Search Query Vector Construction for Knowledge Base Retrieval
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current search techniques for knowledge base articles in CRM environments rely on human intervention, leading to inefficiencies and irrelevant results due to the inclusion of specific, colloquial, or data-specific terms with high TF-IDF scores, which are not correlated across formal knowledge articles.
Innovation Solution
A method utilizing TF-IDF scoring to select candidate root words, building graphs for knowledge articles, merging these graphs, and creating a search query vector by including only terms with correlated neighbors across both service requests and knowledge articles, thereby excluding irrelevant terms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If keywords are manually selected from service requests to create search queries, then the search can capture colloquial and data-specific terms, but the results include irrelevant terms with high TF-IDF scores that are not correlated across formal knowledge articles
Solution Approach 1:
The patent introduces an intermediary component (the search query generation system) that mediates between service requests and knowledge base articles. This intermediary automatically processes service request text, identifies relevant terms through TF-IDF scoring, and constructs search queries that bridge the informal language of service requests with the formal language of knowledge articles, resolving the contradiction between capturing colloquial terms and maintaining result relevance.
Solution Approach 2:
The system changes the parameter of term selection by using automated TF-IDF scoring instead of manual keyword selection. It dynamically adjusts which terms are included in search queries based on their statistical significance across the corpus, transforming the static manual process into a dynamic parameter-driven process that adapts to the specific characteristics of each service request and the knowledge base.
2Productivity
If automated search query generation is implemented using TF-IDF scoring, then human effort is reduced, but the system may include terms with high TF-IDF scores that are specific to individual correspondence and not correlated across formal articles
Solution Approach 1:
The system implements feedback mechanisms where search results and their relevance are analyzed to refine future term selection. The TF-IDF scoring is continuously optimized based on feedback from actual search outcomes, allowing the system to learn which high-scoring terms actually lead to relevant results and adjust its selection criteria accordingly, thereby maintaining productivity while improving term selection accuracy over time.
Solution Approach 2:
The patent replaces the mechanical manual process of keyword selection with an automated computational system based on TF-IDF mathematics. This substitution eliminates human effort and inconsistency while introducing algorithmic precision, where term selection is driven by objective statistical measures rather than subjective human judgment, thus improving productivity without sacrificing accuracy.
3Adaptability or versatility
If all high-scoring terms from service requests are included in search queries, then comprehensive coverage is achieved, but irrelevant and data-specific terms reduce the quality of search outcomes
Solution Approach 1:
The system applies local quality by treating different terms in service requests differently based on their specific characteristics. Instead of uniformly including all high-scoring terms, it selectively processes each term through TF-IDF analysis, evaluating its correlation with formal knowledge articles and applying different inclusion criteria to different terms. This localized approach ensures comprehensive coverage of relevant terms while filtering out irrelevant data-specific terms.
Solution Approach 2:
The system performs preliminary filtering and analysis of terms before finalizing the search query. By pre-processing service request text to identify and evaluate potential terms using TF-IDF scoring against the knowledge base corpus, it prepares a refined set of candidate terms that are more likely to yield quality results, thus achieving comprehensive coverage without including irrelevant terms in the final search query.
Data Source
AI summary
Methods, devices, and systems are described for creating and implementing search query vectors for knowledge base articles or other formal articles, the query vectors automatically created from informal correspondence such as a service request email to an information technology (IT) department. Term frequency-inverse document frequency (TF-IDF) scores are calculated for rarewords in the correspondence with respect to a corpus of other service requests. High scoring terms with the same neighbors as those in the corpus of formal articles are added to the search query vector, while high scoring terms that do not share the same neighbors are thrown out. The query vector is then used to run a search of the knowledge base for relevant articles.