Neural Network Session Model for User Intent Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computing systems that interact with users via text require users to express their queries in a way compatible with the system's textual processing model, limiting user interactions and requiring knowledge of the system's internal workings, and are not effective in understanding user intent over multiple interactions.
Innovation Solution
The use of neural networks trained via reinforcement learning to determine actions during a session, allowing the system to understand user queries and provide relevant information without requiring users to understand the system's internal model, by evaluating state and possible actions separately and adapting to user expressions over multiple interactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If text-based processing models are used to interact with users, then the system can process information through established textual methods, but users must express queries in a way compatible with the system's internal model, limiting ease of operation
Solution Approach 1:
The patent introduces an intermediary layer between the user's natural language input and the system's textual processing model. This intermediary translates user expressions into the system's internal representation without requiring users to understand or conform to that internal model, thereby improving ease of operation while maintaining system complexity management.
Solution Approach 2:
The patent replaces rigid, rule-based textual processing mechanisms with a learned model that can adapt to various user expressions. This substitution allows the system to understand natural language more flexibly without requiring users to follow strict syntactic rules, improving ease of operation.
2Measurement precision
If the system requires users to understand its internal processing model, then processing accuracy can be maintained, but user accessibility and ease of operation deteriorate
Solution Approach 1:
The intermediary translation layer preserves processing accuracy by maintaining the semantic meaning of user queries while transforming them into the system's internal representation. Users continue to express themselves naturally without needing to understand the system's internal model, thus maintaining both accuracy and accessibility.
Solution Approach 2:
The system changes the parameters of interaction by accepting a broader range of user expressions and transforming them into standardized internal representations. This parameter transformation allows the system to maintain processing accuracy while accommodating diverse user input styles, improving accessibility.
3Speed
If the system processes each interaction independently, then processing speed is maintained, but the system cannot understand user intent over multiple interactions
Solution Approach 1:
The system performs preliminary processing of user inputs during interactions, extracting and storing relevant contextual information that can be quickly retrieved in subsequent interactions. This preliminary action enables faster processing while maintaining the ability to understand user intent across multiple interactions.
Solution Approach 2:
The system maintains continuous understanding of user intent by preserving contextual information across interactions. Rather than processing each interaction in complete isolation, the system continuously builds and updates its understanding of user needs, maintaining both speed and comprehension.
4Ease of manufacture
If the system uses traditional text processing methods, then implementation simplicity is maintained, but the system cannot effectively adapt to user expressions over time
Solution Approach 1:
The system transitions from static, rule-based text processing to a dynamic model that can adapt to user expressions over time. This dynamic approach allows the system to learn and adjust to various user input styles while maintaining a relatively simple implementation through the use of trained models rather than complex rule systems.
Solution Approach 2:
The system changes its processing parameters based on learned patterns from user interactions. By adjusting its internal parameters through training on user data, the system becomes more adaptable to different expression styles while maintaining implementation simplicity through a unified model architecture.
Data Source
AI summary
A processing unit can determine a first feature value corresponding to a session by operating a first network computational model (NCM) based part on information of the session. The processing unit can determine respective second feature values corresponding to individual actions of a plurality of actions by operating a second NCM. The second NCM can use a common set of parameters in determining the second feature values. The processing unit can determine respective expectation values of some of the actions of the plurality of actions based on the first feature value and the respective second feature values. The processing unit can select a first action of the plurality of actions based on at least one of the expectation values. In some examples, the processing unit can operate an NCM to determine expectation values based on information of a session and information of respective actions.