Here’s What Happened at ATA’s Expert-in-the-Loop and AI Conference
Speakers at the ATA MT and AI virtual conference coincided in a message urging attendees to embrace the changes AI brings to the way translators and interpreters work.
Language Industry (Artificial) Intelligence — Slator Answers
MT
Domain Adaptation On-The-Fly: Boosting In-Context Machine Translation with Coherence
Researchers from John Hopkins University explore the role of domain and local coherence in in-context machine translation revealing improvements in quality.
Why Large Language Models Hallucinate When Machine Translating ‘in the Wild’
A group of researchers found that large language models produce hallucinations when machine translating ‘in the wild’ that are different from traditional models.
How Does Large Language Model BLOOM Perform for Machine Translation?
Researchers test BLOOM’s capabilities for producing good quality machine translation. They find that training the large language model makes a big difference for all language pairs.
How Google Wants to Improve Evaluation of Low-Resource Dialect Machine Translation
Google launches a new dataset and benchmark to address the lack of region-awareness in machine translation (MT) systems and support under-resourced dialects.
How Large Language Models Can Get Better at Machine Translation of Rare Words
Meta AI researchers experimented with dictionary data prompting on known LLMs in order to improve MT. Results look promising for rare words and domain transfer.
How Large Language Models Can Improve Machine Translation Quality Evaluation
Google demonstrates the capability of LLMs to create synthetic datasets that can be used to train semantic similarity metrics for evaluating MT quality.
Revisiting Controlled Language for Better Machine Translation Quality
Researchers revisit controlled language as a means of improving MT quality. A new methodology based on stylistic quality is said to eliminate repeated fine tuning.
No Post-Editing Output: Microsoft Releases Massive Machine Translation Test Set
Microsoft’s NTREX-128, the second largest human-translated test set, is another benchmark for the evaluation of massively multilingual machine translation research.
A Realistic Dataset for Evaluating Gender Bias in Machine Translation by Amazon
Amazon releases MT-GenEval, a realistic dataset for evaluating gender bias in machine translation, to better understand how MT models perform on gender translation accuracy.
Working Around Machine Translation’s Need for Large-Scale Training Data
Unlike some fields in machine learning, machine translation still requires large sets of training data. The solution? Creating more data when none (or not enough) exists.
Smartling’s Bryan Murphy on How Translation Creates Growth as Budgets Become Tight
At SlatorCon Remote December 2022, Bryan Murphy, CEO of Smartling, talks about leveraging AI and automation to drive growth and reduce costs in the event of a recession.
Intento CEO Outlines Seven Stages of Machine Translation Adoption
At SlatorCon Remote December 2022, Konstantin Savenkov, CEO at Intento, on machine translation bottlenecks, value drivers, ROI, and the future of the global language market.
How Fast Can You Post-Edit Machine Translation?
All you need to know about how you score against postediting translation speed standards and the factors affecting your performance.
Reader Polls: Should Language Service Providers Diversify Beyond Translation?
LSPs think of diversification, DeepL retains its edge after five years, cultural neutralization of source marketing language, automatic transcription far from general adoption.
Your information will not be shared with third parties. No Spam.