Scientists Use AuthorRank-Like Logic to Assess Article Quality on Wikipedia
Xiangju Qin and Pádraig Cunningham from UCD have just published an interesting paper which discusses the challenges in quality assessment of Wikipedia articles. Their approach to quality scoring is based on three main modes:
- Edit contribution
- Contributor authoritativeness measures
- Combination of the two
The hypothesis is that Wikipedia pages with a significant number of contributions from authoritative editors are likely to be of a high quality. Cunningham and Qin measure user authority by using centrality metrics in Wikipedia talk and co-author networks:
- Eigenvector centrality
The idea is great but personally I find little benefit in different modes of authority calculation. It seems as if basic PageRank treatment will do the job. The main issue I see is in the basic sources used to calculate authority. Given that we’re limited to Wikipedia only here, I don’t see how this could be improved, other than including temporal and revision-based metrics (which are in the proposed future work anyway).
What is interesting is that this calculation has great potential on the web in general and could help organise search results in such way that articles with high AuthorRank return higher than low AuthorRank (or no AuthorRank) articles.
The only way I see this actually working is through Google’s authorship and standardised identity networks (currently between Facebook and Google+). In the ideal world Google would use both, but if we’re realistic, it’s going to be Google+ and nothing else.
Assessing the Quality of Wikipedia Pages Using Edit Longevity and Contributor Centrality
Xiangju Qin, Pádraig Cunningham, School of Computer Science & Informatics, University College Dublin