| CPC G06F 40/35 (2020.01) [G06F 16/35 (2019.01); G06F 40/284 (2020.01); G06N 20/00 (2019.01); G06Q 30/01 (2013.01); H04L 51/02 (2013.01); G06F 16/3329 (2019.01)] | 14 Claims |
|
1. A system comprising:
a processing resource; and
a non-transitory computer-readable medium, coupled to the processing resource, having stored therein instructions that when executed by the processing resource cause the system to:
receive, via a chatbot, free text input from a user for a product support case, wherein the free text input describes an issue associated with a product line of a vendor;
create a vector representation of the issue by tokenizing and vectorizing at least a portion of the free text input using a word association model of a plurality of word association models corresponding to the product line, wherein:
each of the plurality of word association models is based on a set of historical support cases relating to a plurality of supported issue categories for a respective product line of a plurality of product lines of the vendor that are remotely resolvable and meet a predefined threshold of criticality,
each of the plurality of word association models are generated by applying a topic model to identify clusters of the set of historical support cases according to a taxonomy at a component level within the respective product line and at a category level within the plurality of supported issue categories for the respective product line, and
the component level corresponds to components of the respective product line;
determine whether the issue matches a category within the plurality of supported issue categories for the product line by applying an intermediate classification model to the vector representation to determine whether the vector representation can be matched to a corresponding vector representation output by the word association model and representing a particular issue category supported by the chatbot,
wherein:
the intermediate classification model comprises a product-line specific long short-term memory (LSTM) model, and
the LSTM model is trained via a supervised learning process for each of the plurality of product lines to train product line specific LSTM models; and
responsive to an affirmative determination, initiate an automated, interactive, conversational troubleshooting dialog via the chatbot with the user based on a decision tree for the category within the product line.
|