Haven OnDemand Offers Machine Learning As A Service
Written by Nikos Vaggalis
Thursday, 24 March 2016
Hewlett Packard Enterprise's already established IDOL OnDemand platform has been rebranded to signify its new dynamicity in providing a wider variety of API's that expose machine learning as a service.
The underlying notion is that applications, cloud based or otherwise, are not monolithic; they typically rely on the seamless cooperation between a wide variety of components.
In the case of PaaS, these components take the shape of API's. As cloud based applications become cleverer and more complex, working with just one dedicated API isn't enough anymore as the application is expected to cover many and versatile aspects
Haven OnDemand steps into this market by providing a large number (60+) of self contained API's which can be further integrated into workflows where one API's output can become the next one's input, forming a virtual pipeline that simulates more complex logic.
So what do these API's have to offer and how do they pertain to machine learning ?
The Advanced Text Analysis API works on a variety of digital formats, text,audio,video and can amongst others identify the language, extract key concepts as well entities present in the medium, properties that facilitate sentiment analysis and subject classification. One such practical use is in translating text transcripts into live audio streams.
The Format Conversion API can access, extract and convert information from a variety of sources.
The Image Recognition and Face Detection is also able to do barcode recognition and OCR, and can be used for tasks such as identifying the gender of an individual or other key information present in an image.
The Graph Analysis API can make predictions based on relationships and behavioural patterns,something useful in analysing social media data and visualizations.
The Predict and Recommend API can be used to train prediction models that can then be used in building self-learning functions that can analyse huge amounts of data and make accurate predictions.
Two of the most interesting API's are those concerned with search and speech recognition.
The Speech recognition API can extract information from audio and video that can subsequently be fed into the Concept Extraction API, which in turn will return the concept, context and summary of the of things talked about in the audio or video file
This also serves as a prime example of component composition, in the form of the API's, since the Concept Extraction API can be also be called and utilized from other API's like the Text Analysis one.
To start using the Search API you have to create an Index of information or use one from a number of public text Indexes like Wikipedia. Then you can start performing searches on them through the Query Text Index API by building queries using natural language text, keywords, and Boolean expressions. Its potential can be maximized by combining it with the Find Related Concepts,the Find Similar, or the Text Tokenization APIs.
Since the Search facility is expected to be one of the most widely used APIs, it has been also instantiated as a separate entity, taking shape as the Search OnDemand platform, which utilizes HPE's API's to their full extend,satisfying requirements such as searching within different data types or connecting to heterogeneous services like Dropbox or Sharepoint through the Webconnectors APIs.
Voice input is all the rage and it's an interesting new modality. The real question is how are we going to make any money out of it? The obvious answer is to introduce advertising, but this might not [ ... ]
Android Studio 2.3 is out and it is an improvement. Even so its users probably would like it to have a better sense of direction and to be given the impression of a project that knows where it's going [ ... ]