Information Pushing via Phrase Segmentation and TF-IDF Relevance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing keyword filtering methods in search engine marketing have low accuracy in matching relevant keywords for advertisers, leading to inefficient advertising strategies.
Innovation Solution
A method and apparatus that divide target text into phrases, calculate term frequency-inverse document frequency, and determine similarities between candidate keywords and phrases to select and push relevant keywords based on relevance thresholds, improving the accuracy of information pushing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If literal relevance calculation based on purchased keywords is used, then the matching process is simple, but the accuracy of keyword matching is low
Solution Approach 1:
The patent segments the target text into multiple phrases using different phrase lengths (e.g., unigrams, bigrams, trigrams). This segmentation allows the system to capture keyword relevance at multiple granularities, improving matching accuracy by considering both individual words and multi-word expressions that may better represent user intent.
Solution Approach 2:
The patent introduces a new dimension to keyword matching by calculating relevance based on phrase similarities rather than just literal keyword presence. It computes similarity scores between candidate keywords and extracted phrases, adding a semantic similarity dimension to the traditional exact-match approach, thereby improving accuracy without excessive complexity.
2Reliability
If traditional keyword filtering method is used, then the processing speed is fast, but the relevance of pushed information is low
Solution Approach 1:
The patent performs preliminary phrase extraction and similarity calculation on the target text before the actual keyword matching process. By pre-processing the text to identify relevant phrases and their similarities with candidate keywords, the system prepares relevance scores in advance, ensuring high information relevance while maintaining processing efficiency through optimized pre-computation.
3Measurement precision
If phrase-based similarity calculation is used, then the accuracy of relevance determination is improved, but the computational complexity increases
Solution Approach 1:
The patent applies partial action by selecting a limited set of phrase lengths (e.g., unigrams, bigrams, and optionally trigrams) rather than analyzing all possible phrase combinations. This selective approach captures the most relevant information while avoiding excessive computational resource consumption, achieving a balance between accuracy and efficiency.
Data Source
AI summary
The present disclosure discloses a method and apparatus for pushing information. A specific embodiment of the searching method comprises: dividing words in a target text to generate a set of phrases; calculating a term frequency inverse document frequency for each of the phrases in the set in the target text; acquiring similarities between preset candidate keywords and the each of the phrases; calculating relevance of the candidate keywords to the target text, based on the term frequency inverse document frequency for the each of the phrases in the target text and the similarities between the candidate keywords and the each of the phrases; and selecting a to-be-pushed candidate keyword from the candidate keywords, and pushing the to-be-pushed candidate keyword to a first user terminal based on a preset relevance threshold, or based on the relevance and a preset pushing quantity.