Expert recommendation leveraging topic clusters derived from unstructured text data
An apparatus comprises a processing platform configured to implement an expert recommender engine. The expert recommender engine receives information relating to a communication from a user device, and identifies at least one subject matter expert for the communication based on the received information and unstructured text data of a service events database. The expert recommender engine is associated with a clustering module that separates the unstructured text data into topic clusters. The expert recommender engine comprises a collaborative filtering module that receives the topic clusters from the clustering module and utilizes the topic clusters to identify the subject matter expert. The user device is connected with an expert device corresponding to the identified subject matter expert. The expert recommender engine may utilize structured data, social media data and customer satisfaction survey data in combination with the received information and the topic clusters to identify the subject matter expert.
1. An apparatus comprising:
a processing platform comprising one or more processing devices each comprising a processor coupled to a memory;
the processing platform being configured:
to receive information relating to a communication from a user device, the communication comprising a service request;
to identify at least one subject matter expert for the communication based on the received information and unstructured text data of a service events database, the unstructured text data comprising a plurality of documents associated with previous service requests, the plurality of documents comprising at least one unstructured service request summary comprising one or more problem summaries and one or more corresponding solution summaries;
to separate the unstructured text data into topic clusters for a plurality of topics, at least a subset of the plurality of topics being determined automatically from the unstructured text data without reference to a set of rules characterizing predefined topics; and
to connect the user device with an expert device corresponding to the identified subject matter expert
wherein determining at least the subset of the plurality of topics automatically from the unstructured text data without reference to a set of rules characterizing predefined topics comprises:
processing the unstructured service request summaries of the plurality of documents to construct a term index of terms utilized in the unstructured service request summaries;
generating, for a domain comprising the unstructured service request summaries of the plurality of documents, an in-domain dictionary by processing the term index utilizing automatic lemmatization and synonym extraction;
constructing a topic model by processing the in-domain dictionary; and
determining a list of topics utilizing the topic model, wherein the list of topics comprises at least one topic elevated as a set of related terms from the unstructured request summaries of the plurality of documents;
wherein connecting the user device with the expert device corresponding to the identified subject matter expert further comprises delivering one or more visualizations to the expert device, the one or more visualizations comprising at least one of:
a bigram view visualization of a plurality of term pairs from a selected topic cluster;
a summarization view visualization of representative term sequences from the selected topic cluster; and
a unigram and aggregate probability view visualization of a plurality of individual terms from the selected topic cluster, the aggregate probability comprising a combination of individual probabilities that respective ones of the terms appear in the selected topic cluster.
2. The apparatus of claim 1 wherein the user device comprises a device associated with one of a customer and a support person.
3. The apparatus of claim 1 wherein the processing platform is configured to assign each of the documents to one or more of the topic clusters.
4. The apparatus of claim 3 wherein the processing platform is configured to assign the documents to the topic clusters in a manner that permits documents not sharing any common technical phrases to be assigned to the same topic cluster.
5. The apparatus of claim 1 wherein the processing platform is configured to determine one or more of the topic clusters that are most closely related to the received information and to utilize the determined one or more topic clusters to identify the subject matter expert.
6. The apparatus of claim 1 wherein the processing platform is configured to obtain structured data of the service events database and to utilize the structured data in combination with the received information and the topic clusters to identify the subject matter expert, and wherein the structured data comprises service event data stored in one or more structured data fields of the service events database.
7. The apparatus of claim 1 wherein the processing platform is configured to obtain social media data and to utilize the social media data in combination with the received information and the topic clusters to identify the subject matter expert.
8. The apparatus of claim 1 wherein the processing platform is configured to obtain customer satisfaction survey data and to utilize the customer satisfaction survey data in combination with the received information and the topic clusters to identify the subject matter expert.
9. The apparatus of claim 1 wherein the processing platform is configured to generate a ranked listing of subject matter experts and to identify an available subject matter expert from the ranked listing, and wherein the user device is connected with the expert device corresponding to the available subject matter expert.
