[Paper Review] Improving Schema Matching with Linked Data

Ahmad Assaf, Eldad Louw+4 more|arXiv (Cornell University)|May 11, 2012

Semantic Web and Ontologies20 references3 citations

TL;DR

This paper presents a framework that enhances schema matching for tabular data integration by leveraging Linked Data—specifically Freebase's rich types—to improve accuracy. By mapping column headers to semantic types and cell values to linked entities, the approach significantly boosts matching quality in Google Refine, enabling more reliable data integration for business intelligence applications.

ABSTRACT

With today's public data sets containing billions of data items, more and more companies are looking to integrate external data with their traditional enterprise data to improve business intelligence analysis. These distributed data sources however exhibit heterogeneous data formats and terminologies and may contain noisy data. In this paper, we present a novel framework that enables business users to semi-automatically perform data integration on potentially noisy tabular data. This framework offers an extension to Google Refine with novel schema matching algorithms leveraging Freebase rich types. First experiments show that using Linked Data to map cell values with instances and column headers with types improves significantly the quality of the matching results and therefore should lead to more informed decisions.

Motivation & Objective

• Address the challenge of integrating heterogeneous, noisy tabular data from external sources with enterprise data.
• Improve schema matching quality in data integration workflows where data formats and terminologies vary widely.
• Enable business users to perform semi-automated data integration with minimal technical expertise.
• Leverage the semantic richness of Linked Data to resolve ambiguities in column headers and cell values.
• Demonstrate that integrating external knowledge from Freebase enhances matching precision and reliability.

Proposed method

• Extend Google Refine with schema matching algorithms that utilize Freebase's rich type system.
• Map column headers to semantic types using Freebase's ontology and type hierarchy.
• Link cell values to specific entities in Freebase by leveraging entity recognition and disambiguation.
• Use type-based similarity and entity-based matching to compute alignment scores between schema elements.
• Integrate the matching pipeline into a user-friendly interface for interactive data cleaning and integration.
• Apply a hybrid matching strategy combining lexical, structural, and semantic techniques enhanced by Linked Data.

Experimental results

Research questions

• RQ1 Can leveraging Linked Data improve the accuracy of schema matching in semi-automated data integration tools?
• RQ2 How does using Freebase's semantic types enhance the disambiguation of column headers and cell values?
• RQ3 To what extent does integrating external knowledge reduce errors in schema alignment for noisy, heterogeneous data?
• RQ4 Can a framework combining Google Refine with Linked Data be effective and usable for non-technical business users?
• RQ5 What is the impact of entity linking and type inference on the overall quality of schema matching results?

Key findings

• The integration of Freebase's rich types significantly improves schema matching accuracy compared to traditional methods.
• Mapping column headers to semantic types reduces ambiguity and increases confidence in schema alignment decisions.
• Linking cell values to specific entities in Freebase enhances the precision of attribute matching, especially for ambiguous or short values.
• First experiments demonstrate a measurable improvement in matching quality, suggesting better data integration outcomes.
• The framework enables business users to achieve more reliable data integration with minimal manual effort, thanks to semantic enrichment.
• The approach effectively handles noisy and heterogeneous data by grounding values in a standardized, externally validated knowledge base.