On-Device Intent Detection via Streaming IPC for Low-Latency Apps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current intent detection methods rely on Application Programming Interfaces (API) and servers, consuming significant resources and time, and require cloud storage, which is inefficient for on-device intent recognition.
Innovation Solution
A method and system for on-device intent detection using streaming pipe inter-process communication (IPC) without APIs or servers, employing an AI algorithm to recognize user intents and utilizing the Android Interface Definition Language (AIDL) interface for inter-process communication between service and partner applications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If API and server are used for intent detection, then intent detection can be performed, but resource consumption and execution time increase significantly
Solution Approach 1:
The patent extracts the intent detection functionality from the server environment and implements it directly on the user device through a service application. The AI model is downloaded and executed locally, removing the dependency on remote servers and APIs, thereby eliminating the resource consumption and latency associated with network communication while maintaining intent detection capability.
Solution Approach 2:
The patent introduces a service application as an intermediary component that bridges the gap between the user device and partner applications. This service application manages the AI model execution and facilitates communication between components through inter-process communication (IPC) mechanisms, enabling efficient local intent detection without requiring external server mediation.
2Reliability
If API and server are used for intent detection, then intent detection can be performed, but execution time increases
Solution Approach 1:
The patent extracts the intent detection functionality from the server environment and implements it directly on the user device through a service application. The AI model is downloaded and executed locally, removing the dependency on remote servers and APIs, thereby eliminating the resource consumption and latency associated with network communication while maintaining intent detection capability.
Solution Approach 2:
The patent implements preliminary action by downloading and caching the AI model on the user device before it is needed for intent detection. The model is prepared and stored in advance, allowing immediate local execution without waiting for server responses during actual detection operations, thus reducing execution time.
3Quantity of substance
If cloud storage is used to store detected intents, then data can be stored, but additional resources and complexity are required
Solution Approach 1:
The patent removes the cloud storage dependency by implementing local storage mechanisms on the user device. Detected intents and user data are stored in the device's local storage through the service application, eliminating the need for cloud storage infrastructure and associated system complexity while maintaining data storage capability.
Solution Approach 2:
The service application provides self-service functionality by managing its own data storage and retrieval operations locally on the device. It handles intent detection results, user profiles, and interaction data without requiring external cloud storage services, thereby reducing system complexity and enabling autonomous operation.
4Use of energy by moving object
If on-device intent detection is implemented without API or server, then resource consumption is reduced, but inter-process communication mechanism is needed
Solution Approach 1:
The patent introduces a service application as an intermediary component that bridges the gap between the user device and partner applications. This service application manages the AI model execution and facilitates communication between components through inter-process communication (IPC) mechanisms, enabling efficient local intent detection without requiring external server mediation.
Solution Approach 2:
The service application serves multiple functions: it downloads and manages the AI model, executes intent detection, stores results locally, and communicates with partner applications through IPC. This multi-functional design consolidates what could be separate complex components into a single versatile service, reducing overall system complexity while maintaining resource efficiency.
Data Source
AI summary
The present invention provides a method and a system for an on-device intent detection using streaming pipe inter-process communication. The method (100) and system (300) disclosed herein involves an on-device intent detection (ODI), that enables applications to recognize user interests which are essentially keywords as they type using an AI model or algorithm. These keywords are communicated to a partner application, installed on the user's device without using any Application Programming Interface (API) or server, thereby effectively improving performance. The user's list of keywords is saved in an AI model and sending the keywords to the partner application employing Android Interface Definition Language (AIDL) interface for inter-process communication (IPC).