feed.xml

<?xml version="1.0" encoding="utf-8"?><feed xmlns="http://www.w3.org/2005/Atom" ><generator uri="https://jekyllrb.com/" version="4.2.2">Jekyll</generator><link href="https://discoverylab.ai//feed.xml" rel="self" type="application/atom+xml" /><link href="https://discoverylab.ai//" rel="alternate" type="text/html" /><updated>2025-05-12T13:48:30+02:00</updated><id>https://discoverylab.ai//feed.xml</id><title type="html">Discovery Lab</title><subtitle>Driving Scientific Discovery using Machine Intelligence</subtitle><entry><title type="html">DiscoveryLab members attending and presenting work at NeurIPS2023!</title><link href="https://discoverylab.ai//2023/12/15/The-DiscoveryLab-is-at-NeurIPS2023.html" rel="alternate" type="text/html" title="DiscoveryLab members attending and presenting work at NeurIPS2023!" /><published>2023-12-15T00:00:00+01:00</published><updated>2023-12-15T00:00:00+01:00</updated><id>https://discoverylab.ai//2023/12/15/The%20DiscoveryLab%20is%20at%20NeurIPS2023</id><content type="html" xml:base="https://discoverylab.ai//2023/12/15/The-DiscoveryLab-is-at-NeurIPS2023.html">&lt;p&gt;Dimitrios Alivanistos, Daniel Daza, and Michael Cochez of the Discovery Lab attended NeurIPS2023 in New Orleans!&lt;/p&gt;

&lt;p&gt;Excellent opportunity for inspiration, networking and promoting the work done at the Lab.&lt;/p&gt;

&lt;!--more--&gt;

&lt;p&gt;&lt;img src=&quot;/img/dl_at_neurips.jpg&quot; alt=&quot;image&quot; /&gt;&lt;/p&gt;

&lt;p&gt;&lt;img src=&quot;/img/dl_at_neurips_poster.jpg&quot; alt=&quot;image&quot; /&gt;&lt;/p&gt;

&lt;p&gt;Congratulations to the team!&lt;/p&gt;</content><author><name></name></author><summary type="html">Dimitrios Alivanistos, Daniel Daza, and Michael Cochez of the Discovery Lab attended NeurIPS2023 in New Orleans! Excellent opportunity for inspiration, networking and promoting the work done at the Lab.</summary></entry><entry><title type="html">New journal publication: ‘BioBLP: a modular framework for learning on multimodal biomedical knowledge graphs’</title><link href="https://discoverylab.ai//2023/12/07/BioBLP-gets-published-in-the-Journal-of-Biomedical-Semantics.html" rel="alternate" type="text/html" title="New journal publication: ‘BioBLP: a modular framework for learning on multimodal biomedical knowledge graphs’" /><published>2023-12-07T00:00:00+01:00</published><updated>2023-12-07T00:00:00+01:00</updated><id>https://discoverylab.ai//2023/12/07/BioBLP%20gets%20published%20in%20the%20Journal%20of%20Biomedical%20Semantics</id><content type="html" xml:base="https://discoverylab.ai//2023/12/07/BioBLP-gets-published-in-the-Journal-of-Biomedical-Semantics.html">&lt;p&gt;After almost a year of extensive research, BioBLP is published! We explore the combination of multimodal pretrained attribute encoders with Knowledge Graph Embeddings for biomedical Link Prediction!&lt;/p&gt;

&lt;p&gt;Congrats to Daniel Daza, Dimitrios Alivanistos, Thom Pijnenburg, Payal Mitra, Michael Cochez and Paul Groth!&lt;/p&gt;

&lt;!--more--&gt;

