US 12,260,176 B2
System and method for extracting issues based on trouble ticket mining
Vikrant Vikas Shimpi, Pune (IN); Maitreya Natu, Pune (IN); Monika Bhave, Pune (IN); and Siddartha Kshirsagar, Pune (IN)
Assigned to TATA CONSULTANCY SERVICES LIMITED, Mumbai (IN)
Filed by Tata Consultancy Services Limited, Mumbai (IN)
Filed on Sep. 29, 2022, as Appl. No. 17/936,822.
Claims priority of application No. 202121045391 (IN), filed on Oct. 6, 2021.
Prior Publication US 2023/0108067 A1, Apr. 6, 2023
Int. Cl. G06F 40/279 (2020.01); G06F 16/26 (2019.01)
CPC G06F 40/279 (2020.01) [G06F 16/26 (2019.01)] 15 Claims
OG exemplary drawing
 
1. A processor-implemented method comprising steps of:
receiving, via an input/output interface, a plurality of trouble tickets representing one or more information technology (IT) related issues, wherein the plurality of trouble tickets comprising a machine generated and a user generated trouble tickets;
inferring, via the one or more hardware processors, description of the machine generated, and the user generated trouble tickets based on language variations, Part of Speech (POS) tagging, sentence correctness and completeness, and presence of non-dictionary terms;
identifying, via the one or more hardware processors, patterns of the trouble ticket descriptions of the plurality of trouble tickets by understanding domain-specific parameters from the plurality of trouble tickets and identify patterns of the trouble ticket descriptions in the user generated trouble tickets by understanding jargons from the trouble tickets, cluster ambiguous and verbose user generated descriptions using vectorization by decoding semantics of issues;
clustering, via the one or more hardware processors, the machine generated, and the user generated trouble tickets separately based on the identified patterns of the trouble ticket descriptions of machine generated trouble tickets and further extract issues from the ticket descriptions by identifying top k-gram keywords within each cluster and creating a sub-cluster of all those descriptions having those top k-grams keywords, systematically choses an optimum ‘k’ value by following a k vs cluster-size matrix and continue performing these steps iteratively until issues are extracted from the trouble ticket descriptions;
selecting, via the one or more hardware processors, one or more techniques for the machine generated and user generated trouble tickets based on data properties, historical performance, and user feedback on each of the plurality of trouble tickets, wherein the one or more techniques comprising a pattern matching technique, a text similarity matching technique, and a keyword frequency based matching technique;
identifying, via the one or more hardware processors, one or more columns from a predefined service catalog based on the pattern matching technique, wherein the service catalog comprising one or more annotated issues;
comparing, via the one or more hardware processors, description of the plurality of the trouble tickets having weighted keywords with the identified one or more columns from the service catalog for matching with a similarity score;
labelling, via the one or more hardware processors, descriptions of the machine generated trouble tickets using the text similarity matching technique based extraction;
aggregating, via the one or more hardware processors, the labelled descriptions of the machine generated trouble ticket and the descriptions of the user generated trouble tickets to obtain trouble ticket descriptions with labels or without labels; and
learning, via the one or more hardware processors, one or more issues of the machine generated trouble tickets from the obtained trouble ticket descriptions with labels.