Deep Learning, Data Mining, NLP, Digital Humanities: Overview of Artificial Intelligence
This video tries to give an overview about artificial intelligence and its subfields.
It is an improved version of a video I made and It was liked by many people. It shortly describes fields like data mining, text mining, digital humanities, natural language processing, machine learning, deep learning, evolutionary algorithms and symbolic ai. The aim is to show the differences between those fields and their interaction
Data mining and knowledge discovery is often used to extract additional information from large databases. This area has historically emerged as an extension of the database management systems. The aim is to use statistical and mathematical processes to find information or knowledge that is not explicitly stored in the database.
Text mining is closely related to data mining, with the difference that it uses large collections of text as a data source. These texts can come from the Internet, such as social media and blogs, online magazines, or from other databases such as legal firms, offices, libraries, etc. In addition to the methods of data mining, text mining has its own processes that are specially tailored for text collections. Topic modelling, for example, is a family of algorithms that are used to extract topics from big amounts of text.
The focus of Digital Humanities is on the digital handling of the humanities. Since our cultural goods are increasingly being digitized or digitally generated, the knowledge gained from the humanities is used to develop interdisciplinary methods. Application examples are diverse, from digital archives and museums to digital editions and portals, online dictionaries and tools for capturing humanities data.
Natural Language Processing deals with the digital processing of languages such as English, German, French, Spanish, Arabic or Chinese and many others that are spoken around the globe and used as means of communication. The focus here is on how computers can best understand spoken and written language. Areas of application are diverse: As an example, today we can use tools such as Siri or Amazon Alexa to issue voice commands to our cell phones or computers and thus control them. The spoken language or the texts on the Internet can be simultaneously translated into other languages or from other languages into English.
Machine learning is now at the core of every AI project. Machine learning deals with algorithms that can learn from data. In this context it is also said that these algorithms can be trained. There are problems in life that can only be solved best with the help of machine learning. There are two reasons for this. First, the objective reality often changes very quickly, and second, sometimes a problem is usually far too difficult to grasp. The number of rules or heuristics would explode and the easiest way to do this is for an algorithm to automatically extract them.
As a rule, a distinction is made between supervised and unsupervised learning. Supervised learning means providing additional information from a human processor in addition to the data. For example, before face recognition software is even programmed, the images with faces are separated from all other images and labeled as such. Supervised learning includes procedures such as categorization or classification (for example, the searched face is on a picture or not) and / or prognosis or prediction. The classic algorithms here are Bayesian classifier, which assigns each object to the class to which it belongs with the greatest probability, or decision trees, which create a tree from the decision rules during the training. When classifying, the algorithm runs through this tree and decides on the correct branch or node. But there are also methods that belong to unsupervised learning, such as clustering or component analysis. Here the algorithms try to learn certain regularities based on the data without additional information, to group certain phenomena, events or objects in the data, etc.
Deep learning deals with a special family of algorithms, namely with artificial neural networks. The development of these algorithms was inspired by the functionality of the human brain. They work in a similar way to Bayesian classifier or decision trees, but they differ in their expressive power thanks to the high number of neurons or layers of neurons. That is why the field is also called “deep” or “deep-layered” learning. Another advantage of the artificial neural networks lies in the way how the neurons are activated, also called the activation function in the algorithm. As a rule, a non-linear activation function for neurons is used here. As a result, the networks can represent the complex objects or events from the reality much more precisely. Research in this area now knows many ways how the synapses or, more simply, the connections between neurons can be organized. LSTM (Long Short Term Memory) or convolutional networks can be named as examples for these forms, also known as deep learning architectures.
Teilbereiche der Künstlichen Intelligenz
In diesem Video bespreche ich gängige Teilgebiete der KI anhand eines Posters, welches ich zu diesem Zweck entwickelt habe.
Wozu ist KI heutzutage fähig? Was können die Teilbereiche der KI darüber erzählen?...
Künstliche Intelligenz nimmt immer mehr an Bedeutung zu. Sie ist mittlerweile sehr weit in unser Leben eingedrungen. Man findet sie nicht nur in vielen science fiction filmen, sondern auch in vielen Bereichen unseres Lebens und in vielen Geräten, die wir tagtäglich verwenden. Sie ist in unseren Handys, Kopfhörern oder in den sozialen Medien wie Facebook und Instagram. Ich höre immer wieder Fragen wie: was ist denn die Künstliche Intelligenz?... welche Bestandteile hat sie?... wozu ist sie fähig?... In diesem Video werde ich anhand von einem Poster, welches ich zu diesem Zweck erstellt habe, erklären, was Künstliche Intelligenz ist und aus welchen Teilbereichen sie besteht. Dieses Poster kann man kostenlos aus meiner Homepage auf githup herunterladen.
