/language-detection-py-microservice

Language Detection, lightweight microservice. Fast and accurate

Primary LanguagePythonBSD 3-Clause "New" or "Revised" LicenseBSD-3-Clause

Language detection as a microservice

Language detection and multi-language detection microservice.

The accuracy increases the fewer languages you configure this microservice to detect. Accuracy ranges from 85% to 100% depending on range of languages to be detected, modifiable in the code.

Language support in this microservice

The following languages are supported for detection, with an estimated accuracy of 95%

    ENGLISH
    CHINESE
    MALAY
    JAPANESE
    KOREAN
    RUSSIAN
    THAI
    VIETNAMESE
    TAGALOG
    HINDI

Deployment

Personally, I've setup the CICD using Google Cloud Build, and deployed as a microservice in Cloud Run. There are no configurations to be made in the repository if you happen to fork it. Simply click a few buttons on GCP Cloud Run and Cloud Build to import this repository, then add the "CLIENT" environment as your secret key

2 POST endpoints

  1. /

     singular language detection with confidence values

returns

    {predicted: <language>, confidence: <confidence>}

expects

    {message: <message>, key: <secret client key>}

  1. /multiple

     multiple language detection, list of language and phrases in the sentence that is detected in that language

returns

    {results: [<language>: <sentence containing the <language> language>]}

expects

    {message: <message>, key: <secret client key>}

Example for multiple

    {"message": "what does this mean 'dua ekor burung'",
    "key": "MYSPECIALKEY"}

returns

    "results": [
        {
            "ENGLISH": "what does this mean "
        },
        {
            "MALAY": "'dua ekor burung'"
        }
    ]
}

1 Get Endpoint

  1. /

    Shows the currently supported languages to be detected
No parameters or headers necessary

Authentication

The POST endpoints requires the caller to be authorized via providing a "key" in the request body. The key needs to match the Client key which is fed via an Environment variable, "CLIENT" in the docker container running this microservice.