Accuracy of QA systems by normalizing logical queries

1. A method of processing queries, comprising:

inputting, by one or more computer processors, a query as query data;

generating, by the one or more computer processors, a paraphrase from the query data, wherein generating the paraphrase further comprises:

constructing, by the one or more computer processors, a vector of a dependency path;

feeding, by the one or more computer processors, the vector of the dependency path to a neural network for learning a similarity of a pair of verbal paraphrases from the query data; and

extracting, by the one or more computer processors, one or more words from the query data;

rephrasing, by the one or more computer processors, the one or more words while preserving a logic intent of the query data; and

providing, by the one or more computer processors, a pattern, wherein the pattern encodes a condition under which the phrase is a paraphrase, and wherein the paraphrase expresses a same meaning; and

normalizing, by the one or more computer processors, the generated paraphrase according to predefined annotations of a schema, wherein normalizing the generated paraphrase further comprises:

comparing, by the one or more computer processors, the paraphrase generated with a schema annotation list; and

responsive to determining a match is not found in the comparison of the paraphrase generated with the schema annotation list, updating, by the one or more computer processors, the schema annotation list.

2. The method according to claim 1 , wherein comparing the paraphrase generated with the schema annotation list further comprises:

determining, by the one or more computer processors, whether and where a match exists between the paraphrase generated and the schema annotation list.

3. The method according to claim 1 , further comprising:

responsive to determining a match is found in the comparison of the paraphrase generated with the schema annotation list, returning, by the one or more processors, a query result.

4. The method according to claim 1 being cloud implemented.

5. The method according to claim 1 , wherein the logical operation with a predetermined logical function normalizes input queries according with predefined annotations of the schema, and

wherein the configuration of the annotation schema occurs after creation of the annotation schema with an update according to a comparison of a schema annotation list and a paraphrase list.

6. The method according to claim 1 , wherein inputting a query as query data further comprises:

extracting, by the one or more computer processors, one or more entities from the query data in a natural language format using a sub-tree pattern matching algorithm;

outputting, by the one or more computer processors, the one or more entities to a Structured Query Language conversion unit to convert the one or more entities into a Structured Query Language; and

returning, by the one or more computer processors, one or more query results.

7. The method according to claim 6 , wherein the sub-tree pattern matching algorithm is a rule-based algorithm.

8. The method according to claim 7 , wherein the sub-tree pattern matching algorithm receives as an input a Predicate Argument Structure output and a set of Lexical rules, and wherein the set of Lexical Rules are generated by training a set of Non-Lexical Rules with a Schema Annotation File in an offline setting.

9. A system for processing logical queries, comprising:

a memory storing computer instructions; and

a processor configured to execute the computer instructions to:

input a query as query data;

generate a paraphrase from the query data, wherein the computer instructions to generate the paraphrase further comprise:

construct a vector of a dependency path; and

feed the vector of the dependency path to a neural network for learning a similarity of a pair of verbal paraphrases from the query data; and

extract one or more words from the query data;

rephrase the one or more words while preserving a logic intent of the query data; and

provide a pattern, wherein the pattern encodes a condition under which the phrase is a paraphrase, and wherein the paraphrase expresses a same meaning; and

normalize the generated paraphrases according to predefined annotations of a schema, wherein the computer instructions to normalize the generated paraphrase further comprise:

compare the paraphrase generated with a schema annotation list; and

responsive to determining a match is not found in the comparison of the paraphrase generated with the schema annotation list, update the schema annotation list.

10. The system according to claim 9 , wherein the computer instructions to compare the paraphrase generated with the schema annotation list further comprise:

determine whether and where a match exists between the paraphrase generated and the schema annotation list.

11. The system according to claim 10 , the computer instructions further comprising to:

responsive to determining a match is found in the comparison of the paraphrase generated with the schema annotation list, return a query result.

12. The system according to claim 9 , wherein an annotation schema is used to normalize the generated paraphrases, and

wherein new information is added dynamically to the annotation schema according to a comparison of the generated paraphrases and the annotation schema.

13. The system according to claim 9 is cloud implemented.

14. The system according to claim 9 , wherein the computer instructions to input a query as query data further comprise:

extract one or more entities from the query data in a natural language format using a sub-tree pattern matching algorithm;

output the one or more entities to a Structured Query Language conversion unit to convert the one or more entities into a Structured Query Language; and

return one or more query results.

15. A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions readable and executable by a computer to cause the computer to perform a method, comprising:

inputting a query as query data;

generating a paraphrase from the query data, wherein generating the paraphrase further comprises:

constructing a vector of a dependency path;

feeding the vector of the dependency path to a neural network for learning a similarity of a pair of verbal paraphrases from the query data; and

extracting one or more words from the query data;

rephrasing the one or more words while preserving a logic intent of the query data; and

providing a pattern, wherein the pattern encodes a condition under which the phrase is a paraphrase, and wherein the paraphrase expresses a same meaning; and

normalizing the generated paraphrases according to predefined annotations of a schema, wherein normalizing the generated paraphrase further comprises:

comparing the paraphrase generated with a schema annotation list; and

responsive to determining a match is not found in the comparison of the paraphrase generated with the schema annotation list, updating the schema annotation list.

16. The computer program product according to claim 15 , wherein comparing the paraphrase generated with the schema annotation list further comprises:

determining whether and where a match exists between the paraphrase generated and the schema annotation list.

17. The computer program product according to claim 16 , further comprising:

responsive to determining a match is found in the comparison of the paraphrase generated with the schema annotation list, returning a query result.

18. The computer program product according to claim 15 , wherein inputting a query as query data further comprises:

extracting one or more entities from the query data in a natural language format using a sub-tree pattern matching algorithm;

outputting the one or more entities to a Structured Query Language conversion unit to convert the one or more entities into a Structured Query Language; and

returning one or more query results.