News Query Click-Through Prediction Using Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing search engine technologies face challenges in determining the relevance of specialized content, such as news, for queries related to current events, as click-through rates are not always known or reliable, leading to potential display of irrelevant results that annoy users.
Innovation Solution
A system that processes queries to estimate their news-related intent by extracting features from query text and corpora data, using a trained classifier to predict click-through rates and adapt to changing news events, ensuring relevant news results are displayed alongside other search results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If click-through rate tracking is used to determine relevance of specialized content, then relevance measurement is improved, but it fails for news queries with unknown or changing click-through rates
Solution Approach 1:
The system changes the measurement parameter from actual click-through rate to predicted click-through rate. For news queries where historical CTR data is unavailable or outdated, the system uses a prediction model that takes query features and news content features as input to estimate the likelihood of user clicking, thereby adapting the relevance measurement to dynamic news contexts
Solution Approach 2:
The system performs preliminary action by pre-calculating and storing news content features (such as topic categories, time sensitivity, and source credibility) before queries arrive. When a news query comes in, the system quickly retrieves pre-computed features and combines them with query features to predict CTR, avoiding the need for real-time tracking and enabling rapid adaptation to current events
2Quantity of substance
If specialized content is displayed for click-through data collection, then data availability is improved, but irrelevant results annoy users
Solution Approach 1:
The system uses feedback from the prediction model to control content display. The predicted CTR serves as a quality filter - only news content with predicted CTR above a certain threshold is displayed. This feedback mechanism ensures that displayed content is relevant to the query while still providing sufficient data for analysis, preventing user annoyance from irrelevant results
Solution Approach 2:
The system changes the selection criterion from arbitrary or uniform display rules to predicted CTR-based selection. By using the predicted click-through rate as the filtering parameter, the system identifies and displays only those news items that are most likely to be relevant to the specific query, thereby maintaining data availability while eliminating irrelevant content that would annoy users
Data Source
AI summary
Described is estimating whether an online search query is a news-related query, and if so, outputting news-related results in association with other search results returned in response to the query. The query is processed into features, including by accessing corpora that corresponds to relatively current events, e.g., recently crawled from news and blog articles. A corpus of static reference data, such as an online encyclopedia, may be used to help determine whether the query is less likely to be about current events. Features include frequency-related data and context-related data corresponding to frequency and context information maintained in the corpora. Additional features may be obtained by processing text of the query itself, e.g., “query-only” features.