Semantic Analytics: The accelerator of Artificial Intelligence Digital Markets

Author: 
Main Author(s) Dr. Martin Serrano
 
Data Scientist and Scientific Researcher
IoT and stream processing Unit Head
Sensors Research Lab and Artificial Intelligence and Robotics Systems
Focus Area: 

Selected Topic Artificial Intelligence

Who stands to benefit and how: 

General Audience, experts in Analytics tools and expert on Interpretation about the findings of the data after running an analytics process. Also in the recent hype for Artificial Intelligence (AI), Data experts and specialist on Artificial Intelligence can benefit from this position paper information, fist to situate the current state of the art and send to understand the emergence and use of Semantic Analytics towards enhancing/accelerating Artificial intelligence

Position Paper: 

Data is everything today, everywhere and everybody can produce, process and consume data, from a simple sensor detecting our presence and an actuator providing a message or activating a door to be opened to a more complex system running expert computing like in economy or stocks markets where the data defines the rise or down of the indexes. But what are the elements that make the data so useful and rich? The answer is simple: “Analysis and Interpretation” The phenomenon of everybody producing data continues growing and especially with the growth of Internet of Things technology (IoT), but volume is not all, else to make these data useful. It is necessary to provide meaning and extract information in the data and most of the time transform it into information and then eventually in knowledge enabling intelligence. Internet of Things [1] is expected to reach every aspect of our human activity and similarly cloud computing is now immersed and everybody use it indistinctly if it is for only storage in cloud or larger performance using the cloud it is expected to reach more to virtual activity too increasing exponentially the volume of information, thus it is important to identify clearly the analytics and their tools.

The big diversity on analysis and analytics tools has made the Analytics area so rich in vocabulary that today it is even difficult to differentiate if there is one or multiple type of analytics. Analytics today is a capacity, a feature to understand data and the best way to materialize value (including economic) to the data but what Analytics is about?, what makes analytics so different from one to another domain? and why Analytics methodologies/technology cannot be used equally in every domain? and particular now when Artificial Intelligence (AI) is hype, how analytics will improve the performance of AI systems. To resolve these questions let’s first understand there are analytics tools that are used for different purposes.
The first one are those analytics tools prepared for use data coming from physical devices (also call raw data) like IoT sensors, the second ones are those analytics tools which are web-based (which means that the tool is using online data, commonly refers to analytics interface or analytics API) and the third one those analytics tools for the web which analyse specific structured data in order to improve web user experience. Having those different use cases using processing data the last one is visualising data, the ultimate outcome of the analytics is visualisation and interpretation.

Analytics
The majority of analytics tools are not considered suitable for a single (or several) specific domain(s), which makes difficult to define the best analytic tools, if this does not rely in a particular form of data or a particular domain of application (e.g. stock markets, production, values and offers, etc) . It is not possible that analytics tools seamlessly adapt for multiple domains. For example we can take example of IoT data analytics like Kaa [4], Kaa can be used in a lot of IoT use cases, and for each one Kaa give some examples of use. However, even if the majority of tools are designed for IoT use cases, very of them cannot be used for other than collect and process values in that particular domain. Other analytics tools have been extended to applications and aiming to be generic, the first example like this is AWS-IoT Analytics [2], and even at AWS IoT there are use cases which are: the smart agriculture, the predictive maintenance, the proactive replenishing of supplies and the process efficiency scoring, the objective is to specialize the operation and provide meaningful interpretation about the data. Another tool seen before with specific use cases is Thingsboard [5], Thingsboard is designed for four use cases: the smart metering, the smart energy, the smart farming and the fleet tracking (bus tracking in the demo). Taken just as example the IoT domain it can be understood that analytics and its tools are domain dependent and thus the more specialised toold to a particular domain data the better are the analytics results.

Artificial Intelligence
Considering the analytics tools for IoT, let’s keep in mind the same domain of application but now let’s consider those analytics tools (for any types of data) which are web-based (online). It is important to know that some analytics tools are also platforms which work not only online and then they can be used to analyse website traffic like the users’ clicks or the time spent by the user on the site for example. The most used AI tools today for IoT are Microsoft Azure Stream Analytics, AWS IoT Analytics, SAP Analytics Cloud and IBM Watson IoT Platform. Azure [1], can be integrated with two other tools to make better analytics (in real time for example). The AWS tool is a tool created by Amazon, its advantages is the free trial period during 12 months which allows to practice and understand this platform before to pay [2]. When we talk about Artificial Intelligence (AI) we are talking a totally computer-based technique from the collection of data, processing to analysis, interpretation and visualisation, the human intervention is minimised and for this reason AI is many times mixed with machine learning as a technique, but the main difference is that AI main outcome is knowledge generation while machine learning focuses in enhancement of the data and interpretation.

Semantics
Semantics is well known in the domain of web services, everybody has used a search engine or take the benefits from an indexed structured data base, now imagine all this benefits using a distributed world-wide data base, the benefits are exponential and highly evident, what makes powerful this approach is largely the distribution and the infrastructure but most important is the structure of the data, think in the way you can understand a language when you are traveling from one side to other side of the world, different accents, different interpretation and sometimes different meaning but when the structure exist it is much easier to understand and interpret the meaning of a sentence. In the same example, when you have relevant and useful information and the structured data and at the same time you can find an answer to a question without having to ask a human the benefits exponentially increase, thus the meaning of colloquial speech, using semantics is all advantages from observing uncover specific meanings of words used in foreign languages mixed with our own to get more into real meanings.

This position paper triggers the discussion on how analysis and their analytics tools, can be used in a particular data domain i.e. IoT, and how every types of data with requires specialised tools for better results when analysis is required. Also the web-based analytics tools and analytics tools for web are mention as the way to enhance Artificial Intelligence results.

Full paper in PDF: 20180725 ATHMOSPHERE-Cloudspace-AI-PositionPaper-Serrano.pdf