Learning to Recommend Items to Wikidata Editors

TL;DR

This paper introduces WikidataRec, a hybrid neural recommender system designed to enhance editor engagement on Wikidata by providing personalized item suggestions, supported by large benchmark datasets and promising offline evaluation results.

Contribution

It presents a novel neural mixture of representations model for personalized Wikidata item recommendations and releases extensive benchmark datasets for future research.

Findings

Promising offline evaluation results on large datasets

Effective combination of content-based and collaborative filtering

Availability of datasets for further research

Abstract

Wikidata is an open knowledge graph built by a global community of volunteers. As it advances in scale, it faces substantial challenges around editor engagement. These challenges are in terms of both attracting new editors to keep up with the sheer amount of work and retaining existing editors. Experience from other online communities and peer-production systems, including Wikipedia, suggests that personalised recommendations could help, especially newcomers, who are sometimes unsure about how to contribute best to an ongoing effort. For this reason, we propose a recommender system WikidataRec for Wikidata items. The system uses a hybrid of content-based and collaborative filtering techniques to rank items for editors relying on both item features and item-editor previous interaction. A neural network, named a neural mixture of representations, is designed to learn fine weights for the combination of item-based representations and optimize them with editor-based representation by item-editor interaction. To facilitate further research in this space, we also create two benchmark datasets, a general-purpose one with 220,000 editors responsible for 14 million interactions with 4 million items and a second one focusing on the contributions of more than 8,000 more active editors. We perform an offline evaluation of the system on both datasets with promising results. Our code and datasets are available at https://github.com/WikidataRec-developer/Wikidata_Recommender.