Use prebuilt Azure Translator in Foundry Tools in Fabric with REST API and SynapseML (preview)

# Get workload endpoints and access token

from synapse.ml.fabric.service_discovery import get_fabric_env_config
from synapse.ml.fabric.token_utils import TokenUtils
import json
import requests


fabric_env_config = get_fabric_env_config().fabric_env_config
auth_header = TokenUtils().get_openai_auth_header()

# Make a RESTful request to Foundry Tool
prebuilt_AI_base_host = fabric_env_config.ml_workload_endpoint + "cognitive/texttranslation/"
print("Workload endpoint for Foundry Tool: \n" + prebuilt_AI_base_host)

service_url = prebuilt_AI_base_host + "language/:analyze-text?api-version=2022-05-01"
print("Service URL: \n" + service_url)

auth_headers = {
    "Authorization" : auth_header
}

def print_response(response):
    print(f"HTTP {response.status_code}")
    if response.status_code == 200:
        try:
            result = response.json()
            print(json.dumps(result, indent=2, ensure_ascii=False))
        except:
            print(f"parse error {response.content}")
    else:
        print(f"error message: {response.content}")

import synapse.ml.core
from synapse.ml.services import *
from pyspark.sql.functions import col, flatten

Text translation is the core operation of the Translator service.

service_url = prebuilt_AI_base_host + "translate?api-version=3.0&to=fr"
post_body = [{'Text':'Hello, friend.'}]

response = requests.post(service_url, json=post_body, headers=auth_headers)

# Output all information of the request process
print_response(response)

Output

    HTTP 200
    [
      {
        "detectedLanguage": {
          "language": "en",
          "score": 1.0
        },
        "translations": [
          {
            "text": "Bonjour cher ami.",
            "to": "fr"
          }
        ]
      }
    ]

Text translation is the core operation of the Translator service.

df = spark.createDataFrame([
  (["Hello, what is your name?", "Bye"],)
], ["text",])

translate = (Translate()
    .setTextCol("text")
    .setToLanguage(["zh-Hans", "fr"])
    .setOutputCol("translation")
    .setConcurrency(5))

result = translate.transform(df)\
        .withColumn("translation", flatten(col("translation.translations")))\
        .withColumn("translation", col("translation.text"))

display(result.select("text", "translation"))

Transliteration converts a word or phrase from the script (alphabet) of one language to another, based on phonetic similarity.

service_url = prebuilt_AI_base_host + "transliterate?api-version=3.0&language=ja&fromScript=Jpan&toScript=Latn"
post_body = [
    {"Text":"こんにちは"},
    {"Text":"さようなら"}
]

response = requests.post(service_url, json=post_body, headers=auth_headers)

# Output all information of the request process
print_response(response)

Output

    HTTP 200
    [
      {
        "text": "Kon'nichiwa​",
        "script": "Latn"
      },
      {
        "text": "sayonara",
        "script": "Latn"
      }
    ]

Transliteration converts a word or phrase from the script (alphabet) of one language to another, based on phonetic similarity.

transliterateDf =  spark.createDataFrame([
  (["こんにちは", "さようなら"],)
], ["text",])

transliterate = (Transliterate()
    .setLanguage("ja")
    .setFromScript("Jpan")
    .setToScript("Latn")
    .setTextCol("text")
    .setOutputCol("result"))

result = transliterate.transform(transliterateDf)\
        .withColumn("script", col("result.script"))

display(result.select("text", "script"))

Returns a list of languages that Translator operations support.

service_url = prebuilt_AI_base_host + "languages?api-version=3.0"

response = requests.get(service_url, headers=auth_headers)

# Output all information of the request process
print_response(response)

No steps for SynapseML in this section.