Clustering and tagging engine for use in product support systems

Patent metadata

Field	Value
Publication number	US-11023774-B2
Application number	US-201816122809-A
Country	US
Kind code	B2
Filing date	Sep 5, 2018
Priority date	Jan 12, 2018
Publication date	Jun 1, 2021
Grant date	Jun 1, 2021

Abstract

The present invention relates to a computer-based system for supporting Product Customer Support Systems by means for parameter-free and fully unsupervised: clustering of a selected set documents (e.g., based on a query from some database) with unknown ontology (e.g., cases from Customer Support System); building a taxonomy for sets of documents with unknown ontology/taxonomy; enabling a semi-supervised tagging/navigation/recommendations for documents and cross-learning using auxiliary sources (e.g., linking other fields/metadata in Customer Support Systems such as Knowledge DataBase).

First claim

I claim:

1. A computer-based Product Customer Support System (“PCSS”) comprising one or more memory components and one or more processors in communication over one or more communications networks with a plurality of remote customer-operated devices to provide a product resource, the PCSS comprising:
a Product Knowledge Database (“PKD”) comprising a set of known solution records and a set of historical data records, the set of known solution records being related to at least on product and comprising product resolution data, and the set of historical data records being related at least one product and comprising product inquiry data;
a server system adapted to communicate with the PKD and with remote customer-operated devices and comprising a set of processors, and a set of memory components adapted to store code executable by the set of processors;
a user interface adapted to receive inputs from users and to present users with agent-generated prompts, the inputs and prompts related to a product inquiry, wherein the user interface is adapted to receive product related queries as user inputs, to present to users agent-generated questions related to the received queries as prompts for further information, and to receive user responses to the agent-generated questions;
a tagging engine adapted to identify, extract and tag data from at least one of the inputs and prompts and generate a set of tagged inquiry data, the tagged inquiry data including a service issue feature identified and extracted from the product inquiry and tagged by the tagging engine;
a clustering engine adapted to perform unsupervised hierarchical clustering at a plurality of hierarchical levels in at least one of the following domains: documents-similarity domain; features domain; and joint clustering, the clustering engine further adapted to derive a diagnosis taxonomy for addressing the service issue feature identified and tagged by the tagging engine by generating a set of clusters based on comparing the tagged inquiry data with at least one of known solution records, historical data records, clusters of known solution records and historical data records, the diagnosis taxonomy comprising at least a first primary level topic set representing a first node of a cluster and a second level topic set representing a secondary level set of features representing a set of sub-nodes of the cluster connected with the first node by a set of edges, wherein the cluster comprises a set of nodes and sub-nodes related to the service issue feature, wherein the clustering engine is adapted to perform similarity and feature scoring to determine if the tagged inquiry data is closely associated with at least one of an existing record, document, cluster of records, and cluster of documents stored in the PKD, wherein the PKD receives and stores the tagged inquiry data and generates a new historical data record including at least one topic comprised of at least one of: customer inquiry data, agent question data, user response data, product data, resolution data, and recommendation data; and
a recommendation engine adapted to generate for output a set of documents including recommendations responsive to the product inquiry; and wherein the user interface includes elements adapted to present a set of suggested recommendations from which the user may select as being responsive to the user inquiry.

2. The PCSS of claim 1 , wherein the clustering engine is adapted to connect at least one of documents, product data, problem data, solution data, recommendation data, tagging and classification data, and other product related information into a cluster network.

3. The PCSS of claim 2 , wherein the clustering engine is further adapted to process an additional set of documents to make additional associations for storing in the PKD, and wherein at least one of the additional set of documents are stored in the PKD as historical data records and available for use in clustering by the clustering engine.

4. The PCSS of claim 1 further comprising a discovery engine adapted to extract and tag keyword data to allow analyst-type users to classify and navigate over at least one of historical data records and known solution records to identify trends related to user inquiries, and adapted to provide at least one of cross-mapping and cross-learning using mapping extracted taxonomies from different topical domains associated with at least one of historical data records and known solution records.

5. The PCSS of claim 1 , wherein the tagged inquiry data includes tagged data derived from the agent-generated questions and/or the received user responses.

6. The PCSS of claim 1 , further comprising an adaptive network pruning module adapted to amplify clustering based on adaptive thresholding.

7. The PCSS of claim 1 , wherein the clustering engine if further adapted to provide taxonomy extraction based on at least one of documents-similarity domain and features domain.

8. The PCSS of claim 1 , wherein the tagging engine is adapted to use natural language processing techniques to identify, extract and tag data from at least one of the inputs and prompts and generate the set of tagged inquiry data.

9. The PCSS of claim 1 , wherein the tagging engine is adapted to use term frequency (tf), inverse term frequency (idf), and/or tf-idf functions to identify, extract and tag data from at least one of the inputs and prompts and generate the set of tagged inquiry data.

10. The PCSS of claim 1 , wherein the PKD is integrated with or interconnected with a CRM/ERP system.

11. The PCSS of claim 1 , wherein the PKD includes an existing knowledge database having data related to an existing first set of products, and the PCSS is adapted to use features and similarities based on at least one of tagged information and known solutions to generate a set of known solution records for use with a new product not included in the existing first set of products.

Assignees

• Thomson Reuters Entpr Centre Gmbh

Inventors

• Nefedov Nikolai

Classifications

• G06F18/23

  Clustering techniques · CPC title

• G06F16/35 Primary

  Clustering; Classification · CPC title

• H04L41/5074

  Handling of user complaints or trouble tickets · CPC title

• G06Q30/015

  Providing customer assistance, e.g. assisting a customer within a business location or via helpdesk · CPC title

• G06Q10/06316

  Sequencing of tasks or work · CPC title