What is the machine learning API service

Cognitive services and machine learning

  • 6 minutes to read

Cognitive Services offers machine learning functions that can be used to solve common problems, such as analyzing text for emotional moods or analyzing images to recognize objects or faces. You don't need any special knowledge of machine learning or data science to use these services.

Cognitive Services is a group of services, each of which supports different generalized prediction functions. The services are divided into different categories that you can use to find the right service.

Service categorypurpose
decisionBuild apps that make recommendations to help you make informed and efficient decisions.
languageLet your apps process natural language with predefined scripts, evaluate sentiments and train them to recognize what users want.
Search,Add Bing search APIs to your apps and take advantage of the ability to search billions of web pages, images, videos and messages with a single API call.
Voice inputConvert speech to text and text to natural sounding speech. Translate from one language to another and turn on speaker verification and recognition.
Image analysisRecognize, identify, subtitle, index, and manage your images, videos, and ink content.

Use Cognitive Services when:

  • You can use a generalized solution.
  • You access a solution from a REST API or an SDK for programming.

Use another machine learning solution if:

  • You have to choose the algorithm and you have to use very specific data to train.

What is machine learning?

Machine learning is a concept where you merge data and an algorithm to solve a specific aspect. Once the algorithm has been trained on that particular data, the output is a model that you can reuse with other data. The trained model provides insights based on the new data.

The process of creating a machine learning system requires some knowledge of machine learning or data science.

Machine learning is delivered through Azure Machine Learning products and services.

What is a cognitive service?

A cognitive service provides some or all of the components of a machine learning solution: data, algorithm, and trained model. A general understanding of your data is required to use these services without any prior experience with machine learning or data science. These services provide both REST API (s) and language-based SDKs. Therefore, you must have some programming language skills to use the services.

In which areas are Cognitive Service and Azure Machine Learning similar?

Both have the ultimate goal of applying Artificial Intelligence (AI) to improve business operations, although they differ in the way in which they provide this in their respective offering.

In principle, the target groups are different:

  • Cognitive services are intended for developers who have no machine learning experience.
  • Azure Machine Learning is tailored for data scientists.

How does a cognitive service differ from machine learning?

A cognitive service provides you with a trained model. This links data and an algorithm that are available via a REST API or an SDK. Depending on your scenario, you can implement this service in minutes. A cognitive service provides answers to common problems, such as key expressions in text or element recognition in images.

Machine learning is a process that typically takes a long time to implement successfully. This time is spent on data collection, cleansing, transformation, algorithm selection, model training, and deployment to achieve the same functionality that is provided by a cognitive service. With machine learning it is possible to provide answers to highly specialized and / or specific problems. These problems to be solved with machine learning require familiarity with the specific topic and data of the problem in question, as well as knowledge of data science.

What kind of data do you have?

Cognitive Services (as a group of services) may require none, some, or all of the custom data for the trained model.

No additional training data required

Services that provide a fully trained model can be downloaded as a Black box be treated. You don't need to know how the services work or what data was used to train them. You feed your data into a fully trained model to get a prediction.

Some or all of the training data required

Some services allow you to bring in your own data and then train a model. This allows you to extend the model using the service's data and algorithm with your own data. The output meets your needs. If you are contributing your own data, you may need to tag the data in a way that is specific to the Service. For example, if you are training a model to recognize flowers, you can provide a catalog of flower images along with the location of the flower in each image to train the model.

A service can provide you with data enableto improve his own data. A service can be from you desirethat you provide data.

Real-time or near-real-time data required

A service may need real-time or near real-time data to build an effective model. Such a service process considerable amounts of model data.

Service requirements for the data model

In the following points, each service is categorized according to the type of data it allows or requires.

* Personalization only requires training data recorded by the service to evaluate your guidelines and data (because the service works in real time). Personalization does not require large datasets with already collected data for pre-training or batch processing training.

Where can you use cognitive services?

The services can be used in any application that can make REST API or SDK calls. Examples of applications are websites, bots, applications for virtual reality or mixed reality, as well as desktop and mobile applications.

How are Azure Cognitive Search and Cognitive Services related?

Azure Cognitive Search is a separate cloud search service that optionally uses Cognitive Services to supplement workload indexing with image and natural language processing. Cognitive Services is exposed in Azure Cognitive Search through built-in capabilities that include individual APIs. You can use a free walkthrough resource, but plan to create and allocate a billable resource for larger volumes.

How Can You Use Cognitive Services?

Each service provides information about your data. You can combine services to chain solutions, e.g. B. Converting speech (audio) to text, translating the text into many languages, and then using the translated languages ​​to get answers from a knowledge base. Although cognitive services can be used to create intelligent solutions based on them, they can also be combined with conventional machine learning projects to supplement models or to accelerate the development process.

The following cognitive services provide exported models for other machine learning tools:

Cognitive serviceModel information
Custom visionExport for TensorFlow for Android, CoreML for iOS11, ONNX for Windows ML

additional Information

Next Steps