A selection of our keynotes, talks, interviews w. Goran S. Milovanović, public debates, workshops, and meetups.
My keynote Exploring the Potential of Generative AI in Digital Analytics, Webit Summer Edition 2023, Sofia, NDK. At the kind invitation of the organizers of the Webit 2023 conference in Sofia, I discussed the potential of modern generative AI systems in digital analytics. I pay special attention to understanding the process of data generation in a user ecosystem that is all supported by generative AI systems directly or through applications, as well as the significance of synthetic data in analytics. My conclusions suggest that the study of a mix of synthetic and human data is inevitable in the future of digital analytics, and that it doesn’t necessarily compromise analytical conclusions as much as is commonly thought today.
A full synopsis of the Keynote is available on LinkedIn: Exploring the Potential of Generative AI in Digital Analytics
The last in a series of extensive, overview talks on data analytics infrastructure and machine learning with Wikidata, held in the fall of 2021 at the Data Science Conference Europe 2021, Belgrade, Crown Plaza. At that time, the Wikidata Analytics system was fully developed and provided Big Data analytics and machine learning on a vast amount of data, primarily through its leading application, the Wikidata Concepts Monitor. After five years with Wikidata (2017 - 2023), leading projects completely from the back-end, through ML and analytics, to Shiny in production independently, I decided it was time for the next step in my professional career, and that the intriguing and challenging Wikidata analytics should be left to the younger generation. This makes that talk very dear to my memory.
On the zeroth day of the ML Conference Palić 2023, held at InspiraHub in Subotica, the DataKolektiv team conducted a one-day, free workshop on Introduction to Machine Learning for a selected group of participants.
The participants were provided with a comprehensive overview of the mathematical foundations of all the discussed models and Python code, enabling them to independently optimize these elementary models used in ML from scratch.
My 2022 Tech Tutorial on Behavioural Decision Theory and Risk Estimation in R, encompassing SSE and MLE procedures to estimate parametric models of Expected Utility and Cumulative Prospect Theory. I have put some effort here to drive attention to the fact that (a) not just any data can be appropriately modelled by (b) just any math model (c) without taking into account the measurement procedures and the structure of the experimental design (that are, btw, formally one and the same thing) - a fact of which behavioural scientists are (or should be) quite aware of, but which is not in turn recognised and respected by practitioners in Data Science and Machine Learning in general. The respective GitHub repo is here: https://github.com/datakolektiv/DSC2022_BDT
Branko, Aleks, and I discussing education in Data Science: the worth of a decent academic background and PhD studies (or not), how to approach your training and education plan, where science stops and engineering begins, and many other important things in one’s path towards entering the market in Data Science, Machine Learning, and AI.
Jan, Martin, and I discussing a question of the distribution of Wikidata user retention time in WikidataCon 2021. A classic churn problem approached from an elaborated perspective, testing the Power Law distributions on large data sets + some XGBoost predictive churn modeling. The code repo might be useful for anyone interested in churn problems in practice and MLE of Power Laws from empirical data.
This workshop offered a hands-on approach to Semantic Web technologies in R by exemplifying how to work with Wikidata and DBpedia in different ways. We start simple (using the Wikidata API, for example) and then slowly progress towards more advanced topics (e.g. your first SPARQL query from R and why it is not as complicated as people think, matching your data set against Wikidata entities in order to enrich it, and similar). I will provide an introduction to Semantic Web on a conceptual level only so that participants will not need a full understanding of the related technical standards (RDF, different serializations, etc) to follow through. Finally, weshow how to process the Wikidata JSON dump from R for those interested to play big games with R and the Semantic Web. We might be playing around with some interactive graph visualizations during the workshop.