&lt;h4 id=&quot;abstract&quot;&gt;Abstract:&lt;/h4&gt;
&lt;p&gt;BioBLP allows to investigate different ways of incorporating multimodal biomedical data for learning representations in KGs. With a particular implementation, we find that incorporating attribute data does not consistently outperform baselines, but improvements are obtained on a comparatively large subset of entities below a specific node-degree. Our results indicate a potential for improved performance in scientific discovery tasks where understudied areas of the KG would benefit from link prediction methods.&lt;/p&gt;</content><author><name></name></author><summary type="html">After almost a year of extensive research, BioBLP is published! We explore the combination of multimodal pretrained attribute encoders with Knowledge Graph Embeddings for biomedical Link Prediction! Congrats to Daniel Daza, Dimitrios Alivanistos, Thom Pijnenburg, Payal Mitra, Michael Cochez and Paul Groth!</summary></entry><entry><title type="html">DiscoveryLab members hosting the local Amsterdam meet-up for Learning on Graphs 2023!</title><link href="https://discoverylab.ai//2023/11/28/DiscoveryLab-members-organizing-the-local-LoG-conference.html" rel="alternate" type="text/html" title="DiscoveryLab members hosting the local Amsterdam meet-up for Learning on Graphs 2023!" /><published>2023-11-28T00:00:00+01:00</published><updated>2023-11-28T00:00:00+01:00</updated><id>https://discoverylab.ai//2023/11/28/DiscoveryLab%20members%20organizing%20the%20local%20LoG%20conference</id><content type="html" xml:base="https://discoverylab.ai//2023/11/28/DiscoveryLab-members-organizing-the-local-LoG-conference.html">&lt;p&gt;Thom Pijnenburg, Dimitrios Alivanistos, Daniel Daza, and Michael Cochez of the Discovery Lab alongside a team of colleagues from the VU university 
organized the local Amsterdam meet-up of the global Learning on Graphs conference &lt;a href=&quot;https://logams.github.io/&quot;&gt;LoGAMS&lt;/a&gt;!&lt;/p&gt;

&lt;p&gt;It took place in the Elsevier headquarters in Radarweg and attracted researchers working on graphs and machine learning!&lt;/p&gt;

&lt;!--more--&gt;

&lt;p&gt;&lt;img src=&quot;/img/log.jpg&quot; alt=&quot;image&quot; /&gt;&lt;/p&gt;

&lt;p&gt;&lt;img src=&quot;/img/log2.jpg&quot; alt=&quot;image&quot; /&gt;&lt;/p&gt;

&lt;p&gt;Congratulations to the team!&lt;/p&gt;</content><author><name></name></author><summary type="html">Thom Pijnenburg, Dimitrios Alivanistos, Daniel Daza, and Michael Cochez of the Discovery Lab alongside a team of colleagues from the VU university organized the local Amsterdam meet-up of the global Learning on Graphs conference LoGAMS! It took place in the Elsevier headquarters in Radarweg and attracted researchers working on graphs and machine learning!</summary></entry><entry><title type="html">DiscoveryLab academic manager gave a keynote at the DL4LD workshop</title><link href="https://discoverylab.ai//2023/09/13/DiscoveryLab-manager-gives-keynote.html" rel="alternate" type="text/html" title="DiscoveryLab academic manager gave a keynote at the DL4LD workshop" /><published>2023-09-13T00:00:00+02:00</published><updated>2023-09-13T00:00:00+02:00</updated><id>https://discoverylab.ai//2023/09/13/DiscoveryLab%20manager%20gives%20keynote</id><content type="html" xml:base="https://discoverylab.ai//2023/09/13/DiscoveryLab-manager-gives-keynote.html">&lt;p&gt;Michael Cochez, academic lab manager of the discovery lab was invited to give a keynote at the &lt;a href=&quot;http://dl4ld2023.mruni.eu/&quot;&gt;3rd Workshop DL4LD: Addressing Deep Learning, Relation Extraction, and Linguistic Data&lt;/a&gt;&lt;/p&gt;

&lt;!--more--&gt;

&lt;p&gt;In his talk, Michael talked about several core themes of the lab: machine learning, knowledge representation, and scientific corpora.&lt;/p&gt;

&lt;h2 id=&quot;abstract&quot;&gt;abstract&lt;/h2&gt;

&lt;p&gt;In the past decade, deep learning has taken a prominent role in the research landscape and some applications, often related to perception or generation, have become mainstream. Already before that, we have seen the uptake of technologies to represent complex interconnected information in the form of a knowledge graph.  In this talk, we will explore several ways in which this world of deep learning and the world of  rich semantic information interact with language. We will do this with the help of several research challenges, like query answering, incorporating labels into link prediction and inductive predictions, and scaling neural network training.&lt;/p&gt;</content><author><name></name></author><summary type="html">Michael Cochez, academic lab manager of the discovery lab was invited to give a keynote at the 3rd Workshop DL4LD: Addressing Deep Learning, Relation Extraction, and Linguistic Data</summary></entry><entry><title type="html">DiscoveryLab members win the LM-KBC competition!</title><link href="https://discoverylab.ai//2022/08/30/DiscoveryLab-members-win-LM-KBC-2022.html" rel="alternate" type="text/html" title="DiscoveryLab members win the LM-KBC competition!" /><published>2022-08-30T00:00:00+02:00</published><updated>2022-08-30T00:00:00+02:00</updated><id>https://discoverylab.ai//2022/08/30/DiscoveryLab%20members%20win%20LM-KBC-2022</id><content type="html" xml:base="https://discoverylab.ai//2022/08/30/DiscoveryLab-members-win-LM-KBC-2022.html">&lt;p&gt;Dimitrios Alivanistos and Michael Cochez of the Discovery Lab alongside a team of colleagues from the university win the &lt;a href=&quot;https://iswc2022.semanticweb.org/&quot;&gt;ISWC2022&lt;/a&gt; &lt;a href=&quot;https://lm-kbc.github.io/&quot;&gt;LM-KBC&lt;/a&gt; competition!&lt;/p&gt;

