genre
No full text
