Annotating schema elements based on associating data instances with knowledge base entities

A method for determining schema element types, comprising:

• pooling potential annotations in a hierarchical structure, for an element of an unlabeled schema, from a plurality of heterogeneous sources, and merging instances of similar schema elements, wherein the hierarchical structures from the plurality of heterogeneous sources are combined into a single hierarchy to correlate all of the potential annotations for the element and wherein each of the plurality of heterogeneous sources provides a respective preliminary score for potential annotations and wherein the preliminary scores are formed by at least two different score assignation schemes;
• scoring with a processor the pool of potential annotations according to structural properties of the potential annotations, such that potential annotations that are further away in the hierarchical structure from a potential annotation that is being scored are assigned smaller scores than schema element types that are closer in the hierarchical structure to the potential annotation being scored, and using a scoring function, which sorts potential annotations from each of the plurality of heterogeneous sources by score and prunes potential annotations after a drop in score between two of the sorted plurality of heterogeneous sources that is greater than a threshold, to combine multiple preliminary scores from the remaining heterogeneous sources to produce a relevancy score, wherein the scoring function gives larger weights to potential annotations that are returned by multiple sources from the plurality of heterogeneous sources;
• and
• annotating the element of the unlabeled schema using the most relevant potential annotations based on the relevance score.

A system for determining schema element types, comprising:

• a query module configured to pool potential annotations, for an element of an unlabeled schema, from a plurality of heterogeneous sources, wherein each of the plurality of heterogeneous sources provides a respective preliminary score for potential annotations and wherein the preliminary scores are formed by at least two different score assignation schemes;
• a data module comprising a processor configured to merge instances of similar schema elements and to score the pool of potential annotations according to relevancy using instance information from the plurality of heterogeneous sources and a scoring function, which sorts potential annotations from each of the plurality of heterogeneous sources by score and prunes potential annotations after a drop in score between two of the sorted plurality of heterogeneous sources that is greater than a threshold, to combine multiple preliminary scores from the remaining heterogeneous sources into a relevancy score, wherein the scoring function gives larger weights to potential annotations that are returned by multiple sources from the plurality of heterogeneous sources;
• and
• a data linker configured to annotate the element of the unlabeled schema using the most relevant potential annotations based on the relevancy score.

A system for determining schema element types, comprising:

• a query module configured to pool potential annotations in a hierarchical structure, for an element of an unlabeled schema, from a plurality of heterogeneous sources, wherein the hierarchical structures from the plurality of sources are combined into a single hierarchy to correlate all of the potential annotations for the element and wherein each of the plurality of heterogeneous sources provides a respective preliminary score for potential annotations and wherein the preliminary scores are formed by at least two different score assignation schemes;
• a data module comprising a processor configured merge instances of similar schema elements and to score the pool of potential annotations according to structural properties of the potential annotations, such that potential annotations that are further away in the hierarchical structure from a potential annotation that is being scored are assigned smaller scores than schema element types that are closer in the hierarchical structure to the potential annotation being scored, and using a scoring function, which sorts potential annotations from each source of the plurality of heterogeneous sources by score and prunes potential annotations after a drop in score between two of the sorted plurality of heterogeneous sources that is greater than a threshold, to combine multiple preliminary scores from the remaining heterogeneous sources to produce a relevancy score, wherein the scoring function gives larger weights to potential annotations that are returned by multiple sources from the plurality of heterogeneous sources;
• and
• a data linker configured to annotate the element of the unlabeled schema using the most relevant potential annotations based on the relevancy score.

A non-transitory computer readable storage medium comprising a computer readable program, wherein the computer readable program when executed on a computer causes the computer to:

• pool potential annotations, for an element of an unlabeled schema, from a plurality of heterogeneous sources, and merge instances of similar schema elements, wherein each of the plurality of heterogeneous sources provides a respective preliminary score for potential annotations and wherein the preliminary scores are formed by at least two different score assignation schemes;
• score the pool of potential annotations according to relevancy with a processor using information using instance information from the plurality of heterogeneous sources and a scoring function, which sorts potential annotations from each of the plurality of heterogeneous sources by score and prunes potential annotations after a drop in score between two of the sorted plurality of heterogeneous sources that is greater than a threshold, to combine multiple preliminary scores from the remaining heterogeneous sources into a relevancy score, wherein the scoring function gives larger weights to potential annotations that are returned by multiple sources from the plurality of heterogeneous sources;
• and
• annotate the element of the unlabeled schema using the most relevant potential annotations based on the relevancy score.