Methods and system for providing real-time business intelligence using natural language queries

• US 9,665,662 B1
• Filed: 06/13/2014
• Issued: 05/30/2017
• Est. Priority Date: 06/13/2013
• Status: Active Grant

• Current Assignee: Progress Software Corporation
• Original Assignee: DataRPM Corporation (Progress Software Corporation)
• Inventors: Gautam, Shyamantak, Phukan, Ruban, Katkar, Vishal, Sanghavi, Sundeep
• Primary Examiner(s): Shmatov, Alexey
• Application Number:
  US14/304,487
• Time in Patent Office: 1,082 Days
• Field of Search: 707708
• CPC Class Codes: G06F 16/24522 Translation of natural language queries to structured queries
• CPC Class Codes: G06F 16/26 Visual data mining; Browsing structured data
• CPC Class Codes: G06F 16/3328 using graphical result space presentation or visualisation
• CPC Class Codes: G06F 16/903 Querying for retrieval from the web G06F16/953
• CPC Class Codes: G06F 40/40 Processing or translation of natural language natural language analysis G06F40/20; semantic analysis G06F40/30

Abstract

The methods and systems for providing real-time business intelligence using natural language queries facilitate a user to search within a data warehouse using a natural language question. Such business intelligence platform may receive a natural language based question, extract one or more key words from the natural language based question, determine a first dependency graph of the one or more key words based on a relationship among the one or more key words, determine a second dependency graph of the one or more key words based on previously stored search indices, merge the first and the second dependency graphs to generate an integrated dependency graph, and generate a formatted search string based on the integrated dependency graph.

Claims

1. A system, comprising:

a processor; and

a memory operatively coupled to the processor, the memory storing processor-readable instructions executable by the processor to;

receive a natural language based question relating to a user desired data analytics variable via a user interface;

extract a plurality of key words including a first key word relating to the user desired data analytics variable and a second key word from the natural language based question;

generate each of a language dependency graph and a data dependency graph in parallel, the data dependency graph being generated based on, and including, relationships between the plurality of key words and a plurality of search terms in a plurality of data sources, the language dependency graph being generated based on, and including, relationships between the plurality of key words and natural language sentence structure determined using a natural language typed-dependency graph that maps each key word node of a plurality of key word nodes included in the natural language typed-dependency graph to a portion of the remaining key word nodes of the plurality of key word nodes based on sentence structure rules;

merge the language dependency graph and the data dependency graph to form an inter-dependency graph including a mapping between the plurality of key words and (1) a set of previously stored query terms and (2) a set of operational commands, based on the relationships included in the data dependency graph and the relationships included in the language dependency graph, the set of previously stored query terms being stored in a data table from a data source from the plurality of data sources;

store the inter-dependency graph in an index file linked to the plurality of data sources;

determine a previously stored query term for the second key word based on a mapping of the second key word and the previously stored query term that is included in the inter-dependency graph;

determine a computational operation for the first key word based on a mapping of the first key word and the computational operation that is included in the inter-dependency graph; and

translate the natural language based question into a formatted query string based on the index file, the computational operation, and the previously stored query term,

the user desired data analytics variable being derived by applying the computational operation on a query result relating to the previously stored query term.

2. The system of claim 1, wherein the natural language based question is submitted by a user by at least one of:

typing a question, or speaking the question to a voice input element.

3. The system of claim 1, wherein the plurality of key words are extracted by parsing the natural language based question into phrases and eliminating ancillary phrases based on heuristics.

4. The system of claim 1, wherein the inter-dependency graph is defined based on a context and a sentence structure of the natural language based question.

5. The system of claim 1, wherein the processor-readable instructions are further executable by the processor to:

compare the inter-dependency graph and the mapping to eliminate inaccurate query term mappings.

6. The system of claim 1, wherein the processor-readable instructions are further executable by the processor to:

instantiate a scoring mechanism to prioritize different query term mappings when there are multiple query terms mapped to one key word.

