Method and system for search structured data from a natural language search request

• US 8,762,384 B2
• Filed: 08/19/2010
• Issued: 06/24/2014
• Est. Priority Date: 08/19/2010
• Status: Active Grant

• Current Assignee: SAP SE
• Original Assignee: SAP AG (SAP SE)
• Inventors: Heidasch, Robert
• Primary Examiner(s): SYED, FARHAN M
• Application Number:
  US12/859,552
• Publication Number:
  US 20120047145A1
• Time in Patent Office: 1,405 Days
• US Class Current: 707/741
• CPC Class Codes: G06F 16/245 Query processing

Abstract

A method and system for performing a semantic search on structured data. An unstructured search query is received from a requestor. The query is evaluated within a computer to identify a best structured request based on the unstructured search query. The selected structured request is applied to a set of structured data. The result of the application of the structured request is then returned to the requestor.

Claims

1. A computer implemented method comprising:

creating an index of structured input attributes to a unique value for each attribute and each combination of attributes;

mapping a set of semantic attributes to the input attributes

accepting an unstructured search query from a requester;

analyzing the query in a computer to identify a best structured request, wherein analyzing includes hashing any input attribute in the search query, transforming any semantic attribute in the search query to an underlying input attribute, hashing any underlying input attribute, summing hash values of all input attributes and underlying input attributes, and searching the index based on the sum of the hash values;

applying the structured request to a set of structured data;

returning, to the requester, results of the application of the structured request to the set of structured data;

suggesting, to the requester, an input attribute for a semantic attribute for which no mapping exists; and

mapping the semantic attribute to an input attribute corresponding to an accepted suggestion.

2. The method of claim 1 wherein the creating comprises:

hashing each attribute; and

summing the hashes of each constituent attribute of a combination.

3. The method of claim 1, further comprising restricting the search by the number of total attributes in the search query.

4. The method of claim 1 further comprising accepting a remapping of a semantic attribute to a different input attribute.

5. A system comprising:

a processor;

an electronic display;

a graphical user interface (GUI) to accept an unstructured search query in a search window;

a persistent storage unit retaining at least one set of structured data;

an indexer to create an index of structured input attributes to a unique value for each attribute and each combination of attributes; and

a query analyzer to convert the unstructured query into a best structured request by hashing any input attribute in the search query, transforming any semantic attribute in the search query to an underlying input attribute;

hashing any underlying input attribute, summing hash values of all input attributes and underlying input attributes and searching the index based on the sum of the hash values to find a closest match;

a mapper to map semantic attributes to input attributes; and

a suggestion engine to suggest, through the GUI, possible mappings between semantic attributes and input attributes.

6. The system of claim 5 wherein the query analyzer comprises:

a semantic layer; and

an attribute indexer.

7. The system of claim 6 wherein the attribute indexer comprises:

a hash engine; and

an adder.

8. The system of claim 5 wherein the semantic layer comprises:

a lookup table; and

a comparison engine.

9. A non-transitory computer readable medium having instructions stored therein which when executed by a processor cause the processor to:

create an index of structured input attributes to a unique value for each attribute and each combination of attributes;

map a set of semantic attributes to the input attributes;

accept an unstructured search query from a requester;

analyze the query in a computer to identify a best structured request, the analysis including hashing any input attribute in the search query, transforming any semantic attribute in the search query to an underlying input attribute;

hashing any underlying input attribute, summing hash values of all input attributes and underlying input attributes and searching the index based on the sum of the hash values;

apply the structured request to a set of structured data;

return, to the requester, results of the application of the structured request to the set of structured data;

suggest, to the requester, an input attribute for a semantic attribute for which no mapping exists; and

map the semantic attribute to an input attribute corresponding to an accepted suggestion.

10. The computer readable medium of claim 9 wherein the instructions causing the computer to create comprise instructions causing the processor to:

hash each attribute; and

sum the hashes of each constituent attribute of a combination.

11. The computer readable medium of claim 9 having further instructions thereon that cause the processor to:

restrict the search by the number of total attributes in the search query.

12. The computer readable medium of claim 9 having further instructions thereon that cause the processor to:

accept a remapping of a semantic attribute to a different input attribute.