{"id":9839,"date":"2023-12-06T06:00:56","date_gmt":"2023-12-06T14:00:56","guid":{"rendered":"https://www.microsoft.com\/en-us\/translator/blog\/?p=9839"},"modified":"2024-05-03T11:16:58","modified_gmt":"2024-05-03T18:16:58","slug":"azure-ai-custom-translator-neural-dictionary-delivering-higher-terminology-translation-quality","status":"publish","type":"post","link":"https://www.microsoft.com\/en-us\/translator/blog\/2023\/12\/06\/azure-ai-custom-translator-neural-dictionary-delivering-higher-terminology-translation-quality\/","title":{"rendered":"Azure AI Custom Translator Neural Dictionary: Delivering Higher Terminology Translation Quality\u00a0"},"content":{"rendered":"

\"\"<\/p>\n

Today, we are super excited to announce the release of neural dictionary, a significant translation quality improvement to our platform. In this blog post, we will explore the neural dictionary feature.<\/span><\/p>\n

Introduction\u202f<\/span>\u00a0<\/span><\/h3>\n

Neural dictionary is an extension to our <\/span>dynamic dictionary<\/span><\/a> and <\/span>phrase dictionary<\/span><\/a>\u00a0features in Azure AI Translator. Both allow our users to customize the translation output by providing their own translations for specific terms or phrases. Our previous method used verbatim dictionary, which was an exact find-and-replace operation. Neural dictionary improves translation quality for sentences which may include one or more term translations by letting the machine translation model adjust both the term and the context to produce more fluent translation. At the same time, it preserves the high term translation accuracy.\u202f<\/span>\u00a0<\/span><\/p>\n

The following English-German example demonstrates differences in translation outputs between both methods when a custom terminology translation is requested:<\/span>\u00a0<\/span><\/p>\n\n\n\n\n\n
Input:\u202f<\/span><\/b>\u00a0<\/span><\/td>\nBasic Knowledge of <mstrans:dictionary translation=”regelm\u00e4\u00dfiges Testen”>Periodic Maintenance<\/mstrans:dictionary> \u202f<\/span>\u00a0<\/span><\/td>\n<\/tr>\n
Verbatim dictionary:\u202f<\/span><\/b>\u00a0<\/span><\/td>\nGrundkenntnisse <\/span>der<\/span> regelm\u00e4\u00dfig<\/span>es<\/span> Test<\/span>en<\/span>\u00a0<\/span><\/td>\n<\/tr>\n
Neural dictionary:\u202f<\/span><\/b>\u00a0<\/span><\/td>\nGrundkenntnisse <\/span>des<\/span> regelm\u00e4\u00dfig<\/span>en<\/span> Test<\/span>ens<\/span>\u00a0<\/span><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

Quality improvement<\/span>\u00a0<\/span><\/h3>\n

The chart below illustrates the significant improvements the new feature brings on common publicly available terminology test sets in Automotive (<\/span>https:\/\/aclanthology.org\/2021.eacl-main.271<\/span><\/a>), Health (<\/span>https:\/\/aclanthology.org\/2021.emnlp-main.477<\/span><\/a>) and Covid-19 domains (<\/span>https:\/\/aclanthology.org\/2021.wmt-1.69<\/span><\/a>) using our general translation models.<\/span>\u00a0<\/span><\/p>\n

\"\"<\/p>\n

We also conducted a series of customer evaluations on Custom Translator platform and neural dictionary models. We measured the translation quality gains on customer data between models with and without the neural dictionary extension. Five customers participated, covering German, Spanish, and French in different business domains. <\/span><\/p>\n

The chart below shows the average improvement of COMET<\/a> in the education domain for English-German, English-Spanish, and English-French; for general models on the left, and for customized models on the right. BLUE color bars represent general translation quality without neural dictionary and ORANGE color bars represent translation quality using neural dictionary. These are overall average improvements on the entire test sets. For segments including one or more customer\u2019s dictionary entries (between 19% and 63%), the improvement is as high as +6.3 to +12.9 COMET points.<\/span>\u00a0<\/span><\/p>\n

\"\"<\/span><\/p>\n

\u00a0<\/span>Supported languages\u00a0<\/span>\u00a0<\/span><\/h3>\n