&lt;!--more--&gt;

&lt;p&gt;A diverse team of 6 colleagues from different labs: Dimitrios Alivanistos (DiscoveryLab), Selene Baéz Santamaría (CLTL, Hybrid Intelligence), Michael Cochez (DiscoveryLab), Jan-Christoph Kalo (DReaMS Lab), Emile van Krieken (L&amp;amp;R VU), Thiviyan Thanapalasingam (INDE Lab)
collaborated, participated and &lt;strong&gt;&lt;em&gt;won&lt;/em&gt;&lt;/strong&gt; the &lt;em&gt;Language Model for Knowledge Base Completion&lt;/em&gt; (LM-KBC) competition of ISWC2022!&lt;/p&gt;

&lt;p&gt;The team used OPENAI’s GPT-3 model, performed prompt engineering and gained insights on how large language model perform on KBC. Their approach uses few-shot prompting with only 4 examples, hand designed for each relation type.
One interesting finding: Simple triple-based prompts appear to work better than natural language prompts!&lt;/p&gt;

&lt;p&gt;Congratulations Dimitrios, Selene, Michael and the rest of the team for the win!&lt;/p&gt;

&lt;ul&gt;
  &lt;li&gt;The paper: &lt;a href=&quot;https://arxiv.org/abs/2208.11057&quot;&gt;paper&lt;/a&gt;&lt;/li&gt;
  &lt;li&gt;The code: &lt;a href=&quot;https://github.com/HEmile/iswc-challenge&quot;&gt;git&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</content><author><name></name></author><summary type="html">Dimitrios Alivanistos and Michael Cochez of the Discovery Lab alongside a team of colleagues from the university win the ISWC2022 LM-KBC competition!</summary></entry><entry><title type="html">Masoud Mansoury organizes MORS@RecSys2022</title><link href="https://discoverylab.ai//2022/08/29/Masoud-Mansoury-organizes-a-workshop-at-RecSys.html" rel="alternate" type="text/html" title="Masoud Mansoury organizes MORS@RecSys2022" /><published>2022-08-29T00:00:00+02:00</published><updated>2022-08-29T00:00:00+02:00</updated><id>https://discoverylab.ai//2022/08/29/Masoud%20Mansoury%20organizes%20a%20workshop%20at%20RecSys</id><content type="html" xml:base="https://discoverylab.ai//2022/08/29/Masoud-Mansoury-organizes-a-workshop-at-RecSys.html">&lt;p&gt;After a sucessful edition &lt;a href=&quot;https://sites.google.com/view/mors-workshop/home&quot;&gt;last year&lt;/a&gt; Masoud Mansoury, who is a researcher from the discovery lab, will organize the 2nd Workshop on Multi-Objective Recommender Systems at the RecSys conference. Check the homepage of the workshop &lt;a href=&quot;https://sites.google.com/view/mors-2022/home&quot;&gt;MORS@RecSys2022&lt;/a&gt; for more information.&lt;/p&gt;

&lt;!--more--&gt;</content><author><name></name></author><summary type="html">After a sucessful edition last year Masoud Mansoury, who is a researcher from the discovery lab, will organize the 2nd Workshop on Multi-Objective Recommender Systems at the RecSys conference. Check the homepage of the workshop MORS@RecSys2022 for more information.</summary></entry><entry><title type="html">The discovery lab at ICT.Open 2022!</title><link href="https://discoverylab.ai//2022/04/04/Approximate-query-answerring-ICTOpen.html" rel="alternate" type="text/html" title="The discovery lab at ICT.Open 2022!" /><published>2022-04-04T00:00:00+02:00</published><updated>2022-04-04T00:00:00+02:00</updated><id>https://discoverylab.ai//2022/04/04/Approximate-query-answerring-ICTOpen</id><content type="html" xml:base="https://discoverylab.ai//2022/04/04/Approximate-query-answerring-ICTOpen.html">&lt;p&gt;At the &lt;a href=&quot;https://www.ictopen.nl/&quot;&gt;ICT.Open 2022&lt;/a&gt; event, we will be presenting our work on approximate query answerring.
Many interesting quesitons can be formulated as graph queries.
However, in some cases, the graph does not have all the information to answer them.
In this line of work we use machine learning to still find the best answer.&lt;/p&gt;