10. An information processing system comprising the apparatus of claim 1 .
11. A method comprising:
receiving information relating to a communication from a user device, the communication comprising a service request;
obtaining unstructured text data from a service events database, the unstructured text data comprising a plurality of documents associated with previous service requests, the plurality of documents comprising at least one unstructured service request summary comprising one or more problem summaries and one or more corresponding solution summaries;
separating the unstructured text data into topic clusters for a plurality of topics, at least a subset of the plurality of topics being determined automatically from the unstructured text data without reference to a set of rules characterizing predefined topics;
utilizing the received information and the topic clusters to identify at least one subject matter expert for the communication; and
connecting the user device with an expert device corresponding to the identified subject matter expert;
wherein the receiving, obtaining, separating, utilizing and connecting are performed by a processing platform comprising one or more processing devices;
wherein determining at least the subset of the plurality of topics automatically from the unstructured text data without reference to a set of rules characterizing predefined topics comprises:
processing the unstructured service request summaries of the plurality of documents to construct a term index of terms utilized in the unstructured service request summaries;
generating, for a domain comprising the unstructured service request summaries of the plurality of documents, an in-domain dictionary by processing the term index utilizing automatic lemmatization and synonym extraction;
constructing a topic model by processing the in-domain dictionary; and
determining a list of topics utilizing the topic model, wherein the list of topics comprises at least one topic elevated as a set of related terms from the unstructured request summaries of the plurality of documents;
wherein connecting the user device with the expert device corresponding to the identified subject matter expert further comprises delivering one or more visualizations to the expert device, the one or more visualizations comprising at least one of:
a bigram view visualization of a plurality of term pairs from a selected topic cluster;
a summarization view visualization of representative term sequences from the selected topic cluster; and
a unigram and aggregate probability view visualization of a plurality of individual terms from the selected topic cluster, the aggregate probability comprising a combination of individual probabilities that respective ones of the terms appear in the selected topic cluster.
12. The method of claim 11 wherein separating the unstructured text data into topic clusters comprises assigning each of the documents to one or more of the topic clusters.
13. The method of claim 11 wherein utilizing the received information and the topic clusters to identify at least one subject matter expert further comprises at least one of:
obtaining structured data of the service events database and utilizing the structured data in combination with the received information and the topic clusters to identify the subject matter expert wherein the structured data comprises service event data stored in one or more structured data fields of the service events database;
obtaining social media data and utilizing the social media data in combination with the received information and the topic clusters to identify the subject matter expert; and
obtaining customer satisfaction survey data and utilizing the customer satisfaction survey data in combination with the received information and the topic clusters to identify the subject matter expert.
14. A non-transitory processor-readable storage medium having program code of one or more software programs embodied therein, wherein the program code when executed by at least one processing device of a processing platform causes the processing device:
to receive information relating to a communication from a user device, the communication comprising a service request;
to obtain unstructured text data from a service events database, the unstructured text data comprising a plurality of documents associated with previous service requests, the plurality of documents comprising at least one unstructured service request summary comprising one or more problem summaries and one or more corresponding solution summaries;
to separate the unstructured text data into topic clusters for a plurality of topics, at least a subset of the plurality of topics being determined automatically from the unstructured text data without reference to a set of rules characterizing predefined topics;
to utilize the received information and the topic clusters to identify at least one subject matter expert for the communication; and
to connect the user device with an expert device corresponding to the identified subject matter expert;
wherein determining at least the subset of the plurality of topics automatically from the unstructured text data without reference to a set of rules characterizing predefined topics comprises:
processing the unstructured service request summaries of the plurality of documents to construct a term index of terms utilized in the unstructured service request summaries;
generating, for a domain comprising the unstructured service request summaries of the plurality of documents, an in-domain dictionary by processing the term index utilizing automatic lemmatization and synonym extraction;
constructing a topic model by processing the in-domain dictionary; and
determining a list of topics utilizing the topic model, wherein the list of topics comprises at least one topic elevated as a set of related terms from the unstructured request summaries of the plurality of documents;
wherein connecting the user device with the expert device corresponding to the identified subject matter expert further comprises delivering one or more visualizations to the expert device, the one or more visualizations comprising at least one of:
a bigram view visualization of a plurality of term pairs from a selected topic cluster;
a summarization view visualization of representative term sequences from the selected topic cluster; and
a unigram and aggregate probability view visualization of a plurality of individual terms from the selected topic cluster, the aggregate probability comprising a combination of individual probabilities that respective ones of the terms appear in the selected topic cluster.
15. The processor-readable storage medium of claim 14 wherein the storage medium comprises at least one of an electronic memory and a storage disk.
16. The processor-readable storage medium of claim 14 wherein the program code when executed by at least one processing device further causes said processing device:
to obtain at least one of structured data, social media data and customer satisfaction survey data; and
to utilize the obtained data in combination with the received information and the topic clusters to identify the subject matter expert.
17. The apparatus of claim 3 wherein the processing platform is configured to update the service events database to indicate the assignment of each of the documents to one or more of the topic clusters.
18. The apparatus of claim 1 wherein the service request of the received communication is associated with a given user, and wherein the plurality of documents comprise documents associated with previous service requests from a plurality of users including at least one user other than the given user.
19. The apparatus of claim 1 wherein constructing the topic model comprises utilizing latent Dirichlet allocation (LDA) with asymmetric priors.
20. The apparatus of claim 1 wherein constructing the topic model comprises utilizing at least one of probabilistic latent semantic analysis (pLSA) and canonical-correlation analysis (CCA).