Adding intelligence to applications using Azure Cognitive Services
Frequently, customers wish to engage in intelligent application development or targeted data science workloads, but they aren’t sure where to get started.
Microsoft has released its Cognitive Services algorithms to help eliminate barriers to entry by offering easy, scenario-specific get-started packages for developers interested in key AI scenarios.
To make these more accessible to customers, Neal Analytics has created an offer for Microsoft Cognitive Services that enables customers to quickly and easily modernize their applications with cognitive services. These unique new offerings can make applications more intelligent, enable new application build-out scenarios, or enrich existing applications through a simple API call utilizing the best data.
Below are common scenarios for each technology and a brief breakdown of Cognitive Services’ use cases.
One of the leading offerings in the Cognitive Services suite is Anomaly Detector. This simple algorithm is designed to detect anomalies in a continuous stream of data.
This time-series-focused algorithm is handy for detecting changes in temperature, vibration, or other steady-state sensors designed to operate within a series of constraints. It’s useful as a fundamental feature in predictive maintenance suites, providing notice when machines begin to operate outside of established constraints. Neal Analytics frequently uses Anomaly Detector in Predictive Maintenance solutions to serve as a basis for more prescriptive and detailed algorithms and their triggers.
Metrics Advisor is useful for a host of different scenarios, but Neal Analytics typically uses it in data validation processes for data warehouses and reporting. In this scenario, configurations for individual metrics allow them to pass through Metrics Advisor as a part of data movement operations. This ability to quickly sanity-check metrics is often augmented by more detailed logical and physical dataset tests, acting alongside Metrics Advisor.
Similarly, this Cognitive Service can rapidly audit expense and other types of data with proper configuration. Configuring Metrics Advisor in this way can help with spotting outliers and trends within the data to prevent fraud.
Personalizer is similarly diverse. As the name suggests, it is useful for personalization applications, where reinforcement learning is required to model decisions and behavior. This reinforcement learning means that this approach can be target-fit for all scenarios involving diverse decisions and reinforcement learning in addition to personalization and customer engagement scenarios.
The reinforcement learning capabilities in Personalizer means that it can support everything from basic scenarios involving content catalogs to basic route automation based on trained inputs. Examples of this include recommendations of auto-entries from pre-selected form lists in ERPs or advertisements for marketing mix decisions.
Microsoft’s autonomous systems reinforcement learning technology provides a great deal of potential and, when combined with Neal’s knowledge of the Bonsai platform, unlocks the ability to create complex, nuanced solutions leveraging reinforcement learning.
Speech / Translation
Speech is a simple, but effective, cognitive service. It imitates human speech, with a very high degree of accuracy. This cognitive service set is useful for making apps that talk to customers and apps that listen to them.
From QSR order-taking to intelligent kiosks and robotics, this Cognitive Service is very useful in creating a new user interface paradigm for applications. Special considerations are occasionally required, and it is often a good idea to pair Speech with a specialized dev kit or device with an appropriate microphone array.
Vision enables apps to see and recognize images to distinguish between basic abstract concepts. This Cognitive Service does everything from identifying animals to tagging cancerous moles by leveraging the training and context provided by tagged images within the service.
This service is not to be confused with Live Video Analytics, another product offering from Microsoft. Although similar, vision is limited to image recognition, while Live Video Analytics (LVA) is a newer technology designed around real-time vision and video processing. Microsoft frequently bundles LVA is with a particular dev kit described here.
Form recognizer automates clerical duties and comes with an app designed to do tagging for the clerical automation desired. This app has virtually endless possibilities and can save massive amounts of time as form processing, reading, data entry, and other essential tasks are automated.
Microsoft has integrated this Cognitive Service with its Power Platform to take advantage of its ease-of-use. Additionally, many ready-to-go accelerators exist, making it possible to leverage this app in many different form factors.
Adding enterprise-scale search capabilities to an organization can increase workers’ efficiency and even surface information to save time and costs.
Many organizations end up paying extra for unnecessarily repeated processes of gathering, curating, or procuring information. Cognitive Search can alleviate these woes out of the box and are extensible to many use cases. Neal has built skills to extend Cognitive Search capabilities for enterprise analytics and redundant information detection. These use-case specific extensions demonstrate the flexibility of Cognitive Search architecture to solve complex analytical problems through Cognitive Services.
Make it real
Neal Analytics has created an engagement offer to make it easier to build business cases and implement these different cognitive services within intelligent applications. The offer enables customers to deploy a proof-of-concept leveraging any of these cognitive services with Neal’s expert implementation team’s help. Often, proving the concept and validating an intelligent application’s market viability is the first step on a more extensive development journey, and Neal Analytics is eager to help.
This article has been updated on March 1, 2022, and was originally published on Sept. 1, 2021.