&lt;iframe width=&quot;560&quot; height=&quot;315&quot; src=&quot;https://www.youtube-nocookie.com/embed/mF94YDg4KPE&quot; title=&quot;YouTube video player&quot; frameborder=&quot;0&quot; allow=&quot;accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture&quot; allowfullscreen=&quot;&quot;&gt;&lt;/iframe&gt;

&lt;p&gt;We will also present a general overview of our work in the lab.&lt;/p&gt;

&lt;iframe style=&quot;display:block;&quot; width=&quot;560&quot; height=&quot;315&quot; src=&quot;https://www.youtube-nocookie.com/embed/qrPDNhqxPqc&quot; title=&quot;YouTube video player&quot; frameborder=&quot;0&quot; allow=&quot;accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture&quot; allowfullscreen=&quot;&quot;&gt;&lt;/iframe&gt;</content><author><name></name></author><summary type="html">At the ICT.Open 2022 event, we will be presenting our work on approximate query answerring. Many interesting quesitons can be formulated as graph queries. However, in some cases, the graph does not have all the information to answer them. In this line of work we use machine learning to still find the best answer. We will also present a general overview of our work in the lab.</summary></entry><entry><title type="html">Paper accepted at ICLR2022!</title><link href="https://discoverylab.ai//2022/02/14/paper-accepted-at-ICLR2022.html" rel="alternate" type="text/html" title="Paper accepted at ICLR2022!" /><published>2022-02-14T00:00:00+01:00</published><updated>2022-02-14T00:00:00+01:00</updated><id>https://discoverylab.ai//2022/02/14/paper-accepted-at-ICLR2022</id><content type="html" xml:base="https://discoverylab.ai//2022/02/14/paper-accepted-at-ICLR2022.html">&lt;p&gt;For the second year straight, the Discovery Lab will be at ICLR! This time members of the lab will present StarQE: Query Answering on hyper-relational KG’s.&lt;/p&gt;

&lt;!--more--&gt;

&lt;p&gt;&lt;strong&gt;&lt;em&gt;StarQE: Query Embedding on Hyper-Relational KG’s&lt;/em&gt;&lt;/strong&gt; is the product of an international collaboration between Dimitrios Alivanistos (VU Amsterdam, DiscoveryLabs), 
Max Berrendorf (LMU Munich), Michael Galkin (Mila Quebec &amp;amp; McGill University) and Michael Cochez (VU Amsterdam, DiscoveryLabs).&lt;/p&gt;

&lt;p&gt;There are many papers in the literature that experiment with graph embeddings for answering queries over knowledge graphs. 
In their paper, members of the Discovery Lab, together with colleagues from Germany and Canada, have extended those algorithms to a richer family of graphs, called “hyper-relational graphs”.
These graphs allow for edges to be labelled with key-value pairs, instead of just a single label. 
This richer representation is gaining popularity as the go-to choice for public KG’s (WikiData) and in daily use in some
of the knowledge graphs constructed and maintained by Elsevier.&lt;/p&gt;

&lt;p&gt;This is the first paper to explore query answering with the use of embeddings for such hyper-relational graphs. 
The authors also introduced a new hyper-relational query dataset - &lt;strong&gt;WD50K-QE&lt;/strong&gt; which they used for their experiments. 
They further analyse the impact of qualifier information in the performance of query answering systems and show a great 
boost when qualifiers are available!&lt;/p&gt;

&lt;ul&gt;
  &lt;li&gt;The preprint: &lt;a href=&quot;https://arxiv.org/abs/2106.08166&quot;&gt;paper&lt;/a&gt;&lt;/li&gt;
  &lt;li&gt;The repo: &lt;a href=&quot;https://github.com/DimitrisAlivas/StarQE&quot;&gt;git&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</content><author><name></name></author><summary type="html">For the second year straight, the Discovery Lab will be at ICLR! This time members of the lab will present StarQE: Query Answering on hyper-relational KG’s.</summary></entry><entry><title type="html">Knowledge Graphs Book published</title><link href="https://discoverylab.ai//2021/11/20/knowledge-graphs-book-published.html" rel="alternate" type="text/html" title="Knowledge Graphs Book published" /><published>2021-11-20T00:00:00+01:00</published><updated>2021-11-20T00:00:00+01:00</updated><id>https://discoverylab.ai//2021/11/20/knowledge-graphs-book-published</id><content type="html" xml:base="https://discoverylab.ai//2021/11/20/knowledge-graphs-book-published.html">&lt;p&gt;In a broad collaboration with top researchers in the field, Michael Cochez, academic manager of the discovery lab worked towards a comprehensive book which covers many aspects of Knowledge Graphs
The book is available from &lt;a href=&quot;https://kgbook.org/&quot;&gt;https://kgbook.org/&lt;/a&gt; and in print from &lt;a href=&quot;https://www.morganclaypoolpublishers.com/catalog_Orig/product_info.php?products_id=1683&quot;&gt;Morgan&amp;amp;Claypool&lt;/a&gt;&lt;/p&gt;

