Cognitive conversational agent for providing personalized insights on-the-fly
A system, method and computer program product, which given in input a question in natural language format, delivers personalized insights related to the answer. Personalized insights are selected among candidate insights mined from the data and ranked based on closeness to (mined) user-preference, relevance to the question, and surprise factor. Two core components include: Question analysis and meaningful insight look up and Multi-dimensional insight ranking. The Question analysis and meaningful insights lookup module performs a semantic analysis of the questions and, uses techniques including “templates” to build new questions which could uncover insights from the data. The Multi-dimensional insight ranking module takes in input a list of insights returned from Question analysis and meaningful insights lookup and rank such insights based on such factors as: relevance to the query, surprise factor, and user preferences.
1. A computer-implemented method for supplementing insights from data in a question/answer system comprising:
receiving, at a hardware processor, via a user input interface, a question in a natural language (NL) format;
identifying, using the hardware processor, entities in said received NL question;
initiating, a query processor to conduct a first search of a data resource to obtain an answer to said received NL question;
generating, using the hardware processor, multiple candidate expanded questions based on said identified entities and said obtained answer;
conducting, using the query processor, a second search at the resource to obtain corresponding answers to said multiple candidate questions;
selecting, using the hardware processor, a sub-set of candidate questions and corresponding answers from among said candidate expanded questions based upon one or more criteria, the selecting a sub-set of candidate questions comprising
ranking, using the hardware processor, said multiple candidate questions and the corresponding answers based on a user-preference, a relevance to the received NL question, and a surprise factor; and
presenting to said user, via said user interface, additional personal insight data related to said answer of said received original question based upon said selected sub-set of corresponding answers.
2. The computer-implemented method of claim 1 , wherein said identifying said entities in said received NL question comprises:
conducting, using a natural language processor, a semantic analysis of the received NL question to generate a dependency tree structure relating words of said received question and associated word types, and
determine said entities based on said relating words and word types.
3. The computer-implemented method of claim 2 , wherein said generating the multiple candidate expanded questions comprises:
applying, using the hardware processor, one or more templates for building said multiple candidate questions for use in discovering additional insights from the data.
4. The computer-implemented method of claim 3 , wherein said templates are based on one or more of: a trend in a pre-defined time period, a content-hierarchy, a user preference, a similarity to an identified entity.
5. The computer-implemented method of claim 1 , wherein said
ranking, using the hardware processor, said multiple candidate questions and the corresponding answers is further based on one or more of: a detected anomaly, and a time series.
6. The computer-implemented method of claim 5 , wherein said user preference comprises one of: a user job role, one or more historical interactions, a time of the day a type of device used to query the system, and a behavior of a cohort of users.
7. The computer-implemented method of claim 1 , conducting, using said hardware processor, a semantic query optimization for efficiently forming queries that limit a number of said multiple candidate expanded questions to obtain said answers.
8. An advisory system comprising:
a memory storage device; and
a hardware processor coupled to said memory storage device and configured to perform a method to:
receive, via a user input interface, a question in a natural language (NL) format;
identify entities in said received NL question;
initiate a query processor to conduct a first search of a data resource to obtain an answer to said received NL question;
generate multiple candidate expanded questions based on said identified entities and said obtained answer;
conduct, using the query processor, a second search at the resource to obtain corresponding answers to said multiple candidate questions;
select a sub-set of candidate questions and corresponding answers from among said candidate expanded questions based upon one or more criteria, wherein to select a sub-set of candidate questions, the hardware processor is further configured to:
rank said multiple candidate questions and the corresponding answers based on a user-preference, a relevance to the received NL question, and a surprise factor; and
present to said user, via said user interface, additional personal insight data related to said answer of said received original question based upon said selected sub-set of corresponding answers.
9. The advisory system of claim 8 , wherein to identify said entities in said received NL question, said hardware processor is further configured to:
conducting, using a natural language processor, a semantic analysis of the received NL question to generate a dependency tree structure relating words of said received question and associated word types, and
determine said entities based on said relating words and word types.
10. The advisory system of claim 9 , wherein to generate the multiple candidate expanded questions, said hardware processor is further configured to:
apply one or more templates for building said multiple candidate questions for use in discovering additional insights from the data.
11. The advisory system of claim 10 , wherein said templates are based on one or more of: a trend in a pre-defined time period, a content-hierarchy, a user preference, a similarity to an identified entity.
12. The advisory system of claim 8 , wherein to select said sub-set of candidate questions and corresponding answers, said hardware processor is further configured to:
rank said multiple candidate questions and the corresponding answers based further on one or more of: a detected anomaly, and a time series.
13. The advisory system of claim 12 , wherein said user preference comprises one of:
a user job role, one or more historical interactions, a time of the day a type of device used to query the system, and a behavior of a cohort of users.
14. The advisory system of claim 8 , wherein said hardware processor is further configured to:
conduct a semantic query optimization for efficiently forming queries that limit a number of said multiple candidate expanded questions to obtain said answers.
15. A non-transitory computer readable medium comprising instructions that, when executed by at least one processor comprising hardware, configure the at least one processor to:
receive, via a user input interface, a question in a natural language (NL) format;
identify entities in said received NL question;
initiate a query processor to conduct a first search of a data resource to obtain an answer to said received NL question;
generate multiple candidate expanded questions based on said identified entities and said obtained answer;
conduct, using the query processor, a second search at the resource to obtain corresponding answers to said multiple candidate questions;
select a sub-set of candidate questions and corresponding answers from among said candidate expanded questions based upon one or more criteria, wherein to select a sub-set of candidate questions, the at least one processor is further configured to:
rank said multiple candidate questions and the corresponding answers based on a user-preference, a relevance to the received NL question, and a surprise factor; and
present to said user, via said user interface, additional personal insight data related to said answer of said received original question based upon said selected sub-set of corresponding answers.
16. The non-transitory computer readable medium according to claim 15 , wherein said instructions further comprise instructions to:
conduct, using a natural language processor, a semantic analysis of the received NL question to generate a dependency tree structure relating words of said received question and associated word types, and
determine said entities based on said relating words and word types.
17. The non-transitory computer readable medium according to claim 15 , wherein to generate the multiple candidate expanded questions, said instructions further comprise instructions to:
apply one or more templates for building said multiple candidate questions for use in discovering additional insights from the data, said templates being based on one or more of: a trend in a pre-defined time period, a content-hierarchy, a user preference, a similarity to an identified entity.
18. The non-transitory computer readable medium according to claim 15 , wherein to select said sub-set of candidate questions and corresponding answers, said instructions further comprise instructions to:
rank said multiple candidate questions and the corresponding answers based further on one or more of: a detected anomaly, and a time series.
19. The non-transitory computer readable medium according to claim 18 , wherein said user preference comprises one of: a user job role, one or more historical interactions, a time of the day a type of device used to query the system, and a behavior of a cohort of users.
20. The non-transitory computer readable medium according to claim 15 , wherein said instructions further comprise instructions to:
conduct a semantic query optimization for efficiently forming queries that limit a number of said multiple candidate expanded questions to obtain said answers.