0.0 Lecture Overview

This video belongs to the openHPI course Knowledge Graphs - Foundations and Applications. Do you want to see more?

0.0 Lecture Overview

Time effort: approx. 17 minutes

An error occurred while loading the video player, or it takes a long time to initialize. You can try clearing your browser cache. Please try again later and contact the helpdesk if the problem persists.

Scroll to current position

00:00Welcome to Knowledge Graphs.
00:03I am Hartsock and let me introduce you to my team.
00:07We have such a Bruins' Antun, Tubby, Tietz and Masa War Fire
00:13and what I'm going to do now is of course we will introduce you
00:17to the subjects of this lecture. Let's start with week number
00:23One. Week Number One. We are doing Knowledge Representation with graphs.
00:28First of all, we have to make clear what is knowledge and how
00:31does it differ from data. Pure data is raw, it doesn't have any
00:36meaning, and we have to put context, information and stuff everything together
00:40to come up in the end with knowledge, how this works, and to
00:44analyze this in a bit more precise way. We have the very
00:49first part of the lecture from data to knowledge.
00:52Of course knowledge representation is one of the important
00:56issues that we are after, so that will be the subject that follows
01:00directly our first part of the lecture. and when we are talking
01:04about knowledge, of course we also have to talk about how to understand
01:09that. Knowledge understanding is more than just reading.
01:14It's an art form. We have to correctly interpret information
01:19that is brought to us. So this is the first part of the lecture series here in
01:25week number one. and then we continue with an intuitive art or way
01:30of knowledge representation which is simply by graphs on the one hand side
01:35and on triples. Putting things together. then are the so called
01:40Knowledge Graphs. We are explaining what exactly is a knowledge
01:44graph to represent knowledge here based on graph technology
01:49and the vision. where that's this leads to is the so called
01:52Semantic Web. So we will introduce you into the architecture
01:56of the Semantic Web Technology Stick. So this is all of the technology
02:01that we will be talking about in the subsequent weeks. how to
02:04represent knowledge and how to make use of that Later on based on Knowledge Graphs
02:11and one of the simple things, there are specific principles
02:15that are followed and this is first the linked Data principles.
02:18We will learn about what is linked data as a specific form
02:22of the Semantic Web which is closely related also to Knowledge Graphs.
02:26And of course what happens if we take together all that linked
02:30data that is on the web and then we will reach something which is called
02:34the Web of Data. For all of these weeks as well as also in the
02:39first week we have additional hands on created for you that introduce you
02:44really to things, how to do then by yourself for example graph
02:49creation from text or the Art of Understanding with natural
02:52language processing or how to resolve natural language processing ambiguities.
02:57Hi I'm Sasha Bronze and I will give you a short overview of the lecture
03:01to basic Knowledge Graph infrastructure.
03:04In 2.1 we will talk about how to identify, distinguish
03:08and access things on the web. In particular, we'll talk about
03:12apples and why they are not as simple as it seems.
03:15Then we will talk about how to represent simple facts with RDF
03:19triples and what idea of it is and why do we need it.
03:22Then we will move to the total serialization yes, a way to in
03:26court idea of graphs. In 2.4, we will talk about vocabulary, semantics and meaning.
03:33We will talk about RDF schema that is used
03:37to describe facts and things and so on.
03:40In 2.5, we will move to the complex data structures.
03:44We will talk about lists, containers, collections, and how to
03:48work with them, how to
03:51achieve something morphism in exclusion. One, we will talk about
03:55idea of verification and idea of star. Sometimes idea of triples
04:00are just not enough and we will need some kind of some way to
04:04make statements about statements and this is what we will talk about.
04:09In 2.6, we will talk about logical inference with
04:12RDF(S) and in particular how to deduce or bring new knowledge to
04:19knowledge, what is explicitly mentioned
04:22and in exclusion Tool we will talk about idea of one, RDF
04:26and the web, how to connect RDF and HTML.
04:30And of course we will have two practical hands on where we will
04:35show you how to work with RDf and Jupiter notebooks or
04:40Collab notebooks. Yes, we will talk about how to serialize or
04:44visualize our graphs and also how to manipulate the graphs.
04:48How to bring new information, how to delete information,
04:52and how to benefit from it.
04:54Hello, I'm Tobia Tietz and I will give you a brief overview
04:59of lecture number
05:01three. Last week you learned how what knowledge graphs are and how
