Word Boundary Likelihoods for Incomplete Query Suggestions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Providing accurate query suggestions for long-tail queries is challenging, leading to degraded user experience due to irrelevant search results for incomplete queries, and existing solutions fail to offer timely and relevant search results.
Innovation Solution
The method involves determining word boundary likelihoods by analyzing query logs to identify query sequences, determining word and non-word boundary counts, and providing search results based on these likelihoods, allowing for context-sensitive predictions and delayed result provisioning to enhance relevance and precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If search results are provided for every keystroke in long-tail queries, then user experience is improved by providing timely results, but precision deteriorates due to irrelevant results from incomplete queries
Solution Approach 1:
The system pre-computes word boundary likelihoods for all possible query sequences during an offline phase by analyzing query logs. This preliminary action stores probabilistic information about where words are likely to begin and end, enabling the online system to quickly determine whether to provide results without performing complex analysis in real-time.
Solution Approach 2:
The patent introduces an intermediary mechanism - the word boundary likelihood model - that acts as a filter between user input and search result generation. This intermediary uses pre-computed probabilities to decide whether the current input represents a complete word boundary, thereby mediating when results should be provided to balance timeliness and precision.
2Ease of operation
If query suggestions are provided for long-tail queries, then user experience is improved, but device complexity increases due to the need for sophisticated prediction algorithms
Solution Approach 1:
The patent segments the complex task of query completion into smaller, manageable components: unigram word boundary detection, bigram context analysis, and n-gram sequence matching. By dividing the problem into these discrete segments, the system achieves sophisticated prediction capabilities while maintaining manageable complexity through modular processing.
Solution Approach 2:
The system computes word boundary likelihoods for multiple n-gram lengths (unigrams, bigrams, and potentially longer sequences), using more information than strictly necessary. This excessive computation in the offline phase creates a rich pre-computed model that simplifies online decision-making, trading offline computational effort for online simplicity.
3Speed
If word boundary detection is performed without context, then processing speed is improved, but measurement precision deteriorates due to inability to distinguish homographs and context-dependent words
Solution Approach 1:
The system pre-computes context-aware word boundary likelihoods by analyzing actual query logs during an offline phase. This preliminary action captures contextual patterns from real user queries, storing probabilistic information that reflects how words are actually used in different contexts, thereby eliminating the need for complex runtime context analysis.
Solution Approach 2:
The patent uses feedback from query logs to train and refine the word boundary likelihood model. By continuously analyzing actual user queries and their outcomes, the system learns contextual patterns and adjusts its probability estimates, creating a feedback loop that improves detection accuracy without increasing online processing complexity.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for determining word boundary likelihoods in potentially incomplete text. In one aspect, a method includes selecting query sequences from the query, each query sequence being at least a portion of a word n-gram, the word n-gram being a subsequence of up to n words selected from the second sequence of words of the query, and for each query sequence: determining one or more query sequence keys for the query sequence; determining at least one of a word boundary count and a non-word boundary count for each query sequence key, each word-boundary count and non-word boundary count being dependent on the context of the query sequence; and associating, in a data storage device, the at least one word boundary count and non-word boundary counts with each query sequence key.