Microsoft Connect(); 2018 : Cognitive Services Container LUIS support & Custom Translation in Translator
At Microsoft Connect(); 2018 event, these cognitive-related news announced;
- Cognitive Services Container offers Language Understanding (LUIS) service for use in your container
- Custom translation available for Translator Text
Language Understanding (LUIS) Container
- Container Services
You can use these Docker environments:
- Local on Windows*/macOS/Linux (*be configured to support Linux containers on Windows)
- Azure Kubernetes Service
- Azure Container Instances
- Kubernetes cluster deployed to Azure Stack
This time I introduce it with Docker in local PC (Windows).
- LUIS subscription from Azure Portal
It's OK to use free F0 tier.
Please get endpoint (access URL) and API key, which is for billing when you use it in container. (No data except count usage shouldn't be send to Microsoft.)
- LUIS App
Language Understanding is instant NLP services (modeling GUI tools and APIS for its management and use).
Once you congfigure LUIS App (NLP model), you can export it locally for container.
Store into local folder where Docker can access.
Configure and use LUIS in container
Copy and download LUIS container to local
Using docker command, pull LUIS container image with "latest" option.
docker pull docker pull mcr.microsoft.com/azure-cognitive-services/luis:latest
Run LUIS container
Run LUIS image with your downloaded LUIS app.
docker run --rm -it -p 5000:5000 --memory 4g --cpus 2 --mount type=bind,src=c:\demo\container\luis,target=/input --mount type=bind,src=C:\demo\container\luis,target=/output mcr.microsoft.com/azure-cognitive-services/luis Eula=accept Billing=https://westus.api.cognitive.microsoft.com/luis/v2.0 ApiKey=xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
This is sample to set parameters as:
- Port 5000, 2 CPU, Memory 4GB
- Mount LUIS App in input folder (c:\demo\container\luis in this case)
- Save log in output folder (c:\demo\container\luis in this case)
- Billing (set your endpoint copied from Azure Potal - West US in this case)
- ApiKey (set your ApiKey copied from Azure Portal)
Try LUIS container
Once docker image started, you can access using browser at http://localhost:5000
See swagger (http://localhost:5000/swagger), and you can try analyzing natural sentence by LUIS container.