# 199 The State-of-the-Art in Machine Translation with Language Weaver’s Bart Maczynski
Enterprise Machine Translation and Language Weaver
In this section, Bart Machinski discusses enterprise machine translation and the background of Language Weaver.
Bart Machinski's Career Background
- Bart is the VP of machine learning at Language Weaver, a translation tech brand of RWS.
- Bart started his career in machine translation in 2000 and has held various positions at companies like Trados, SDL, and RWS.
- His work has focused on enterprise and government customers, starting with cat tools and moving to machine translation.
Story Behind Language Weaver
- Language Weaver was founded in 2002 by researchers from the University of Southern California.
- Early work on machine translation at Language Weaver was done for the US government through a grant from DARPA.
Evolution of Language Weaver Technology
This section covers the evolution of technology at Language Weaver leading up to its acquisition by RWS.
Technological Advancements
- Transitioned from statistical solutions to neural Revolution and Transformer-based architectures.
- RWS acquired SDL and merged its MTT team with Iconic, bringing back the original brand due to its recognition in the government space.
Here, Bart discusses some client projects he worked on involving language technology.
Client Projects
New Section
In this section, the speaker discusses the importance of Enterprise-grade machine translation and the specific considerations that need to be taken into account for successful implementation.
Importance of Enterprise Machine Translation
- Enterprise machine translation is optimized not just for linguistic outcomes but also for specific business outcomes.
- Considerations for Enterprise MT include data security, scalability, ease of integration, adaptability, user experience, risk mitigation, and licensing models.
- Solutions must meet all requirements consistently to differentiate from consumer-grade MT.
New Section
The discussion revolves around the increased exposure to AI in various industries, including translation, and the potential benefits and risks associated with it.
Increased Exposure to AI
- AI narratives are prevalent across knowledge-based industries.
- Exposure to AI can lead to unlocking budgets and engaging decision-makers.
- Freelance community tends to embrace AI more readily than enterprise customers due to different risk-reward ratios.
New Section
Discussing the process of improving translations through automated services and human intervention.
Potential Results in Translation Process
- The process involves categorizing potential results as good, adequate, or bad.
- Good translations are ignored, while bad and adequate ones are sent to the LLM for enhancement.
- After editing, the sentence is back-propagated to the QE model for further improvement up to three times.
New Section
Exploring the impact of improved productivity on savings and turnaround time in translation processes.
Impact on Productivity and Savings
- RWS has a large network of language specialists internally and externally.
- Even small improvements in productivity can lead to significant savings and reduced turnaround time.
- Conducted tests with major customers focusing on large localization programs to gather empirical data on efficiency.
New Section
Discussing the successful implementation of an evolved approach in translation processes.
Successful Implementation of Evolved Approach
- Positive feedback from companies like Dell interested in trying out Evolve.
- Plans to expand Evolve to support more languages by the end of the year.
- Emphasis on balancing speed, cost, and complexity in implementing Evolve for efficient translation outcomes.
New Section
Addressing the integration of Evolve into existing systems through API access for requesting specific types of translations.
Integration with Existing Systems
- Evolve will be available as an API for requesting different types of translations.
- Users can specify their preferred translation type (e.g., regular language Weaver Mt or evolve Mt) from various tools like Trados.
New Section
Discussion on the left and right of translation, content generation, AI promise, and extractive summarization capabilities.
Left and Right of Translation
- The left side involves content generation, offering support, and expanding into Pro providing unique positions with access to content management technologies.
- ROI for authoring is clear with generative AI promising benefits. Proper tuning is essential to understand customer nuances.
- Content analytics focus on extractive summarization capabilities across languages for reports or articles.
- Utilizing technology like LLM for entity recognition, data cleanup, and universal capabilities without building separate solutions.
New Section
Discussing the impact of orchestrators in the MT landscape and partnerships with companies like Blackbird.
Impact of Orchestrators
- Exploring Zapier-style orchestrators like Blackbird and Phrase Orchestrator impacting both MT and translation management landscapes.
- Partnerships with companies such as Blackbird show promise in simplifying workflows and enhancing innovation pace.
New Section
Evolution in translation workflows towards nonlinear processes and automated solutions.
Evolution in Translation Workflows
- Embracing low-code automation tools like Bety Blocks for nonlinear workflows and iterative translation needs based on feedback.
- Decentralization of translation production within corporations leading to the need for automated solutions alongside traditional localization programs.
New Section
Drivers for success in the evolving landscape including voice support development.
Success Drivers & Voice Support
- Factors driving success include decentralization, internal review needs based on expertise, and evolving workflow structures.