Method and apparatus for predictive enrichment of search in an enterprise

• US 9,058,362 B2
• Filed: 04/10/2012
• Issued: 06/16/2015
• Est. Priority Date: 07/26/2011
• Status: Active Grant

• Current Assignee: [24]7.ai, Inc.
• Original Assignee: 24/7 Customer Incorporated
• Inventors: Vijayaraghavan, Ravi, Kumar, Gangadharan
• Primary Examiner(s): Choi, Yuk Ting
• Application Number:
  US13/443,782
• Publication Number:
  US 20130031081A1
• Time in Patent Office: 1,162 Days
• Field of Search: 707/727, 707/706
• US Class Current: 1/1
• CPC Class Codes: G06F 16/24575 using context
• CPC Class Codes: G06F 16/248 Presentation of query results
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06F 16/9538 Presentation of query results
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

With regard to searches and, more particularly, to searches performed on information repositories belonging to an enterprise, a centralized management system is used by the enterprise to manage the predictive search experience for users. A system offers a rich resolution experience to the end users based on user intent as determined from a variety of mechanisms, such as keywords, end user journey, clustered journey, etc. Also disclosed herein is a system that derives and offers various suggestions to end users to help them accomplish their objectives.

Claims

1. A computer implemented enterprise search apparatus, comprising:

one or more repositories comprising electronic storage media available within an enterprise;

an electronic database storing therein information related to one or more users;

a processor implemented search engine for receiving user queries via a user interface and, responsive thereto, performing searches on said one or more repositories;

upon receiving a user query from a user, said search engine identifying the user and fetching information related to the user from said database, the information fetched from said database comprising;

prior search words and results for searches performed by the user and other users who may be associated with the user;

a trail of the user in an enterprise domain;

a trail of the user prior to visiting the enterprise domain;

text mined chat transcripts;

meta data of chats;

text mined voice transcripts;

information present in social networks;

information related to the user, including attributes from a CRM system about the user such as billing data, past bills, mobile plan;

data from enterprise pages; and

forms filled by the user;

wherein information fetched from said database also depends on said user query;

a processor implemented recommendation engine for receiving from said search engine information fetched from said database by said search engine;

said recommendation engine probabilistically predicting a user issue type based upon both of the information related to the user fetched from the database and the search query from the user;

said probability engine determining probability for a given issue type/resolution need as a function of any of user attributes from CRM and other sources, history of prior contacts through all channels, behavior on a website, behavior on the Web before coming into a website and the search query from the user;

said recommendation engine executing an algorithm for predicting a probability that a given issue is an issue of interest for the user and returns results to the user related to said issue of interest; and

wherein meta data of chats comprises average handle time, average response time, disposition, standard deviation of response times, number of customer lines, and number of agent lines.

2. The apparatus of claim 1, wherein said search engine checks if the predicted search query provided by the recommendation engine needs further refining.

3. The apparatus of claim 2, wherein a decision for further refining of the search query provided by the recommendation engine is made by the user who has entered the search query;

wherein, if further refinement is to be done, said search engine informs a refinement engine.

4. The apparatus of claim 2, wherein a decision for further refining the query search query provided by the recommendation engine is automatically decided by the search engine based on a plurality of factors;

wherein, if further refinement is to be done, said search engine informs a refinement engine.

5. The apparatus of claim 4, wherein said factors comprise a number of results that are returned by the predicted search query.

6. The apparatus of claim 1, further comprising:

a refinement engine for checking if a predicted search query provided by said recommendation engine needs further refining.

7. The apparatus of claim 6, wherein said refinement engine, based upon any of a predicted search query from said recommendation engine, a search query entered by the user, and/or other information provided by said search engine, prepares a list of questions, which is sent to the user;

and

wherein, upon receiving a response from the user, said refinement engine further refines said predicted search query and sends said refined search query to said search engine.

8. The apparatus of claim 7, wherein upon receiving said refined search query, said search engine fetches corresponding results from said one or more repositories and displays said results to the user.

9. The apparatus of claim 1, wherein said search engine monitors user responses to search results displayed to the user by tracking any of clicks made by the user, information viewed by the user, and an amount of time the user spends on each piece of information;

wherein responses of the user are stored in said database and are used to improve searches made by the user and/or other users in the future.

10. The apparatus of claim 1, wherein said search engine requests that the user take a survey on the efficacy of the search results;

wherein any information provided by the user is used by the search engine to improve searches made by the user and/or other users in the future.

11. The apparatus of claim 1, further comprising:

a summarization engine for fetching chat and voice transcripts from conversations with the one or more users, and social media/blog/forum transcripts from said one or more repositories;

wherein said summarization engine text mines said transcripts and provides a summary content that is used to build a knowledge repository automatically and at scale;

wherein said transcripts are filtered using structured attributes comprising any of highest recommendation score, highest customer satisfaction score (CSAT), and highest resolution score for a given issue type; and

wherein said transcripts are summarized and stored in said database using Natural Language Processing based summarization techniques.

12. The apparatus of claim 1 said predicted search query comprising any of a key word or key string.

13. The apparatus of claim 1, said algorithm comprising any of Naï

ve Bayes, logistic regression, and/or support vector machines (SVM) algorithms.

14. The apparatus of claim 1, said user interface comprising any of a web site and an application available to the user.

15. A computer implemented method for enterprise searching, comprising:

providing a processor implemented search engine receiving a user search query from the user and fetching relevant information from an electronic database;

the information fetched from said database comprising;

prior search words and results for searches performed by the user and other users who may be associated with the user;

a trail of the user in an enterprise domain;

a trail of the user prior to visiting the enterprise domain;

text mined chat transcripts;

meta data of chats;

text mined voice transcripts;

information present in social networks;

information related to the user, including attributes from a CRM system about the user such as billing data, past bills, mobile plan;

data from enterprise pages; and

forms filled by the user;

wherein information fetched from said database also depends on said user query;

said search engine sending information fetched from said database to a processor implemented recommendation engine;

said recommendation engine probabilistically predicting an issue type given the information related to the user fetched from the database and the search query from the user;

a probability engine determining probability for a given issue type/resolution need as a function of any of user attributes from CRM and other sources, history of prior contacts through all channels, behavior on a website, behavior on the Web before coming into a website and the search query from the user;

said recommendation engine executing an algorithm for predicting a probability that a given issue is an issue of interest for the user and returns results to the user related to said issue of interest; and

wherein meta data of chats comprises average handle time, average response time, disposition, standard deviation of response times, number of customer lines, and number of agent lines;

said recommendation engine sending a resulting predicted search query to the search engine; and

said search engine returning corresponding results to said user.

16. The method of claim 15, wherein said predicted search query is in the form of any of a key word or key string.

17. The method of claim 15, said recommendation engine using any of Naï

ve Bayes, logistic regression, and/or support vector machines (SVM) algorithms to predict the probability of a given issue being the issue of interest for the user.

18. The method of claim 15, further comprising:

said search engine determining if the predicted search query provided by said recommendation engine needs further refining.

19. The method of claim 18, wherein, if further refinement is to be done, said search engine informing a refinement engine;

said refinement engine, based on any of predicted search query from said recommendation engine, a search query entered by the user, and other information provided by said search engine, asking questions of the user;

wherein, upon receiving a response from the user, said refinement engine finalizing an issue/query category, identifying at least one key article in a knowledge base, and returning said at least one key article to said user.

20. The method of claim 19, wherein if refinement is not required, said search engine finalizing said predicted search query;

and

said search engine fetching appropriate results from said repositories and displaying the results to the user.