Search Query Spell Correction via Term Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing search engine spell-check and spell-correct features often result in delays when processing lengthy search queries due to the need to check and correct multiple terms, as they typically only correct the initial terms encountered, potentially missing misspelled terms further in the query.
Innovation Solution
A computer-implemented method that assesses the entire search query to determine language model and character n-gram scores for each term, selects the most likely misspelled terms based on these scores, and provides them to a spell correction system, allowing for contextual correction while minimizing latency by determining a predetermined number of terms to correct and considering surrounding context.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If spell checking is performed on all terms in a search query, then spell correction accuracy is improved, but processing time increases
Solution Approach 1:
The patent segments the search query into individual terms and evaluates each term separately using language model scoring. Terms are divided into different categories (common misspellings, proper nouns, technical terms) and processed differently based on their characteristics and likelihood of being misspelled.
Solution Approach 2:
The patent applies different spell-checking strategies to different terms based on their local characteristics. High-scoring terms (likely misspelled) receive intensive correction attention, while low-scoring terms receive minimal or no correction processing. The correction depth and methodology are adapted locally to each term's needs.
2Reliability
If the number of terms selected for spell correction is increased, then spell correction coverage is improved, but system latency increases
Solution Approach 1:
The patent performs preliminary language model scoring on all query terms before initiating the spell correction process. This preliminary evaluation ranks terms by their likelihood of being misspelled, allowing the system to pre-select which terms will undergo correction and in what order, optimizing the balance between coverage and latency.
Solution Approach 2:
The patent implements partial action by correcting only the most likely misspelled terms rather than all terms. The system applies a threshold-based filtering approach where only terms exceeding a certain confidence score undergo correction, achieving sufficient coverage while limiting processing overhead.
Data Source
AI summary
A computer implemented method and system for spell correcting terms within a string of terms that a computer system receives from a computer readable data string representative of a user search query.