Software Incident Vectorization for Support Record Retrieval
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Software system incidents often require manual intervention by support technicians, who rely on personal knowledge or documentation, leading to inefficiencies and knowledge gaps, especially when subject matter experts are scarce.
Innovation Solution
A computer-implemented incident management system that processes textual incident reports to generate a regularized report, determine importance measures, create an incident matrix, and apply machine learning models to identify relevant software support records and modules, acting as a virtual subject matter expert.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If text-based searches are used to identify relevant support records and modules, then the system can process incidents automatically, but the accuracy and efficiency are insufficient
Solution Approach 1:
The patent transforms the incident identification process from simple text-based keyword matching to a vector-based semantic representation. By converting incident reports, support records, and module documentation into vectors using word embeddings (such as Word2Vec or GloVe), the system captures semantic meaning rather than relying on exact keyword matches. This parameter change from discrete text tokens to continuous vector representations enables both higher efficiency (through vector similarity computations) and improved accuracy (through semantic understanding).
Solution Approach 2:
The patent replaces the mechanical text-search mechanism with a machine learning-based vector similarity system. Instead of manually searching through text databases using keywords, the system uses pre-trained word embedding models to automatically convert text into vectors and compute cosine similarity. This substitution eliminates the need for manual search operations and provides consistent, scalable, and accurate relevance identification across large volumes of support records and documentation.
2Ease of operation
If support technicians rely on personal knowledge and documentation, then they can address incidents, but knowledge gaps and inefficiencies occur when experts are scarce
Solution Approach 1:
The patent implements a self-service incident management system that automatically identifies relevant support records and software modules without requiring deep expert knowledge from technicians. The vector-based retrieval system enables technicians to query the database using natural language incident descriptions, and the system autonomously retrieves the most relevant historical support records and identifies potentially affected software modules. This self-service capability democratizes access to expert knowledge stored in the support records, making it available to all technicians regardless of their individual expertise level.
Solution Approach 2:
The patent introduces a vector similarity search system as an intermediary between the incident report and the support records database. Instead of requiring technicians to directly search through vast amounts of documentation and rely on their personal knowledge, the vector intermediary automatically bridges the gap by computing semantic similarity between the incident description and stored support records. This intermediary system captures and leverages embedded knowledge from historical incidents, making it accessible to all users without requiring them to be subject matter experts.
3Loss of information
If manual searches through incident reports and documentation are performed, then relevant information can be found, but significant time and computing resources are consumed
Solution Approach 1:
The patent applies preliminary action by pre-computing vector representations for all support records, module documentation, and software modules before they are needed for incident resolution. These vectors are stored in advance, allowing the system to quickly retrieve relevant information through simple vector similarity computations rather than performing complex text analysis during incident handling. This pre-processing step significantly reduces the time required to search and retrieve relevant information during actual incident responses.
Solution Approach 2:
The patent replaces time-consuming manual text searches with efficient vector similarity computations. By substituting the mechanical process of reading and comparing text documents with automated vector-based semantic matching, the system achieves both completeness in relevance detection and speed in retrieval. The vector representation allows the system to process and compare large volumes of documentation simultaneously, eliminating the sequential nature of manual searches.
Data Source
AI summary
A method comprises receiving an incident report comprising a textual description of an incident; generating a regularized incident report in which out-of-vocabulary terms in the received incident report are replaced with in-vocabulary terms; determining importance measures for a plurality of incident report terms, wherein each of the incident report terms is in the regularized incident report; generating an incident matrix in which similarity values are defined for combinations of terms in the incident report and terms in a predetermined term set; generating an incident vector based on the incident matrix and the importance measures for the terms in the incident report; applying one or more machine learning (ML) models that identify, based on the incident vector, relevant software support records and/or software modules, wherein the relevant software support records and the software modules are potentially relevant to the incident; and outputting data identifying relevant software support records and/or software modules.