Natural Language Processing Explained
This is a video I made to explain Natural Language Processing and its subfields
What is the goal of Natural Language Processing and how it effects our daily life
As the name suggests Natural Language Processing is a field of study which deals with the processing or digital application of natural human language. If we put it into simple words, it is mostly about how to make computers to understand or to produce human language. NLP is wildly used in many applications on your computer or on your phone, in the tools like Siri or Amazon Alexa. Even your google assistant can understand your spoken commands. The field has a bright future because the langue is the best and the easiest way to for us humans to communicate. So, what do we mean with natural languages? These are languages like English, French, German, Chinese and many others we use in day-to-day communication. It presents itself mostly in one of the two forms: first in the form of a spoken language which can also be recorded and secondly in the form of written language, like a text in internet or social media. Classically, the field of the study of languages is called linguistics. In fact, it has a tradition of more than two thousand years of research. Usually, linguistics is divided into the study of sounds called phonetics which investigates the smallest building blocks for a language called phonemes, the study of the meaning of words, also called Lexicology, the study of morphology which deals with the inner structure of words or how they change in order to express certain abstract categories like plural in nouns or past in verbs and into a study of syntax how for examples words can be combined in order to build valid sentences.
Foundations of Artificial Neural Networks
This is a video I made to explain the foundation of Artificial Neural Networks
How does an ANN work under the hood mathematically and algorithmically?...
This video gives a very simple and very basic introduction to Artificial Neural Networks. It is meant for people who have no background in Artificial Neural Networks or Deep Leaning, especially for the people from disciplines other than Computer Science like Humanities. The video walks you around the topics how to imagine a neural network as a function, how to calculate an error function in order to find the difference between the predicted value and taget label and how to implement it in python. In the video I also discuss briefly why it is a good idea to vectorise the inputs, because in the real life the many deep learning algorithms use operations from linear algebra in oder not only to make the calculations easier to understand, but also to improve the running time. I will create another videos explaining the perceptron algorithm with examples and the another one explaining how artificial neural networks with deep hidden layers work and how they can be trained with the usage of Gradient Decent. This video will give you already some of the terminology and basic ideas in order to dig to these topics.
Related and Dependent Subfields of AI and Data Science
What fields does AI include?
Machine/Deep Learning, NLP, Symbolic, Digital Humanities and more...
This video tries to give an overview about artificial intelligence and its subfields. It shortly describes fields like data mining, text mining, digital humanities, natural language processing, machine learning, deep learning, evolutionary algorithms and symbolic ai. The aim is to show the differences between those fields and their interaction.
Luna Corpus Tools
The LuNa Open Toolbox for the Luxembourgish Language
How can text analysis and text mining be done for the Luxembourgish languge...
This video is a short description of the open toolbox Luna Corpus Tools. This tool is developed for working with minority languages. It was designed for creating extensive linguistic annotations with the help of graphical user interface. Annotations are created in the process of tokenisation, sentence splitting, normalisation, pos-tagging, morphological analysis, topic modelling and sentiment analysis.
Luna ATM
Automated Text Markup(ATM) with the help of regular expressions)
One of the earlier versions of LuNa for automated text markup with regular expressions
This video was made by a student of mine Ursula Schultze at the University of Trier.
LUNA ist ein Programm, das an der Universität Trier entwickelt wurde und speziell der korpuslinguistischen Arbeit dient. In diesem Tutorial werden grundlegende Funktionen von LUNA und rudimentäre Schritte der korpuslinguistischen Arbeit vorgestellt.
About Me
I earned my master’s degree in Computational Philology from the University of Würzburg (Germany) and my PhD in Digital Humanities from the University of Trier (Germany). Since 2012 I have been involved in different projects at the Universities of Trier and Luxembourg. I am interested in research on appling machine learning in different fields of Humanities. These could be big databases, corpora, portals or language corpora and big language models, especially the ones which were created with the help of the newest deep learning algorithms.
Additional Information
I am about to finish my book about Artificial Intelligence