Feature-Vector Translation Control for Low-Resource Electronic Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electronic devices face resource limitations and security risks in providing a smooth translation function due to the need for multiple encoders and decoders for various languages, and on-device translation without a network or server increases hacking risks.
Innovation Solution
An electronic apparatus transmits and receives feature vectors to and from an external device using a first encoder and decoder, leveraging neural machine translation models to convert user voice into text, and communicates with external devices or servers based on decoder availability, employing compression and decompression to optimize resource use.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If encoders and decoders for all languages are provided in the electronic apparatus, then translation function coverage is improved, but device resource consumption increases
Solution Approach 1:
The translation system is segmented into two parts: the electronic apparatus contains only the encoder for the user's language, while the external device contains decoders for multiple languages. This segmentation allows the apparatus to maintain low resource consumption while still providing comprehensive translation coverage through the external device's decoder library.
Solution Approach 2:
The encoder output (feature vector) acts as an intermediary that bridges the user's language and multiple target languages. The encoder processes the user's input once, and the resulting feature vector can be decoded into any number of target languages by the external device, eliminating the need for the apparatus to store multiple encoders and decoders.
2Reliability
If text is transmitted to external server for translation, then security risks are reduced, but translation smoothness and speed deteriorate
Solution Approach 1:
The decoder functionality is extracted from the electronic apparatus and placed in the external device. This allows the apparatus to transmit only compact feature vectors rather than raw text, reducing network transmission overhead and enabling faster, smoother translation while maintaining security by not exposing the apparatus's full text processing capabilities.
Solution Approach 2:
The system changes the parameter being transmitted from raw text to compressed feature vectors. This parameter transformation reduces the data size and transmission time while maintaining the essential information needed for translation, thereby improving translation smoothness without compromising security.
3Productivity
If on-device translation is implemented, then translation smoothness is improved, but security risks increase due to hacking and privacy exposure
Solution Approach 1:
The decoder is extracted from the electronic apparatus and relocated to the external device. This extraction allows the apparatus to perform only the encoding step locally (maintaining smoothness), while the decoding into various languages occurs externally, reducing the security risks associated with storing multiple language models and sensitive translation data in the apparatus.
4Adaptability or versatility
If multiple encoders and decoders are provided for various languages, then translation versatility is improved, but device complexity increases
Solution Approach 1:
The translation system is divided such that the complex multi-language decoder functionality is segmented out and placed in the external device, while the electronic apparatus retains only the simple encoder for the user's language. This segmentation dramatically reduces device complexity while preserving translation versatility through the external device's comprehensive decoder support.
Data Source
AI summary
An electronic apparatus includes: a microphone; a communication interface including communication circuitry; a memory configured to store a first encoder corresponding to a first language and a first decoder corresponding to the first language; and a processor configured to: based on a user voice in the first language being received through the microphone, acquire text in the first language corresponding to the user voice, acquire a first feature vector by inputting the text in the first language to the first encoder, control the communication interface to transmit the first feature vector to an external device, and based on a second feature vector being received from the external device through the communication interface, acquire text in the first language corresponding to the second feature vector by inputting the second feature vector to the first decoder.