Generative Model Query Rewriting for Chat Search Intent
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional search engines struggle to accurately determine the intent of queries in chat-style interfaces, where user queries often omit information and rely on context from previous messages, leading to irrelevant search results.
Innovation Solution
A computer-implemented method using a generative model to process input queries and multi-turn query data, generating a contextually aware query that includes additional details from the chat session history, and using a machine-learned embedding model to determine query embeddings and intent clusters for retrieving relevant search results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a traditional search engine processes queries in isolation without chat history, then the search system is simple and fast, but the search results are not responsive to user intent when queries use pronouns or omit context
Solution Approach 1:
The system performs preliminary processing by generating a contextually aware query that incorporates chat history before executing the search. The query rewriting module enriches the current query with relevant context from previous turns, so that when the search engine processes the query, it already contains sufficient information to determine user intent accurately without requiring complex real-time analysis during search execution
Solution Approach 2:
The patent introduces a query rewriting module as an intermediary between the user's concise query and the search engine. This module acts as a mediator that translates the user's context-dependent query into a contextually aware query by integrating chat history, thereby enabling accurate search results without modifying the search engine itself or requiring it to directly access and process entire chat histories
2Measurement precision
If the system incorporates chat history to determine query intent, then search results become more responsive to user intent, but the processing time and computational resources increase
Solution Approach 1:
The system extracts only the essential contextual information needed to resolve pronouns and understand query intent, rather than processing the entire chat history. The query rewriting module identifies and incorporates specific relevant details from previous turns that are necessary to make the current query contextually complete, thereby reducing unnecessary processing while maintaining accuracy
Solution Approach 2:
The system applies partial action by selectively enriching the query with only the necessary context from chat history rather than incorporating all previous information. This approach provides sufficient context to determine user intent accurately without the overhead of processing and analyzing the complete chat history, thus balancing accuracy with efficiency
3Ease of operation
If the system requires users to provide complete descriptive information in each query, then search results are accurate, but the chat-style conversational interface becomes less natural and requires more user effort
Solution Approach 1:
The system enables the query to serve itself by automatically enriching it with necessary context from the chat history. Instead of requiring users to manually provide complete information in each query, the query rewriting module autonomously identifies and incorporates relevant contextual details, allowing users to input concise natural queries while still achieving accurate search results
Data Source
AI summary
Systems and methods for performing a search based on chat data can include determining search results for user queries by generating multi-turn aware queries with a generative language model to include additional details associated with a determined intent of the queries. The generative language model may receive a user query and multi-turn chat data indicative of a chat session of the user and determine an intent of the user query based on the user query and the multi-turn chat data. The generative language model may generate a multi-turn aware query by rewriting the user query to include details associated with the determined intent of the user query, and the multi-turn aware query may be utilized for search result determination. A generative language model may be leveraged to tune an embedding model that may be used for query intent determination.