[Paper Review] Multilingual Schema Matching for Wikipedia Infoboxes

Thanh Nguyen, Viviane P. Moreira+3 more|arXiv (Cornell University)|Oct 30, 2011

Topic Modeling27 references9 citations

TL;DR

This paper proposes WikiMatch, a training-free, automated method for multilingual schema matching of Wikipedia infoboxes that combines multiple similarity signals—semantic, lexical, and structural—without relying on syntactic similarity or external resources. It achieves high precision and recall across diverse language pairs, including morphologically distinct languages like Vietnamese and English, and outperforms state-of-the-art techniques in cross-lingual attribute alignment.

ABSTRACT

Recent research has taken advantage of Wikipedia's multilingualism as a resource for cross-language information retrieval and machine translation, as well as proposed techniques for enriching its cross-language structure. The availability of documents in multiple languages also opens up new opportunities for querying structured Wikipedia content, and in particular, to enable answers that straddle different languages. As a step towards supporting such queries, in this paper, we propose a method for identifying mappings between attributes from infoboxes that come from pages in different languages. Our approach finds mappings in a completely automated fashion. Because it does not require training data, it is scalable: not only can it be used to find mappings between many language pairs, but it is also effective for languages that are under-represented and lack sufficient training samples. Another important benefit of our approach is that it does not depend on syntactic similarity between attribute names, and thus, it can be applied to language pairs that have distinct morphologies. We have performed an extensive experimental evaluation using a corpus consisting of pages in Portuguese, Vietnamese, and English. The results show that not only does our approach obtain high precision and recall, but it also outperforms state-of-the-art techniques. We also present a case study which demonstrates that the multilingual mappings we derive lead to substantial improvements in answer quality and coverage for structured queries over Wikipedia content.

Motivation & Objective

• To address the challenge of supporting multilingual structured queries over Wikipedia by enabling schema matching across infoboxes in different languages.
• To develop a scalable, automated approach that does not require training data, making it applicable to low-resource and morphologically distinct language pairs.
• To overcome limitations of existing methods that depend on syntactic similarity or large annotated training sets, especially for under-represented languages.
• To improve answer quality and coverage in multilingual queries by discovering reliable attribute correspondences across languages.
• To demonstrate the practical benefits of multilingual schema alignment through a case study on real-world query workloads.

Proposed method

• WikiMatch combines multiple similarity signals—semantic, lexical, and structural—using a weighted fusion strategy to identify multilingual attribute correspondences.
• It prioritizes high-confidence matches through iterative refinement, minimizing error propagation in the matching process.
• The method does not require training data, external dictionaries, or machine translation systems, relying instead on intrinsic Wikipedia structure and content.
• It leverages cross-language links and value co-occurrence patterns to infer semantic similarity between attributes across languages.
• The approach uses a similarity thresholding mechanism that dynamically adjusts based on confidence scores to balance precision and recall.
• It applies a systematic combination of features such as n-gram similarity, semantic embeddings, and attribute-value alignment to improve robustness across heterogeneous infobox schemas.

Experimental results

Research questions

• RQ1 Can a training-free approach achieve high-precision multilingual schema matching for Wikipedia infoboxes across languages with distinct morphologies?
• RQ2 How does WikiMatch perform compared to state-of-the-art methods in terms of precision, recall, and scalability on low-resource language pairs?
• RQ3 To what extent do the derived multilingual mappings improve answer coverage and quality for structured queries over multilingual Wikipedia content?
• RQ4 Does the method remain effective when attribute names and structures vary significantly across languages due to schema drift, polysemy, and synonymy?
• RQ5 Can the approach be applied effectively to under-represented languages like Vietnamese and Portuguese without requiring large-scale annotated training data?

Key findings

• WikiMatch achieves high precision and recall in multilingual schema matching, outperforming state-of-the-art techniques on the Portuguese-English and Vietnamese-English language pairs.
• The method is effective for languages with distinct morphologies, such as Vietnamese and English, where syntactic similarity-based methods fail.
• It does not require training data or external resources like machine translation systems, making it scalable and applicable to low-resource languages.
• The case study shows that using WikiMatch-derived mappings increases answer coverage and quality for multilingual queries by enabling translation of queries into English and leveraging English Wikipedia content.
• The approach maintains strong performance even on entity types with high cross-language heterogeneity and limited data instances.
• Experiments show that WikiMatch is robust to schema drift, polysemy, and synonymy, outperforming methods like Ziggurat and COMA++ in both precision and recall.