Paper 8
Probabilistically Ranking Web Article Quality Based on Evolution PatternsAuthors: Jingyu Han, Kejia Chen, and Dawei Jiang |
AbstractUser-generated content (UGC) is created, updated, and main- tained by various web users, and its data quality is a major concern to all users. We observe that each Wikipedia page usually goes through a se- ries of revision stages, gradually approaching a relatively steady quality state and that articles of different quality classes exhibit specific evolution patterns. We propose to assess the quality of a number of web articles using Learning Evolution Patterns (LEP). First, each article’s revision history is mapped into a state sequence using the Hidden Markov Model (HMM). Second, evolution patterns are mined for each quality class, and each quality class is characterized by a set of quality corpora. Finally, an article’s quality is determined probabilistically by comparing the article with the quality corpora. Our experimental results demonstrate that the LEP approach can capture a web article’s quality precisely. |