13 of the 28 MT engines support a glossary and 8 support tone of voice through their API, and 14 support neither.
We send a glossary to an MT engine in one of three ways. We upload it in advance, and each request references it. We send the terms found in each segment with the request. Or we put those terms into a free-text instructions field of the API.
Tone of voice is a parameter with values the vendor sets, or a free-text instruction at Translated Lara 2.
Glossary and tone of voice support in the 28 MT engines
| Model | Vendor | Type | Glossary | Glossary languages (of 16) | Tone of voice | Tone values | Tone languages |
|---|---|---|---|---|---|---|---|
| ALL-Lab | ALL-Lab | MT engine | in each request | 16 | API parameter | tone: formal, natural, expressive, precise | |
| DeepL Next Gen | DeepL | LLM-based MT engine | uploaded in advance | 14 | API parameter | formality: more, less, prefer_more, prefer_less | 13 |
| Google Cloud Advanced Translation API | MT engine | uploaded in advance | 16 | not supported | |||
| Google Translation LLM | LLM-based MT engine | uploaded in advance | 15 | not supported | |||
| ModernMT Enterprise Edition | Translated | MT engine | uploaded in advance | 15 | not supported | ||
| Naver Papago NMT Commercial | Naver | MT engine | not supported | API parameter | honorific setting | 2 | |
| Supertext LLM | Supertext | LLM-based MT engine | in each request | 16 | API parameter | politeness: more, less | 26 |
| SYSTRAN PNMT API | SYSTRAN | MT engine | uploaded in advance | 14 | not supported | ||
| Tilde | Tilde | LLM-based MT engine | uploaded in advance | 11 | not supported | ||
| Translated Lara 2 | Translated | LLM-based MT engine | in the instructions field | 16 | in the instructions field | instructions: 'Be formal', 'Be informal' | |
| Widn AI Tower Anthill | TransPerfect | LLM-based MT engine | in each request | 16 | API parameter | tone: formal, informal | |
| Widn AI Tower Sugarloaf | TransPerfect | LLM-based MT engine | in each request | 16 | API parameter | tone: formal, informal | |
| Widn AI Tower Vesuvius | TransPerfect | LLM-based MT engine | in each request | 16 | API parameter | tone: formal, informal | |
| Yandex | Yandex | MT engine | in each request | 13 | not supported | ||
| Alibaba | Alibaba | MT engine | not supported | not supported | |||
| Amazon | Amazon | MT engine | not supported | not supported | |||
| AppTek NMT | AppTek | MT engine | not supported | not supported | |||
| Baidu | Baidu | MT engine | not supported | not supported | |||
| Elia Elhuyar Machine Translation System | Elia | MT engine | not supported | not supported | |||
| iFlyTek MT | iFlyTek | MT engine | not supported | not supported | |||
| Microsoft Translator | Microsoft | MT engine | not supported | not supported | |||
| NiuTrans | NiuTrans | MT engine | not supported | not supported | |||
| Oracle | Oracle | MT engine | not supported | not supported | |||
| Pangeanic | Pangeanic | MT engine | not supported | not supported | |||
| Tarjama | Tarjama | MT engine | not supported | not supported | |||
| TartuNLP | TartuNLP | MT engine | not supported | not supported | |||
| Tencent Cloud TMT | Tencent | MT engine | not supported | not supported | |||
| Youdao | Youdao | MT engine | not supported | not supported |
Units — Glossary languages (of 16): languages; Tone languages: languages
Tone languages: the number of languages in which the model accepts the tone parameter, where it limits them. Glossary languages: of the 16 with a glossary, the languages for which the model accepted it.