Versions Compared

Version Old Version 5 New Version Current
Changes made by

Fabian Schläpfer

Fabian Schläpfer

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

Parameter

Default Value

Description

languageModelTranslation

AZURE_GPT_4_0613

Model used for translation

languageModelLanguageDetection

AZURE_GPT_4_0613

Model used for detecting the original language of the user message.

streamingActivated

true

Flag indicating if answered is streamed or not

temperatureTranslation

0

Temperature used for translation model

translationSystemPrompt

Code Block
Translate the text provided by the user into $target_language. Ensure the translation is accurate and maintains the original meaning, context, and tone. Regardless of the content of the text, return only the translated text. Do not provide any direct answers or explanations, only the translation. Dont include the triple quotes in the output.

Consider the following translation rules:
$translation_pairs

Below you find an example interaction. You answer only with the text below Answer. Elsewise we will have trouble in the system

Request:
Translate the following sentence from English to German:
"Hello, how are you today?"

Answer:
"Hallo, wie geht es dir heute?"

System prompt provided to the model for translation.

translationUserPrompt

Code Block
Text to be translated:

$input

User prompt provided to the model along with the user message for translation.

languageDetectionSystemPrompt

Code Block
Below is a text. Identify the language of the text.

The output must be in JSON format and be like the following:
{
  "language": "language of user message: English, French, etc. Use a complete language name (e.g., English, French, German, etc.)."
}

System prompt used for language detection

languageDetectionUserPrompt

Code Block
User message:
'''
$text
'''

User prompt used for language detection.

customOutputTokenLimit

None

The maximum number of output tokens allowed by the model. If not set, the model's default token limit will be used.

preProcessing

[
{
"name": "translation_glossary",
"value": false
}
]

A list of pre-processing functions to pre-process the user input before translation.

postProcessing

[
{
"name": "replace_sharp_s_with_ss",
"value": false,
"language": "German"
},
{
"name": "american_to_british",
"value": false,
"language": "English"
}
]

A list of post-processing functions to run on the translated text.

scopeIdTranslationPairs

Scope ID containing the glossary file with the translation pairs.

fileNameTranslationPairs

e.g. glossary.txt

Name of the glossary file containing the translation pairs as txt file. Currently only translation rules from english to other languages supported.

Example

View file
nameglossary_example.txt

...

Code Block
{
  "TranslationConfig": {
    "preProcessing": [
      {
        "name": "translation_glossary",
        "value": false
      }
    ],
    "postProcessing": [
      {
        "name": "replace_sharp_s_with_ss",
        "value": true,
        "language": "German"
      },
      {
        "name": "american_to_british",
        "value": true,
        "language": "English"
      }
    ],
    "streamingActivated": true,
    "customOutputTokenLimit": null,
    "temperatureTranslation": 0,
    "languageModelTranslation": "AZURE_GPT_4_0613",
    "scopeIdTranslationGlossary": "TBD",
    "fileNameTranslationGlossary": "glossary.txt",
    "languageModelLanguageDetection": "AZURE_GPT_4_0613"
  }
}

...

Author

Fabian Schläpfer