This article was written by Shashi Sastry

One of the key goals of Industry 4.0 is business optimization, whether it’s from predictive maintenance, asset optimization, or other capabilities that drive operational efficiencies. Each of these capabilities is driven by data, and their success is dependent on having the right data at the right time fed into the appropriate models and predictive algorithms.

Too often data analysts find that they are working with data that is incomplete or unreliable. They have to use additional techniques to fill in the missing information with predictions. While techniques such as machine learning or data simulations are being promoted as an elixir to bad data, they do not fix the original problem of the bad data source. Additionally, these solutions are often too complex, and cannot be applied to certain use cases. For example, there are no “data fill” techniques that can be applied to camera video streams or patient medical data.

Any data quality management effort should start with collecting data in a trusted environment. This, in turn, implies that the data sources (machines, IoT devices, etc.) and the data collection processes are all trusted.

Trusting Data from IIoT Devices

When a new device, is added to a connected factory system, it’s onboarded by authenticating it, typically by using its IP address and a password. Then it’s authorized to communicate data over the network. The flow of data across the network is typically protected through the use of a secure protocol like TLS or SSL, so there is a secure channel established for ongoing data transfer. The foundation is set for trustworthy data.

By Martin Heller Contributing Editor, InfoWorld | MAY 9, 2019

Machine learning uses algorithms to turn a data set into a model. Which algorithm works best depends on the problem

Machine learning and deep learning have been widely embraced, and even more widely misunderstood. In this article, I’d like to step back and explain both machine learning and deep learning in basic terms, discuss some of the most common machine learning algorithms, and explain how those algorithms relate to the other pieces of the puzzle of creating predictive models from historical data.

What are machine learning algorithms?

Recall that machine learning is a class of methods for automatically creating models from data. Machine learning algorithms are the engines of machine learning, meaning it is the algorithms that turn a data set into a model. Which kind of algorithm works best (supervised, unsupervised, classification, regression, etc.) depends on the kind of problem you’re solving, the computing resources available, and the nature of the data.

How machine learning works

Ordinary programming algorithms tell the computer what to do in a straightforward way. For example, sorting algorithms turn unordered data into data ordered by some criteria, often the numeric or alphabetical order of one or more fields in the data.

Linear regression algorithms fit a straight line, or another function that is linear in its parameters such as a polynomial, to numeric data, typically by performing matrix inversions to minimize the squared error between the line and the data. Squared error is used as the metric because you don’t care whether the regression line is above or below the data points; you only care about the distance between the line and the points.

By Rana Gujral

Human emotions exist on a spectrum. More than 50 individual emotions within the categories of love, surprise, joy, anger, sadness, and fear represent the spectrum of human conversation. Each of these emotions has hundreds of potential signals – some of them through body language, some through words, and others through tone of voice. The ability to recognize these signals and respond indicates someone’s emotional intelligence, which strongly correlates to their potential success in life.

The same is true for machines. As artificial intelligence becomes more ubiquitous in consumer applications, it must develop a greater understanding of human emotion and how to respond. That’s where emotion AI comes in. In many applications, more than an “emotionless robot” is needed to replace or augment a human counterpart. Emotion AI is enabling engineers to program systems that interact with human users to recognize and respond to signals of human emotion in both vocal patterns and body language. The more these systems become agile in identifying and responding to individual patterns, the more nuanced they will be in a consumer-facing role.

How AI recognizes emotion in voice

Until very recently, AI has been able to recognize and respond to only a small percentage of human communication—the quantifiable words we speak. Even then, it is often a very literal conversation partner. Voice assistants that are trained to process simple queries often fail to identify sarcasm, exasperation, or hyperbole, among other human speech habits. Emotion AI addresses this by identifying and understanding emotional prosody and intonation in voice to understand the subtext of the conversation.

CECILE DESJARDINS Le 25/06 à 06:00

Tendance / De nouveaux chatbot spécialisés sont à même de répondre aux demandes financières, mais aussi d'interagir avec les systèmes.

Wanda est d'une efficacité exceptionnelle :

« Wanda, quel est le chiffre d'affaires du mois dernier ? » 

- « Bien sûr, le voici » 

- « Peux-tu me sortir un graphique comparatif et l'envoyer au directeur général ? » 

- « C'est fait. »

Il n'a fallu que quelques instants à Wanda pour afficher les résultats comme les graphiques attendus. Cette remarquable diligence s'explique par le fait que Wanda n'a rien d'humain. Ni « elle » ni « lui », « it » est le « chatbot », ou nouvel assistant virtuel, et vocal, proposé par l'éditeur d'ERP spécialisé dans le tertiaire Unit4.

Intégré à différents outils comme Messenger ou Skype, Wanda parle déjà cinq langues - l'anglais, le français, l'allemand, le norvégien et le suédois. Pourquoi « Wanda ? » ? Parce que c'était le nom de la première personne qui s'est inscrite pour discuter avec cet « agent conversationnel », ou « programme informatique capable de simuler une conversation ». Proposé depuis deux ans par Unit4, Wanda constitue un axe de développement important pour l'éditeur. « Nous voulons rendre notre ERP le moins visible possible, et nous concentrer sur l'usage. A terme, nos clients pourront s'ils le souhaitent s'adresser à leur outil en s'adressant uniquement à Wanda. C'est particulièrement intéressant pour les utilisateurs ponctuels, comme des consultants qui remplissent leur feuille de temps une fois par semaine en vue d'une facturation », précise Denis Hucafol, responsable des ventes de Unit4 pour la France.

Facilités