AUTOMATED DATABASE SCHEMA ANNOTATION

• US 20160371275A1
• Filed: 06/18/2015
• Published: 12/22/2016
• Est. Priority Date: 06/18/2015
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Cortez Custodio Vilarinho, Eli, Bernstein, Philip A., He, Yeye, Novik, Lev
• Granted Patent: US 10,452,661 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/20 of structured data, e.g. relational data
• CPC Class Codes: G06F 16/24573 using data annotations, e.g. user-defined metadata
• CPC Class Codes: G06F 16/24578 using ranking
• CPC Class Codes: G06F 40/169 Annotation, e.g. comment data or footnotes
• CPC Class Codes: G06F 40/18 of spreadsheets form-filling G06F40/174

Abstract

Techniques and constructs that improve annotating target columns of a target database by performing automated annotation of the target columns using sources. The techniques include calculating a similarity score between a target column and columns extracted from a table that is included in a source. The similarity score is calculated based at least in part on a similarity between a value in the target column of the target database and a column value of the extracted column from the table and on a similarity between an identity of the target column of the target database and column identities of the extracted columns from the table. In some examples, the techniques calculate similarity scores for one or more extracted columns and annotate the target column based on the similarity scores.

Claims

1. A device comprising:

a processor; and

a computer-readable medium including modules, the modules when executed by the processor, configure the device to generate annotations, the modules comprising;

a column discovery module configured to retrieve a table; and

a column annotation module configured to annotate a target column of a target table by;

determining a similarity between the target column of the target table and a column of the table, the similarity based at least in part on similarities between one or more values in the target column of the target table and one or more column values extracted from the column of the table; and

annotating, based at least in part on the similarity, the target column of the target table using a column identity of the column of the table.

2. A device as claim 1 recites, wherein the similarity is further based at least in part on similarities between identities of one or more columns of the target table and column identities of one or more columns of the table.

3. A device as claim 1 recites, wherein the column annotation module is further configured to:

determine a first value of a first annotation based at least in part on the similarity;

determine a second value of a second annotation based at least in part on another similarity between the target column of the target table and another column of the table, the another similarity based at least in part on similarities between the one or more values in the target column of the target table and one or more column values extracted from the another column of the table;

rank the first annotation and the second annotation based at least in part on the first value and the second value; and

annotate the target column based at least in part on the ranking of the first annotation and the second annotation.

4. A device as claim 1 recites, wherein retrieving the table comprises the discovery module accessing a database of sources to discover a source, the source including the table.

5. A device as claim 4 recites, the modules further comprising an extraction module configured to identify the table for extraction from the source based at least in part on an identification of a header row included in the source.

6. A device as claim 4 recites, the modules further comprising an extraction module configured to identify the table for extraction from the source based at least in part on at least one of an identification of a border around a group of cells included in the source, or a group of cells included in the source that is surrounded on at least two sides by blank or empty cells.

7. A device as claim 1 recites, the modules further comprising an indexing module configured to generate an index for the table, the index comprising column values for the table mapped to individual column identities of columns included in the table.

8. A device as claim 7 recites, wherein the index further comprises the individual column identities of the columns included in the table mapped to the column values for the table.

9. A method comprising:

retrieving a table;

determining a similarity between a target column of a target table and a column of the table, the similarity based at least in part on similarities between one or more values in the target column of the target table and one or more column values extracted from the column of the table;

annotating, based at least in part on the similarity, the target column of the target table using a column identity of the column of the table; and

storing the annotated target column.

10. A method as claim 9 recites, wherein the similarity is further based at least in part on similarities between identities of one or more columns of the target table and column identities of one or more columns of the table.

11. A method as claim 9 recites, further comprising:

determining a first value of a first annotation based at least in part on the similarity;

determining a second value of a second annotation based at least in part on another similarity between the target column of the target table and another column of the table, the another similarity based at least in part on similarities between the one or more values in the target column of the target table and one or more column values extracted from the another column of the table;

ranking the first annotation and the second annotation based at least in part on the first value and the second value; and

annotating the target column based at least in part on the ranking of the first annotation and the second annotation.

12. A method as claim 9 recites, wherein the table is included in a spreadsheet, and the method further comprising discovering the spreadsheet.

13. A method as claim 12 recites, further comprising extracting the table from the spreadsheet based at least in part on an identification of a header row within the spreadsheet.

14. A method as claim 12 recites, further comprising extracting the table from the spreadsheet based at least in part on at least one of an identification of a border around a group of cells included in the spreadsheet, or a group of cells included in the spreadsheet that is surrounded on at least two sides by blank or empty cells.

15. A method as claim 9 recites, further comprising generating an index for the table, the index comprising column values for the table mapped to individual column identities of columns included in the table.

16. A method as claim 15 recites, wherein the index further comprises the individual column identities of the columns included in the table mapped to the column values for the table.

17. A computer-readable medium having computer-executable instructions to program a computer to perform operations comprising:

receiving a table;

identifying a column included in the table;

identifying a target column in a target table; and

annotating the target column included in the target table using an identity of the column included in the table.

18. A computer-readable medium as claim 17 recites, the operations further comprising ranking the identity of the column included in the table, the ranking based at least in part on at least one of:

a similarity between one or more values in the target column of the target table and one or more column values of the column included in the table; and

a similarity between identities of one or more columns of the target table that contains the target column and the identities of one or more columns included in the table.

19. A computer-readable medium as claim 17 recites, wherein the table includes a first table and the column includes a first column, the operation further comprising:

receiving a second table;

identifying a second column included in the second table; and

annotating the target column included in the target table using an identity of the second column included in the second table.

20. A computer-readable medium as claim 19 recites, the operations further comprising:

determining a first similarity score based at least in part on similarities between the first column and the target column;

determining a second similarity score based at least in part on similarities between the second column and the target column; and

ranking the identity of the first column and the identity of the second column based at least in part on the first similarity score and the second similarity score,

and wherein annotating the target column using the identity of the first column and the identity of the second column is based at least in part on the ranking.