Description
Custom.MT provides a comprehensive platform for machine translation (MT) and generative AI (GenAI) localization, specifically built to empower localization teams. The platform allows users to manage their stock and custom translation models, connecting them seamlessly with various linguistic tools. A key capability is the ability to evaluate and compare different MT and LLM engines, including popular options like DeepL, Google, and GPT-4, alongside custom models. This evaluation is facilitated through the Custom.MT Console's Best Engine tool, which provides metrics such as BLEU and WER scores, example segments, and engine rankings, enabling users to select the optimal model for their specific localization workflow.
The platform offers robust features for managing translation assets and improving translation quality. Tag processing routines can either preserve or remove tags in software strings during AI translation, ensuring compatibility with different systems. Custom.MT also supports model training and fine-tuning to enhance model accuracy with specific terminology and style guides, leading to reduced edit distances and cost savings. For organizations dealing with confidential content, on-premise machine translation deployment is available, ensuring data remains within their own hardware infrastructure.
Further enhancing localization efficiency, Custom.MT includes services for Translation Memory (TMX) cleaning, which is crucial for maintaining the usefulness of translation memories for translators and preparing data for model training. Machine Translation Automatic Post-Editing (MT-APE) is offered to replace manual review with automated post-editing, potentially cutting linguistic audit costs by up to 70%. The platform also assists in data acquisition by searching for and acquiring in-domain training datasets from public sources and marketplaces, covering areas like medical, legal, patent, and eCommerce. For those seeking to build their own MT solutions, Custom.MT supports building custom machine translation models from open-source data and the latest neural network toolkits. Language data anonymization is also provided to automatically remove personal information while retaining language data for training purposes.
Custom.MT is designed for localization professionals, project managers, and enterprises looking to optimize their translation workflows, improve the quality of machine translations, and leverage the power of GenAI for localization. The platform's flexibility, integration capabilities, and advanced features aim to streamline the entire localization process, from model selection and training to final translation output, ultimately driving efficiency and cost-effectiveness.
Custom.MT's Core Features
Manage and connect stock and custom translation models
Evaluate and compare MT and LLM engines (DeepL, Google, GPT-4, custom)
Best Engine tool with BLEU, WER scores, and engine rankings
Tag processing for preserving or removing tags in software strings
GenAI Services: Model Training and Fine-Tuning
Machine Translation Evaluation with automated and human evaluations
On-Premise Machine Translation deployment
Translation Memory (TMX) Cleaning
Machine Translation Automatic Post-Editing (MT-APE)
Data Acquisition for in-domain training datasets
Open-Source Machine Translation model building
Language Data Anonymization
How to use Custom.MT?
Configure: Connect your MT and GenAI models
Evaluate: Use the Best Engine tool to compare models
Train: Fine-tune models with your terminology and style
Translate: Utilize the AI Translation Platform for workflows
Integrate: Connect with your existing translation and content systems
Optimize: Leverage MT-APE and TM cleaning for efficiency
Custom.MT's Use Cases
- MT Engine Evaluation
- Custom Model Training
- On-Premise Translation
- Automated Post-Editing
- Translation Memory Optimization
- Data Acquisition
- Tag Management
