Text Processing services on the API Business Hub

In order to consume the Topic Detection API SAP Leonardo Machine Learning Foundation service, you will first need to get the service URI, request and response parameters.

Go to https://api.sap.com/.

Then you will be able to search for the SAP Leonardo Machine Learning - Functional Services, then click on the package found.

Select Topic Detection API.

You can also access the page directly from the following address:

• https://api.sap.com/api/topic_detection_api/resource

As you can notice the API has only one resource (or service): /inference_sync.

Now click on the /inference_sync link to expand the section.

Note: the term inference refers to the application phase (scoring) an existing model (as opposed to the training or inception phase) and sync for synchronous.

As stated in the description, the service accepts either:

• an archive file with a zip/tar extensions containing multiple text files
• a single text file
• a list of text files as input

The service returns a list of detected topic with the associated keywords and scores (confidence).

The supported text file formats is plain text only.

The input file, files or archive file will be sent as a FormData query parameter in the service request.

A series of options are also required for the following parameters:

• numTopics: Total number of topics to be detected
• numTopicsPerDoc: Number of most relevant topics to be listed per document
• numKeywordsPerTopic: Number of keywords to be listed per topic
• numFeatures: Maximum number of keywords to be extracted from documents in total

In the top right corner of the page, click on the Log On button.

You will be prompted for your SAP Cloud Platform credentials.

Scroll down to the Parameters section.

In the options field enter the following:

{"numTopics":3, "numTopicsPerDoc":3, "numKeywordsPerTopic":5}

Now let’s build an archive with at least 3 text files (as numTopics is equal to 3).

If you are missing some inspiration, use the following articles content to create your text files:

• Wikipedia article on Machine Learning
• Wikipedia article on Deep Learning
• Wikipedia article on Cheesecake

Create a zip out of these text files.

Click on the Choose File button next to files parameter and pick your zip file.

Now, scroll down and click on Try out.

In the Response Body, you will get for each file, the associated keywords, topics and score.

The output will be sorted by descending topic score.

For example here, for the cheesecake.txt document, the first list of keywords represent to top topic (with id 1) and it’s score is 1.0059799604699027.

{
  "_id": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
  "processed_time": "Mon, 31 Jul 2017 14:52:20 GMT",
    "request": {
      "files": [
        "Desktop.zip"
      ],
      "options": {
        "numKeywordsPerTopic": 5,
        "numTopics": 3,
        "numTopicsPerDoc": 3
      },
      "tenantName": "imgclassif-tech-user",
      "texts": []
    },
    "status": "DONE",
    "tenantName": "imgclassif-tech-user",
    "topicDetection": [
      {
        "docName": "cheesecake.txt",
        "keywords": [
          [ "cheesecake", "cheese", "cream", "bake", "make" ],
          [ "deep", "learn", "neural", "jump", "network" ],
          [ "learn", "machine", "data", "jump", "algorithm" ]
        ],
        "scores": [
          1.0059799604699027,
          0,
          0
        ],
        "topics": [
          1,
          2,
          0
        ]
      },
      {
        "docName": "deep_learning.txt",
        "keywords": [
          [ "deep", "learn", "neural", "jump", "network" ],
          [ "learn", "machine", "data", "jump", "algorithm" ],
          [ "cheesecake", "cheese", "cream", "bake", "make" ]
        ],
        "scores": [
          1.0683225059034496,
          0.0008786062197741417,
          0
        ],
        "topics": [
          2,
          0,
          1
        ]
      },
      {
        "docName": "machine_learning.txt",
        "keywords": [
		  [ "learn", "machine", "data", "jump", "algorithm" ],
		  [ "cheesecake", "cheese", "cream", "bake", "make" ],
          [ "deep", "learn", "neural", "jump", "network" ]
        ],
        "scores": [
          0.890515209507582,
          2.1184791916652968e-9,
          0
        ],
        "topics": [
          0,
          1,
          2
        ]
      }
    ]
  }

Provide an answer to the question below then click on Validate.

You can also try the following Machine Learning Functional Services consuming text content:

• the Document Feature Extraction API
• the Product Text Classification API

For more information, you can also check the online SAP Leonardo Machine Learning Foundation documentation