Machine learning approach for query resolution via a dynamic determination and allocation of expert resources

Abstract

The systems and methods described herein relate to mapping and identifying expert resources. The systems and methods described herein may provide a set of technologies, that work together as one solution, to effectively and efficiently resolve user questions. A cognitive engine may autonomously learn which experts have the knowledge to quickly solve a question or whether a previous question is similar enough to provide a solution instantly. Using machine learning, a know-how map may be created, linking all of the users of the system with their areas of expertise. Expert resources among the users may be mapped by determining connections between topics (and their corresponding tags) and calculating an expert score related to each topic for each user. These connections and expert scores are subsequently used during expert routing for each new question, to find those users with the expertise to give the best possible solution.

Description

CROSS-REFERENCE TO RELATED APPLICATION[0001]This application is a continuation of pending U.S. patent application Ser. No. 15 / 910,372, entitled “MACHINE LEARNING APPROACH FOR QUERY RESOLUTION VIA A DYNAMIC DETERMINATION AND ALLOCATION OF EXPERT RESOURCES”, filed Mar. 2, 2018, which is hereby incorporated herein by reference in its entirety.FIELD OF THE INVENTION[0002]The systems and methods described herein relate to a machine learning-based approach for resolving user queries via a dynamic determination and allocation of expert resources.BACKGROUND OF THE INVENTION[0003]Conventional approaches to mapping and identifying expert resources suffers from one or more drawbacks. For example, in large organizations, the number of users and volume of content can make mapping and identifying expert resources problematic. Models of an organization's expert resources may be unable to factor in all of the available data and differentiate between users with a similar background or level of exper...

AI Technical Summary

Benefits of technology

[0004]The systems and methods described herein describe a comprehensive knowledge management technology tool to address these and other issues. According to one aspect of the invention, the system may provide a set of technologies, that work together as one solution, to effectively and efficiently resolve user queries. The cognitive engine autonomously learns which experts have the knowledge to quickly solve a question or whether a previous question is similar enough to provide a solution instantly. The system uses predictive analytics to find those individuals who can best answer a particular question at the time a question is generated as well as using autocomplete suggestions and automatically searching for similar previously stored questions while the user is inputting a question. In this way, if an answer to the question exists in the system, the system can provide that information to the user. If it is a new question, the system can make a real-time, dynamic determination of who can be respond to the question at that time.

[0005]According to one aspect of the invention, the system uses machine learning to create a know-how map, linking all of its users with their areas of expertise. The system may use unsupervised learning techniques, which are automated and require no human effort. The learning algorithms automatically adapt themselves to the different topics and use-cases that occur in different organizations. In this way, the system creates a dynamically changing knowledge and user map.

[0010]In various implementations, one or more expert users to route the question may be identified based on the set of tags associated with the question and the expert scores for each user related to the set of tags. In doing so, users may be identified whose expertise best matches the particular combination of tags associated with the question. In some implementations, users identified as an expert user to route a question may be identified from a set of users based on the workload of each of the set of users (i.e., load balancing). By analyzing the activity of different users by time and day of the week, we can alert those experts for a new question who are likely to be available to provide a solution without delay. In doing so, the systems and methods described herein may make sure that the highest scoring experts are not demotivated by receiving too many questions. In some implementations, the user identified as the expert user to route the question may be identified from a set of users based on the language proficiencies of each of the set of users. The user identified as the expert user may be provided with the question and prompted to provide a solution.

[0011]Combining all of these factors, the systems and method described herein may obtain an ordered list of users identified as expert users. The ordered list of users may represent the top experts (i.e., the users with the highest expert score) for a given question at a given time. One or more of the identified expert users may be alerted and provided with the question. For example, in some implementations, between four and ten expert users identified for a question may be provided with the question to maximize the likelihood of getting a quality solution or answer to the question as soon as possible.

Problems solved by technology

The system may use unsupervised learning techniques, which are automated and require no human effort.

Images