Query Completion Assistant Using N-gram Language Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional query completion systems rely on memorized frequent queries, which fail to provide effective suggestions for never-before-seen queries, especially in environments like enterprise intranets where historical context is lacking, and may suggest queries that do not retrieve any documents, frustrating users.
Innovation Solution
The system uses a corpus of selected documents to generate query completion suggestions by creating a short text index and employing a language model based on n-grams, ensuring that suggested queries will retrieve at least some search results, thus addressing the limitations of traditional methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If a trie data structure is used to store memorized queries for autocomplete suggestions, then the system can provide fast suggestions (within milliseconds) for known queries, but the system cannot make suggestions for never-before-seen queries or queries with unknown prefixes
Solution Approach 1:
The patent introduces a language model as an intermediary component that bridges the gap between the trie data structure and query completion. The language model processes query prefixes that are not in the trie and generates appropriate completions, allowing the system to handle never-before-seen queries while maintaining the fast lookup capability for known queries through the trie structure.
Solution Approach 2:
The system combines multiple functions into a unified autocomplete mechanism: the trie structure handles exact prefix matching and known queries, while the language model handles novel queries and generates creative completions. This multi-functional approach allows the system to serve both known and unknown query patterns effectively.
2Productivity
If traditional autocomplete systems use memorized frequent queries, then they can provide suggestions for common search terms, but they fail to provide effective suggestions for never-before-seen queries, especially in enterprise intranet environments
Solution Approach 1:
The system uses feedback from the language model to adapt to new query types and contexts. The language model processes the query prefix and generates completions based on the current context, allowing the system to learn and adapt to enterprise-specific terminology and query patterns without requiring pre-programmed knowledge of all possible queries.
Solution Approach 2:
The patent changes the fundamental parameter of how query completions are generated: instead of relying on static memorized queries, the system uses a language model that dynamically generates completions based on the query prefix and contextual information, enabling adaptation to new query types and domains.
3Speed
If autocomplete systems suggest queries based on memorized patterns, then they can complete common queries quickly, but they may suggest queries that do not retrieve any documents, frustrating users
Solution Approach 1:
The patent replaces the mechanical system of memorized query patterns with a language model-based system that uses statistical and contextual understanding to generate completions. This substitution allows the system to maintain speed while improving reliability, as the language model can generate queries that are more likely to retrieve relevant documents based on semantic understanding rather than pattern matching.
Data Source
AI summary
Representative embodiments disclose mechanisms to complete partial queries entered by a user. Users enter a partial query. The partial query is used to search a short text index comprising the titles of documents. The search yields a list results. The top k entries of the list are selected and a language model is created from the top k entries. The language model comprises n-grams from the top k entries and an associated probability for each n-gram. A query completion generator creates query completion suggestions by matching n-grams with the partial query, removing candidate suggestions that to not comply with suggestion rules, and filtering the remaining suggestions according to a filtering criteria. The top N results are returned as suggestions to complete the query.