Multi-Round Search System Using Bi-gram and LSTM Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing artificial intelligence systems fail to accurately determine multi-round search properties and purposes in human-computer interactions, leading to suboptimal results in low-context scenarios, as they rely on simple literal and syntactic features without effective language modeling or context understanding.
Innovation Solution
A method and system that acquire and analyze multi-round search conditions using bi-gram-based language models and deep learning models like LSTM, determine multi-round properties and purposes, generate search results by combining conditions, and rank them using generative or discriminative models like GBDT, while pruning results to provide optimal outputs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If simple literal and syntactic features are used to identify multi-round properties, then the system is easy to implement, but the accuracy of determining multi-round property and case deteriorates
Solution Approach 1:
The patent transforms the identification approach from simple literal/syntactic features to a comprehensive parameter set including semantic features, user behavior features, and context features. This parameter transformation enables accurate multi-round property identification by capturing the underlying semantics and user intent rather than relying on surface-level text patterns.
Solution Approach 2:
The patent replaces the mechanical feature-matching system with an intelligent semantic analysis system using natural language processing and machine learning models. This substitution allows the system to understand user intent and context automatically, significantly improving the accuracy of multi-round case determination.
2Device complexity
If no language model of search conditions is built, then the system is simpler, but the ability to determine multi-round search purpose in low context deteriorates
Solution Approach 1:
The patent performs preliminary actions by building comprehensive language models of search conditions before actual search execution. These models pre-process and structure search condition data, enabling the system to accurately determine multi-round search purposes even in low-context scenarios by leveraging pre-established semantic understanding.
Solution Approach 2:
The patent introduces language models as intermediary components between user inputs and search execution. These models act as mediators that translate and structure raw search conditions into meaningful representations, allowing accurate determination of search purposes without requiring complex real-time analysis.
3Measurement precision
If multi-round search conditions are combined and analyzed, then the understanding of user search purpose improves, but the processing time and computational resources increase
Solution Approach 1:
The patent segments the multi-round search analysis into distinct modular components: semantic feature extraction, user behavior analysis, context understanding, and result synthesis. This segmentation allows parallel processing of different aspects and reduces overall processing time while maintaining comprehensive analysis accuracy.
Solution Approach 2:
The patent implements partial action by selectively analyzing only the most relevant search conditions and features for each specific query, rather than processing all possible parameters uniformly. This approach reduces computational overhead while maintaining accurate understanding of user search purposes.
Data Source
AI summary
A searching method and system based on multi-round inputs and a terminal are provided. The method comprises: acquiring search conditions input by a user in multiple searches; determining a multi-round property between at least two searches of the multiple searches; determining a search purpose of one of the search conditions, and determining that the search purpose of the one of the search conditions is a multi-round search purpose; generating search results based on the multi-round search purpose and search conditions input by the user; and ranking the generated search results, and determining and outputting an optimal search result. According to the searching method provided by the present application, a machine can understand a user's purpose under a continuous multi-round interactions by understanding the context, so that the use initiative of the user is improved.