Context-Aware Neural Machine Translation Using Prompt Enrichment

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Traditional machine translation systems fail to recognize specific contexts and provide inaccurate, contextually inappropriate translations due to their reliance on provided passages alone, often lacking personalized details.

Innovation Solution

Architectures that incorporate contextual data from various sources, such as sensors and user profiles, to generate context-appropriate translations by using neural machine translation models with discretization and signal-to-instruction mapping to refine prompts.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional MT systems translate solely based on the provided passage, then the translation process is simple and fast, but the translation accuracy and contextual relevance deteriorate

Engineering Contradiction:
Improvetranslation accuracyVSAvoidsystem complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system performs preliminary actions by collecting context data from multiple sources (sensors, user profiles, device information) before the translation process. This context data is prepared and stored in advance, allowing the MT model to access relevant contextual information during translation, thereby improving accuracy without significantly increasing operational complexity

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system transitions from traditional two-dimensional translation (source text to target text) by adding a third dimension of context data. This includes sensor data, user profile information, and device characteristics, creating a multi-dimensional input space that enriches the translation process and improves contextual relevance

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Reliability

If context data from multiple sources is incorporated, then translation contextual relevance improves, but system complexity and data processing requirements increase

Engineering Contradiction:
Improvecontextual relevanceVSAvoiddata processing complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The context data collection system is segmented into multiple independent modules, each responsible for a specific data source (sensors, user profiles, device information). This modular architecture allows the system to manage complex data processing through organized, manageable segments rather than a monolithic processing system

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system introduces an intermediary context processing layer between data collection and translation. This intermediary component aggregates, filters, and prepares context data from multiple sources before presenting it to the MT model, simplifying the overall data processing flow while maintaining comprehensive contextual information

Inventive Principle:
Principle #24Intermediary (Mediator)

3Measurement precision

If context data signals are processed through discretization and mapping, then translation precision improves, but processing time and computational resources increase

Engineering Contradiction:
Improvetranslation precisionVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system applies parameter changes by discretizing continuous context data signals into discrete categories and mapping them to standardized instruction formats. This transformation converts raw sensor data and user profile information into structured parameters that the MT model can efficiently process, improving precision while optimizing computational efficiency

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4685688A1Machine translation systems utilizing context data
Publication Date: 2026.01.28 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP4685688A1 patent drawingFigure 1
  • EP4685688A1 patent drawingFigure 2
  • EP4685688A1 patent drawingFigure 3

AI summary

A method for utilizing contextual data in generating machine translations. The method includes receiving a translation request including an initial prompt received via a user interface. The initial prompt includes a first language passage and a translation instruction. The initial prompt also includes a context data signal received via a context data source. The method further includes generating a context instruction based on the context data signal and generating a modified prompt including the initial prompt and the context instruction. The method further includes sending the modified prompt to a neural machine translation (NMT) model to process the modified prompt and receiving a second language translation passage as a response to the modified prompt. The second translation language passage being a second language translation of the first language passage translated according to the translation instruction and the context instruction.