[Paper Review] Growing Wikipedia Across Languages via Recommendation

Ellery Wulczyn, Robert West+2 more|arXiv (Cornell University)|Apr 12, 2016

Wikis in Education and Collaboration15 references4 citations

TL;DR

This paper presents an end-to-end system that recommends missing articles for creation in under-resourced Wikipedia language editions by identifying articles in a source language (e.g., English) not present in a target language (e.g., French), ranking them by predicted importance using page view forecasts, and personalizing recommendations to editors based on their editing history. In a controlled experiment with 12,000 French Wikipedia editors, personalized recommendations increased editor engagement twofold and article creation rates by a factor of 3.2, with recommended articles showing comparable quality to organically created ones.

ABSTRACT

The different Wikipedia language editions vary dramatically in how comprehensive they are. As a result, most language editions contain only a small fraction of the sum of information that exists across all Wikipedias. In this paper, we present an approach to filling gaps in article coverage across different Wikipedia editions. Our main contribution is an end-to-end system for recommending articles for creation that exist in one language but are missing in another. The system involves identifying missing articles, ranking the missing articles according to their importance, and recommending important missing articles to editors based on their interests. We empirically validate our models in a controlled experiment involving 12,000 French Wikipedia editors. We find that personalizing recommendations increases editor engagement by a factor of two. Moreover, recommending articles increases their chance of being created by a factor of 3.2. Finally, articles created as a result of our recommendations are of comparable quality to organically created articles. Overall, our system leads to more engaged editors and faster growth of Wikipedia with no effect on its quality.

Motivation & Objective

• To address the significant imbalance in Wikipedia content coverage across language editions, where most editions contain only a fraction of the total knowledge.
• To reduce knowledge gaps in under-resourced Wikipedia language editions by identifying articles that exist in one language but are missing in another.
• To increase editor engagement and article creation rates by recommending missing, high-importance articles tailored to individual editors’ interests.
• To ensure that recommended articles are of high quality and sustain long-term viability by comparing their metrics to organically created content.

Proposed method

• Leveraging Wikidata’s knowledge graph to identify language-agnostic concepts and map them to language-specific Wikipedia articles, enabling detection of missing articles in target languages.
• Training a machine-learned model to predict future page views of missing articles in the target language as a proxy for their importance.
• Computing content-based similarity between editors’ past editing activity and the topics of missing articles to personalize recommendations.
• Using a bipartite matching algorithm to assign each missing article to the most interested editor, ensuring no article is recommended to multiple editors and each editor receives multiple recommendations.
• Implementing a large-scale randomized email experiment to evaluate the impact of personalized recommendations on editor engagement and article creation.
• Validating article quality by comparing deletion rates, page views, and edit quality between articles created via recommendations and those created organically.

Experimental results

Research questions

• RQ1 Does personalized recommendation of missing articles significantly increase editor engagement in Wikipedia?
• RQ2 To what extent do personalized recommendations increase the rate of article creation in under-resourced language editions?
• RQ3 Are articles created in response to recommendations of comparable quality to those created organically?
• RQ4 Can predicted page views serve as a reliable proxy for article importance in cross-lingual knowledge gap reduction?
• RQ5 How does the system perform when scaled to real-world editor populations across multiple language editions?

Key findings

• Personalized recommendations increased editor engagement by a factor of two in a controlled experiment with 12,000 French Wikipedia editors.
• The article creation rate for recommended articles increased by a factor of 3.2 compared to the baseline, demonstrating a substantial boost in content growth.
• Recommended articles had lower deletion rates and higher page views than organically created articles, indicating sustained relevance and quality.
• Articles created in response to recommendations exhibited comparable edit quality to organically created articles, confirming no degradation in content quality.
• The system effectively identified and prioritized high-importance missing articles using predicted page views as a proxy for future readership.
• The bipartite matching strategy ensured efficient recommendation distribution, maximizing relevance and minimizing redundancy.