System and method for determining user knowledge scores based on topic analysis of mapped content
Various aspects of the subject technology related to systems and methods for mapping communication content based on topic analysis to determine user knowledge scores. A system may be configured to receive communication content including spoken and non-spoken content. The system may convert spoken content to non-spoken content and a data set of the communication content may be generated. The system may construct a generative statistical model identifying a list of topics in the data set. The generative statistical model may also identify the topic percentage of words in the data set that are associated with a given topic. The system may generate a communication content graph identifying user nodes and word edges. The system may calculate a topic strength for each word edge and determine a user knowledge score for each user associated with each identified topic. The system may output the user knowledge score.
1. A computer-implemented method for mapping communication content based on topic analysis to determine user knowledge scores, the method comprising:
receiving a plurality of communication content, wherein the plurality of communication content comprises a non-spoken content;
converting, as converted content, a spoken content to an additional non-spoken content;
generating from the communication content, a data set, wherein the data set comprises a plurality of words contained within the converted content and the non-spoken content;
constructing, from the data set, a generative statistical model that identifies a list of topics from the data set and identifies a topic percentage determined from a probability that a word in the data set is associated with a given topic in the list of topics;
generating a communication content graph comprising user nodes and word edges for each identified topic in the list of topics;
calculating a topic strength for each word edge between two different users, wherein the topic strength is calculated by multiplying the topic percentage by a number of words in the communication content, and is indicative of a likelihood that the communication content shared between the two different users is associated with a particular topic;
determining, based on the topic strengths, a user knowledge score for each user associated with each identified topic, wherein the user knowledge score is determined by adding the topic strength associated with each topic for all adjacent edges in the communication content graph;
outputting the user knowledge score; and
determining that the user is an expert on the topic based on the user knowledge score.
2. The computer-implemented method of claim 1 , further comprising:
reviewing the output of the generative statistical model and manually identifying the list of topics.
3. The computer-implemented method of claim 1 , wherein the plurality of communication content comprises one or more of: call recordings, instant messages, text messages, audio meeting recordings, emails, calendar appointments and video meeting recordings.
4. The computer-implemented method of claim 1 , wherein constructing the generative statistical model is performed using a Latent Dirichlet allocation method, an unsupervised topic modeling method, or a generative probabilistic topic modeling method.
5. The computer-implemented method of claim 1 , wherein generating the communication content graph further comprises normalizing the communication content of the data set.
6. The computer-implemented method of claim 1 , wherein the topic strength is further calculated by further multiplying by a communication content weight.
7. The computer-implemented method of claim 6 , wherein the communication content weight is determined as a function of an age of the communication content or a type of communication content.
8. The computer-implemented method of claim 6 , wherein the communication content weight is provided by the user.
9. The computer-implemented method of claim 1 , wherein the user knowledge score is output for use in a topic search user interface.
10. The computer-implemented method of claim 1 , wherein the user knowledge score is output for use in a topic browsing user interface.
11. The computer-implemented method of claim 1 , wherein the user knowledge score is determined by summing the topic strength associated with each topic for all edges in the communication content graph.
12. A system for mapping communication content based on topic analysis to determine user knowledge scores, the system comprising:
a memory comprising instructions; and
one or more processors configured to execute instructions, which, when executed, cause the one or more processors to:
receive a plurality of communication content, wherein the plurality of communication content comprises a non-spoken content;
convert, as converted content, a spoken content to an additional non-spoken content;
generate, from the communication content, a data set, wherein the data set comprises a plurality of words contained within the converted content and the non-spoken content;
construct, from the data set, a generative statistical model that identifies a list of topics from the data set and identifies a topic percentage determined from a probability that a word in the data set is associated with a given topic in the list of topics;
generate a communication content graph comprising user nodes and word edges for each identified topic in the list of topics;
calculate a topic strength for each word edge between two different users, wherein the topic strength is calculated by multiplying the topic percentage by a number of words in the communication content, and is indicative of a likelihood that the communication content shared between the two different users is associated with a particular topic;
determine, based on the topic strengths, a user knowledge score for each user associated with each identified topic, wherein the user knowledge score is determined by adding the topic strength associated with each topic for all adjacent edges in the communication content graph;
output the user knowledge score for use or display in a topic search user interface or a topic browsing user interface; and
determine that the user is an expert on the topic based on the user knowledge score.
13. The system of claim 12 , wherein the one or more processors are further configured to:
display the output of the generative statistical model to the user and receive user input to identify the list of topics.
14. The system of claim 12 , wherein the plurality of communication content comprises one or more of: call recordings, instant messages, text messages, audio meeting recordings, emails, calendar appointments and video meeting recordings.
15. The system of claim 12 , wherein the generative statistical model comprises a Latent Dirichlet allocation model, an unsupervised topic model or a generative probabilistic topic model.
16. The system of claim 12 , wherein the one or more processors are further configured to normalize the communication content of the data set to generate the communication content graph.
17. The system of claim 12 , wherein the topic strength is calculated by further multiplying by a communication content weight.
18. The system of claim 17 , wherein the communication content weight is determined as a function of an age of the communication content or a type of communication content.
19. The system of claim 12 , wherein the user knowledge score is determined by summing the topic strength associated with each topic for all edges in the communication content graph.
20. A machine readable storage medium containing program instructions for causing a computer to map communication content based on topic analysis to determine user knowledge scores performed by a method of:
receiving a plurality of communication content, wherein the plurality of communication content comprises a non-spoken content generated by employees of an organization;
converting, as converted content, a spoken content to an additional non-spoken content;
generating, from the communication content, a data set, wherein the data set comprises a plurality of words contained within the converted content and the non-spoken content;
constructing, from the data set using a Latent Dirichlet allocation method, a generative statistical model that identifies a list of topics from the data set and identifies a topic percentage determined from a probability that a word in the data set is associated with a given topic in the list of topics;
normalizing the communication content of the data set in the generative statistical model;
generating, based on the communication content, a communication content graph comprising employee nodes and word edges for each identified topic in the list of topics;
calculating a topic strength for each word edge between two different users by multiplying the topic percentage by a number of words in the communication content, which is indicative of a likelihood that the communication content shared between the two different users is associated with a particular topic;
determining, based on the topic strengths, an employee knowledge score for each employee associated with each identified topic, wherein the employee knowledge score is determined by summing the topic strengths associated with each topic for all adjacent edges in the communication content graph;
outputting the employee knowledge score to a database; and
determining that the employee is an expert on the topic based on the user knowledge score.