05:05data is represented. and now of course we have to also learn
05:08how to query this. and this is why the week number three is about
05:12clearing Knowledge Graphs with SPARQL.
05:15In the first lecture, we will learn very basic SPARQL functionalities, how to
05:22perform your first SPARQL queries, what the syntax looks like in SPARQL,
05:26and so on. And then we will take you on a small excursion. We
05:31will take a look at two of the largest knowledge graphs out
05:34there, which are DBpedia and Wikidata.
05:37We will show you how they are created, how to query them,
05:41and then also what kind of interesting knowledge there is to explore for us.
05:46And then we will continue with SPARQL and we will show you
05:50a bit more complex queries,
05:53for example, how to filter the results in interesting ways so that it's
05:58really useful for you and that you can
06:01try out different things. And then we will talk about SPARQL sub-selection
06:07and property Pass, which will also be very handy for you.
06:11And then we will find out in lesson number five that SPARQL is more
06:15than a query language and we can do even more.
06:19And of course, it's also very important that
06:22we talk about quality assurance here, and this is
06:27when we introduce you to the checker constraints.
06:31And as an every week we also have some practical hands-on sessions
06:35prepared for you and that this week we have three call up notebooks for you
06:39and in the first 2 you will learn how to query knowledge
06:43grass was pocket, in the first one with wikidata and then the
06:47second with the DBpedia where we have some really interesting and also fun
06:51queries prepared for you and in the second in the third. Hence on
06:55we will talk about SPARQL Query Federation but we will also introduce some
07:00smaller knowledge graphs to query for example, one about performing arts data.
07:06So stay tuned and I'm looking forward to see you there!
07:10Hi my name is my Sofie and I will give you a brief overview
07:15of what we are going to do in the fourth week of this lecture.
07:19So in the fourth week we will start with a brief history of ontology, from
07:24its definition in philosophy to its definition in computer science. So from
07:29Aristotle to AI, we will together explore the meaning of ontology
07:35and afterwards we will get to the topic of logic.
07:38We expect that you are already familiar with mathematical logics,
07:42but still we will provide a recap of
07:45propositional logic and first order logic in one excursion. But since
07:51these kinds of mathematical logics are not strong enough to
07:54describe ontologies, we will move on to the next excursion where we provide
07:59an introduction to description logics.
08:02Later on we will introduce you to the web ontology language
08:06OWL, and we will show you what possibilities it gives you to create
08:10an ontology and how you can use OWL to provide
08:15definitions for different classes and relations that you want
08:19to have in your ontology.
08:21Eventually we will finish this week up with two hands on in
08:26which you will see how you can create your own ontology using
08:30the unknown version of Proteus or and how you can import this
08:34in the desktop version and what possibilities you have with
08:38potential in creating an ontology.
08:41Hi, my name is Antony. I'm here to give you a brief overview
08:46of Week Five Ontological Engineering for Smarter Knowledge Graphs.
08:51So in the past lecture you've seen or you were introduced to
08:55the web ontology language. Now OWL is very expressive,
09:00but there are limits to OWL and also because of expressivity.
09:04there is also undecidability. So in Excursion seven we will
09:08introduce Semantic Web Rule Language or Swirl, which is a combination of datalog,
09:14a logical rule language and OWL. Swirl is there so that we can
09:20fix the problem of undecidability and also it is computationally efficient for reasoners.
09:27So in Five Point Two, we will introduce to you
09:31a workflow for designing your own ontology.
09:35Here we give you a step by step guide on how to come up with an ontology.
09:40On five Point three, we will then tell you how to petter design your ontologies
09:45and in this case we will talk about ontology evaluation.
09:50In Five Point Four, we will talk about ontological engineering,
09:54particularly ontological alignment and ontological learning.
09:58So learning from text or other information resources.
10:02In Five Point Five, we will tell you how to
10:07construct your knowledge graph. so you have an ontology, how do you feel
10:11your knowledge graph with data? So you can do this using unstructured data
10:15or structured data. And lastly, for this week we will talk about
10:20best practices, particularly in terms of ontology
10:24design, as well as constructing your knowledge graph.
10:28So as with the previous lectures, we also provide practical hands-on.
10:32So we start out with knowledge graph construction from unstructured
10:37text using natural language processing techniques.
