Yanolja Releases AI Translation Model EEVE ROSETTA

SEOUL, South Korea — Softbank Vision Fund-backed Yanolja announced the release of its artificial intelligence translation model, EEVE ROSETTA, on the Hugging Face platform on Friday.

The model, EEVE ROSETTA, received scores on the WMT24++ benchmark evaluation, according to Yanolja.

The company said the scores were in the upper ranges across BLEU, chrF, and MetricX.

Yanolja referred to WMT24++ as an authoritative benchmark for translation performance.

Yanolja NEXT, the company's AI research unit, developed the model. The model supports translation across 32 languages, according to the company.

Languages include Korean, Japanese, Chinese, English, Tagalog, and Farsi.

Yanolja stated that the model operates with 4-12 billion parameters, compared to general-purpose translation models that use more than 70 billion.

The company said EEVE ROSETTA processes both major languages, such as English and Chinese, and less commonly used languages, including Tagalog and Gujarati.

According to Yanolja, the model combines multilingual datasets with context-based translation algorithms that analyze context, nuance, sentence meaning, and sentiment from source text.

The company said it plans to develop language solutions for business applications based on the model. EEVE ROSETTA has been released as open-source.

Potential users include hotels, travel agencies, tourism platforms, e-commerce providers, content producers, and customer support services, according to Yanolja.

A company representative said the placement on Hugging Face is an acknowledgment of Yanolja’s technical capabilities.

The representative added that Yanolja plans to develop additional language and data models designed to improve access to travel services.