Keyword Rewriting via N-gram Mining and Deep Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Search engines often fail to match user queries effectively due to poorly chosen keyword information, leading to irrelevant results for users and inefficient resource usage, resulting in a poor user experience and lost revenue for advertisers and search engine administrators.
Innovation Solution
A computer-implemented technique that modifies original keyword information by providing supplemental information, extracting frequently occurring n-grams, removing noise, and using a deep-structured machine-learned model to assign scores, thereby supplementing or replacing the original keywords with the highest-scoring n-grams to improve query matching.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If original keyword information is used for matching queries, then the search engine can operate with simple matching logic, but the matching accuracy and relevance of results deteriorate due to poorly chosen keywords
Solution Approach 1:
The system performs preliminary action by generating multiple candidate keywords and n-grams before the actual query matching process. The keyword enrichment module creates candidate keywords from the original keyword and its context, and the n-gram extraction module pre-processes supplemental information to identify relevant n-grams. This preliminary keyword generation and selection process improves matching accuracy without increasing the complexity of the core matching logic.
Solution Approach 2:
The system introduces an intermediary keyword enrichment module that bridges the gap between simple original keywords and complex query matching requirements. This module acts as a mediator by transforming original keywords into enriched keyword sets with multiple candidates and associated n-grams, thereby improving matching accuracy while maintaining manageable system complexity through modular design.
2Productivity
If the search engine uses poorly chosen keyword information, then resource consumption remains low, but user experience deteriorates due to irrelevant results and extended search sessions
Solution Approach 1:
The system applies parameter changes by transforming single original keywords into multiple candidate keywords with varying relevance scores and associated n-grams. The keyword enrichment module generates candidate keywords with different properties (relevance scores, frequency weights, contextual associations), enabling the search engine to select the most relevant keywords for matching. This parameter transformation improves both search efficiency and result relevance.
Solution Approach 2:
The system segments the keyword information into multiple components: original keywords, candidate keywords, and associated n-grams. Each component serves a specific function in the matching process. The segmentation allows the search engine to process different keyword elements separately, selecting and weighting them appropriately, thereby improving search efficiency and result relevance without overwhelming resource consumption.
3Ease of manufacture
If original keyword information is used without enrichment, then the system operates with minimal processing, but advertiser revenue is lost through inefficient ad placement
Solution Approach 1:
The system performs preliminary keyword enrichment and candidate generation before the ad placement process. By pre-processing keywords to create enriched sets with relevance scores and contextual n-grams, the system improves ad matching quality without adding significant implementation complexity. The preliminary action ensures that ad placement decisions are based on enhanced keyword information, reducing revenue loss while maintaining system simplicity.
Data Source
AI summary
A computer-implemented technique is described herein for modifying original keyword information to increase the probability that it will match the queries input by users. The technique operates by using a search engine to provide supplemental information that is relevant to the original keyword information. The technique then mines the supplemental information to extract frequently-occurring n-grams. Next, the technique removes n-grams that are considered to represent noise, and then uses a deep-structured machine-learned model to assign score values to the remaining n-grams. Finally, the technique supplements and/or replaces the original keyword information with the highest-scoring n-grams.