10:40and in Five Point Two, we will show you how to construct your
10:44knowledge graph or fill it with data with open refine from a structured
10:50data source. And lastly, we will give you a background or introduction
10:57and also practical hands-on on the Semantic web rule. Language
11:02Week six, there comes the last lecture: Intelligent Applications with Knowledge Graphs
11:08and deep Learning. First of all, what we are going to do is to
11:12look deeper into the graph. In Knowledge Graphs, we all know.
11:16Of course, knowledge graphs are based on graph structures.
11:19However, we have to formally define what exactly is a graph.
11:23and to find out of course, how can we analyze the graph
11:27And how can we make use of the results of the graph analysis for further purposes.
11:33One of the things we have in mind there is also to use for example,
11:37Knowledge Graphs in the terms of Knowledge Graph embeddings.
11:41So this of course is rather similar to distributional semantics in language models.
11:48You might have heard of language models and dear of course
11:51each word or the meaning of words of a text is
11:55expressed by exactly its use within the language. So this is distributional semantics.
12:01However, in the same way, you can also explain
12:05or come up with describing the structure of a Knowledge Graph
12:10and also the properties of a Knowledge Graph
12:13Simply by looking for for example, determining which nodes in
12:17the Knowledge Graph are similar, you are simply looking for
12:20nodes that have a similar environment so that have similar neighbors that are
12:25incased in similar relations with each other for example,
12:30and by using exactly that science same design paradigm, you can create
12:36Knowledge Graph embeddings which are then representation spectral
12:40representations of the Knowledge Graphs usually intense vectors
12:44which on the other hand transport the inherent semantics of
12:49the Knowledge Graph with them. Which means similar nodes are then also
12:54close in closely neighborhood in the vector space as well as
12:58similar relations are in close neighborhood within the vector space.
13:03What use can it be? Of course then you can make Knowledge Graphs
13:07directly accessible for so called graph representation learning.
13:11So you take the graph as input for your machine learning problems
13:17that you want to solve. For example, what we can do is we can solve classification tasks.
13:23However, there is a much more important task. For example, when
13:26we consider Knowledge Graphs we all know that none of these
13:29Knowledge Graphs is really complete and the world of course
13:32it represents. It's also changing dynamically so it's always
13:36clear that there are facts that are missing
13:39and what we can do. Of course based on these embeddings is so
13:42called Knowledge Graph Completion. We could do link prediction
13:47simply to find out stochastically what would be an appropriate link
13:53that is just missing in that knowledge Graph we are looking
13:55for and Difs of course gives way then also to error correction
14:00to completing Knowledge Graphs as well as to fact checking.
14:04So we will see how these things here come together and we will
14:07compare than also Knowledge Graphs to the latest developments
14:11within large language models. We all know that large language models nowadays.
14:16each week or every month a new one is appearing and everybody is in awe
14:20because of the capabilities of this new model and of the understanding and intelligence
14:24it might to to show
14:28the problem there is. As we have all seen, most of these language
14:32models nowadays cannot be fully trusted so they make errors,
14:35they hallucinate and stuff like that
14:37and nearby Knowledge Graphs which x percent which represent explicit
14:43knowledge which can be trusted
14:46is a nice complement for many tasks related with language models like for example
14:51fact checking or explanations
14:56followed by that of course. We then have two tasks at the end
15:00of our lecture. There it's semantic search. So how can we improve
15:04the traditional information retrieval process by including semantic technology?
15:10And moreover then how can we change this traditional retrieval process
15:16and open it up to a exploration of the search space by opening
15:20up the stuff for exploratory search which is closely related
15:25to recommendation and recommender system. So this is the subjects
15:28we are going to talk about in the final week of the lecture.
15:33Complemented is all this again by two hands on. In the first one we will
15:37show you how to do network analysis so we will put in practice
15:41what we have learned here in the very first part in the graph in Knowledge graphs
15:46and in the second hand on we will introduce you to knowledge graph completion
15:50using transit which is a specific knowledge graph embedding model
15:55And with that we come then to an end of the lecture and now
16:00brace yourself or relax. Lean back, enjoy and watch the first lecture.