7. A processor-implemented method, comprising:

receiving a natural language based question;

extracting a plurality of key words from the natural language based question;

mapping the plurality of key words to a set of previously stored query terms and a computational operation connecting the set of previously stored query terms;

determining a first dependency graph of the plurality of key words based on, and including, a grammatical relationship among the plurality of key words determined using a natural language typed-dependency graph that maps each key word node of a plurality of key word nodes included in the natural language typed-dependency graph to a portion of the remaining key word nodes of the plurality of key word nodes based on sentence structure rules;

determining a second dependency graph of the plurality of key words based on, and including, the mapping of the plurality of key words to the set of previously stored query terms and the computational operation;

merging the first and the second dependency graphs to generate an integrated dependency graph including a mapping between the key words and (1) the set of previously stored query terms and (2) a set of operational commands, based on the grammatical relationship included in the first dependency graph and the mapping in the second dependency graph, the integrated dependency graph being saved in an index file linked to a plurality of data sources; and

generating a formatted search string from the natural language based question based on (1) at least one operational command, from the set of operational commands, that is mapped to the plurality of key words, (2) the set of previously stored query terms, and (3) the index file.

8. The method of claim 7, wherein the natural language based question is submitted any of a user manual input interface, or a voice input interface.

9. The method of claim 7, wherein the plurality of key words is extracted by parsing the natural language based question into phrases and eliminating ancillary phrases based on heuristics.

10. The method of claim 7, wherein the first dependency graph has the plurality of key words as nodes, and connects the plurality of key words based on a context and a sentence structure of the natural language based question.

11. The method of claim 7, wherein the second dependency graph has the plurality of key words and the set of previously stored query terms as nodes, and connects each key word from the plurality of key words with a previously stored query term from the set of previously stored query terms.

12. The method of claim 7, wherein the merging the first and the second dependency graphs includes:

comparing the first dependency graph and the second dependency graph to eliminate inaccurate query term mappings.

13. The method of claim 12, wherein the merging the first and the second dependency graphs further includes:

instantiating a scoring mechanism to prioritize different query term mappings when multiple query terms from the set of previously stored query terms are mapped to one key word.

14. The method of claim 13, wherein the scoring mechanism ranks different query term mappings based at least in part on user feedback.

15. A non-transitory processor-readable medium storing code representing processor-executable instructions, the code comprising code to cause the processor to:

receive a natural language based question relating to a user desired data analytics variable via a user interface;

extract a plurality of key words from the natural language based question;

generate an inter-dependency graph of the plurality of key words and previously stored query terms from a plurality of data sources based on each of (1) a first dependency graph generated based on relationships between the plurality of key words and natural language sentence structure determined using a natural language typed-dependency graph that maps each key word node of a plurality of key word nodes included in the natural language typed-dependency graph to a portion of the remaining key word nodes of the plurality of key word nodes based on sentence structure rules and (2) a second dependency graph including relationships between the plurality of key words and the previously stored query terms, the inter-dependency graph including a computational operation to derive the user desired data analytics variable and being saved in an index file linked to the plurality of data sources, the inter-dependency graph including a mapping between the plurality of key words and (1) a set of previously stored query terms and (2) a set of operational commands that includes the computational operation, based on the relationships included in the first dependency graph and the relationships included in the second dependency graph;

translate the natural language based question into a formatted query string of previously stored query terms based on (1) the operational command, (2) the set of previously stored query terms, and (3) the index file;

send a query request to a data source from the plurality of data sources based on the formatted query string;

obtain a query result from the data source in response to the query request; and

derive an output value for the user desired data analytics variable based on the query result and the computational operation.

16. The medium of claim 15, wherein the user interface includes any of a user manual input interface, or a voice input interface.

17. The medium of claim 15, wherein the first dependency graph has the plurality of key words as nodes, and connects the plurality of key words based on a context and a sentence structure of the natural language based question.

18. The medium of claim 17, wherein the code further causes the processor to compare the first dependency graph and the second dependency graph to eliminate inaccurate query term mappings.

19. The medium of claim 15, wherein the code further causes the processor to:

instantiate a scoring mechanism to prioritize different query term mappings when multiple query terms from the previously stored query terms mapped to one key word from the plurality of key words.

20. The medium of claim 15, wherein the code further causes the processor to generate a graphical representation of the query result for display via the user interface.