	US 12,481,497 B2
	Method and system for building and leveraging a knowledge fabric to improve software delivery lifecycle (SDLC) productivity
Nachiket Deshpande, Mumbai (IN); Aarya Karambelkar, Mumbai (IN); Adish Apte, Mumbai (IN); Arindam Bhattacharya, Mumbai (IN); Brijesh Prabhakar, Mumbai (IN); Devanathan Desikan, Plano, TX (US); Meena Malu, Mumbai (IN); and Sandeep Deb, Mumbai (IN)
Assigned to LTI Mindtree Ltd, Mumbai (IN)
Filed by LTI Mindtree Ltd, Mumbai (IN)
Filed on Sep. 21, 2022, as Appl. No. 17/949,211.
Claims priority of application No. 202221041505 (IN), filed on Jul. 20, 2022.
Prior Publication US 2024/0028327 A1, Jan. 25, 2024
Int. Cl. G06F 8/70 (2018.01); G06F 16/36 (2019.01); G06N 5/022 (2023.01)

CPC G06F 8/70 (2013.01) [G06F 16/367 (2019.01); G06N 5/022 (2013.01)]

14 Claims

1. A computer-implemented method, comprising:

constructing, by one or more processors, an NLP-based search engine on a transformers-based architecture to handle sequential data, wherein

the transformers-based architecture comprises a multi-head self-attention mechanism combined with an encoder-decoder structure that considers longer-distanced context around a word and enables parallelization techniques for Graphics Processing Unit (GPU) computing, and

the sequential data comprises at least one of natural language or tasks including translation or text summarization;

receiving, by the one or more processors, a plurality of Software Development Life Cycle (SDLC) artifacts of an SDLC from a plurality of heterogeneous data sources, wherein

the plurality of heterogeneous data sources comprises unstructured documents including word and pdf files, semi-structured documents including JSON and XML files, and structured databases including relational databases and

the plurality of SDLC artifacts comprises at least requirements, test cases, and defect logs;

correlating and clustering, by the one or more processors, the plurality of SDLC artifacts to generate a knowledge fabric, wherein the correlating and clustering comprises:

extracting, by the one or more processors, semantic and contextual data from the plurality of SDLC artifacts based on Natural Language Processing (NLP) and deep text analytics, wherein

the extracting comprises named entity extraction and information extraction for use in building knowledge organization systems or ontologies, and

the NLP includes using the NLP-based search engine on the transformers-based architecture; and

transforming, by the one or more processors, the extracted semantic and contextual data to one or more knowledge graphs to build the knowledge organization systems or the ontologies, wherein the one or more knowledge graphs comprise Resource Description Framework (RDF) triples, semantic web graphs, or labeled property graphs that are queryable by graph query languages;

deriving, by the one or more processors, one or more actionable items based on the knowledge fabric, wherein the one or more actionable items include change impact analysis for new change requests to the requirements, defect hotspot identification, root cause analysis, and contextual search results utilized to improve overall process efficiency of the SDLC;

automatically generating a defect prediction based on the knowledge fabric;

processing data that identifies a file impacted by a new requirement or a defect;

identifying a correction to the defect;

identifying, based on the knowledge fabric, time-series data from one or more application performance management tools;

combining the time-series data with Chaos engineering; and

highlighting performance issues and bottlenecks in one or more infrastructure resources during the SDLC, wherein the one or more infrastructure resources is at least one of CPU utilization, memory usage, network latency, or input/output (I/O) constraints.