&lt;!--more--&gt;

&lt;p&gt;Knowledge Graphs are a graph based abstraction to data that are broadly used to describe, integrate and extract value.
These Knowledge Graphs are at the heart of the discovery lab.
The book covers many topics:&lt;/p&gt;

&lt;ul&gt;
  &lt;li&gt;Formalisms&lt;/li&gt;
  &lt;li&gt;Schema, Identity, and Context&lt;/li&gt;
  &lt;li&gt;Deductive Knowledge (reasoning)&lt;/li&gt;
  &lt;li&gt;Inductive Knowledge (usage in machine learning)&lt;/li&gt;
  &lt;li&gt;Creation and Enrichment&lt;/li&gt;
  &lt;li&gt;Quality Assessment&lt;/li&gt;
  &lt;li&gt;Refinement&lt;/li&gt;
  &lt;li&gt;Publication&lt;/li&gt;
  &lt;li&gt;Knowledge Graphs in Practice&lt;/li&gt;
&lt;/ul&gt;</content><author><name></name></author><summary type="html">In a broad collaboration with top researchers in the field, Michael Cochez, academic manager of the discovery lab worked towards a comprehensive book which covers many aspects of Knowledge Graphs The book is available from https://kgbook.org/ and in print from Morgan&amp;amp;Claypool</summary></entry><entry><title type="html">Best Task Paper Award at SemEval 2021 – ‘MeasEval - Extracting Counts and Measurements and their Related Contexts’</title><link href="https://discoverylab.ai//2021/08/20/Best-Task-paper-award-SemEval-2021.html" rel="alternate" type="text/html" title="Best Task Paper Award at SemEval 2021 – ‘MeasEval - Extracting Counts and Measurements and their Related Contexts’" /><published>2021-08-20T00:00:00+02:00</published><updated>2021-08-20T00:00:00+02:00</updated><id>https://discoverylab.ai//2021/08/20/Best-Task-paper-award-SemEval-2021</id><content type="html" xml:base="https://discoverylab.ai//2021/08/20/Best-Task-paper-award-SemEval-2021.html">&lt;p&gt;Corey Harper et. al. won the best task paper award at SemEval 2021!
See &lt;a href=&quot;https://semeval.github.io/SemEval2021/awards&quot;&gt;https://semeval.github.io/SemEval2021/awards&lt;/a&gt; for the conference organizer’s writeup on it, and &lt;a href=&quot;https://aclanthology.org/2021.semeval-1.38.pdf&quot;&gt;https://aclanthology.org/2021.semeval-1.38.pdf&lt;/a&gt; for the paper itself.&lt;/p&gt;

&lt;!--more--&gt;

&lt;p&gt;SemEval (SEMantic EVALuations) is a highly-regarded conference where each year a number of tasks are proposed for some kind of semantic evaluation of text, images, or data. The tasks are challenges where different research groups develop competing systems. The task organizers provide training and testing data, as well as an evaluation procedure, for the different systems to use. After the end of the competition period, they write a paper on the task and on the results of the various competitors.
Corey proposed a task on extracting measurements and surrounding context from STM text. For example, in addition to spotting measurements like 273.6 MPa, additional information on the property being measured (e.g. fracture pressure), the entity being measured (concrete with a certain additive), and other special information on conditions (.e.g that the samples were cured at a temperature of -10 degrees C).  We expect this kind of additional information will allow us to merge measurements described in the running text with measurements provided databases or in tables in other articles. We think that will be very important as Elsevier continues to develop it data and analytics strategy.&lt;/p&gt;</content><author><name></name></author><summary type="html">Corey Harper et. al. won the best task paper award at SemEval 2021! See https://semeval.github.io/SemEval2021/awards for the conference organizer’s writeup on it, and https://aclanthology.org/2021.semeval-1.38.pdf for the paper itself.</summary></entry></feed>