My talk on Hipcon Light 2022, Belgrade, Dorćol Platz, 2020: Structured semantic knowledge in Data Science. Structured semantic knowledge, represented by semantic graphs like the Wikidata or DBpedia knowledge bases, represents not only the most advanced foundation for the development of future artificial intelligence systems, but also an excellent support for enriching existing data sets in everyday work. Our topic is the mapping between two worlds: the world of knowledge representation through a semantic graph and the typical tabular data frame representation found in data science and the realm of relational databases. Our narrative will start from a standard tabular representation in the R and Python programming languages, which we will then transform into a semantic graph, and subsequently return to a well-known format through the SPARQL language. Along the way, we will discuss the RDF standard for semantic data and its well-known serializations (RDF/XML, Turtle, JSON-LD, N-triples, N-quads…). Anyone who believed that the Semantic Web was an exotic idea for the privileged (and who needs it?) will be able to relax: it’s much easier and more accessible than you think.
My talk on WikidataCon 2019, Berlin, Urania Berlin, 2019: Wikidata Statistics: What, Where, and How. I am presenting the whole Analytics and ML infrastructure for Wikidata Analytics - a rather rare opportunity to get acquitted with how things in Data Science/ML/AI are done on (really!) Big Data from Data Engineering and ML to running Data Products in production.
Material from my MilanoR Meetup, June 25, 2019, Milano, Microsoft House.
These notebooks are essenaitly educational material:
The notebook A. Wikidata From R summarizes several methods to access Wikidata (and Wikipedia content) from R and provides a very light introduction to SPARQL;
Notebooks 00 - 08 present a workflow where Wikidata is accessed from R to perform several typical tasks in Information Retrieval;
The newsrivr package to interface the NEWSRIVER API from R, developed by Michael John Page, is also introduced in this Meetup.
Code Repo: MilanoR2019
My talk on European R Users Meeting eRum2018, Budapest, Budapest, Akvarium Klub, May 16, 2018: Wikidata Concepts Monitor: R in action across Big Data. Wikidata, one of the most prominent Wikimedia projects nowadays, presents a formal ontology that adheres to the RDF standard. All Wikimedia projects (Wikipedia, Wiktionary, Wikivoyage, Wikimedia commons, etc) can reach out to and make use of Wikidata by relying on the features of MediaWiki and Wikibase. Currently, more than 800 Wikimedia websites - some of which are found among the most dynamic places online at all - maintain a client-side Wikidata usage tracking. The resulting data sets present a challenging Big Data problem that calls for SQL transfers to Hadoop via Apache Sqoop, massive HiveQL ETL operations and data wrangling in R just in order to enable for the tracking and analytics of individual Wikidata item usage statistics across the Wikimedia projects.
The Wikidata Concepts Monitor (WDCM) system was developed in R to orchestrate various technologies (SPARQL to access Wikidata, SQL, Apache Sqoop, Hive, and Spark) in order to provide for Wikidata usage tracking and analytics. The WDCM then takes a step beyond and performs machine learning (LDA, t-SNE) across the matrices that encompass Wikidata items and Wikimedia projects in order to gain insight into the distributive semantics (i.e. topic models, dimensionality reduction for similarity maps) of Wikidata usage. The system encompasses a set of Shiny dashboards where the results of the WDCM statistical machinery are visualized and tools for analytical work provided in order to help the understanding of the utterly complex social and behavioral phenomenon of Wikidata usage.
An interview by Filip Milenković, 2020, on everything from Computational Cognitive Science, Semantic Web, to Wikidata and AI. “Goran Milovanović is one of the most knowledgeable people I know. I invited him to lecture about Data Science potential in the public administration at the policy conference that I organized five years ago. We remained friends and I enjoy talking with him about everything that has ever popped in the back of my head. Interested in early Internet development and the role of RAND in immediate postwar America? No worries, he’ll speak 15 minutes about it in one breath.”
A 2018. blog on Wikidata Analytics for Wikimedia Foundation. Wikidata is the central knowledge repository in the Wikimedia universe that now encompasses more than 300 free, online encyclopedias, and more than 800 projects in total. Why would anyone devoted to the representation of the sum of knowledge—which by its complexity must encompass different perspectives, interpretations, points of view—need a central repository like that? Simply because knowledge demands grounding in both logical and empirical truths, a set of constraints that define the limits of its power and usage.
In the fall of 2022, we finally recorded my course “Introduction to R” in Serbian in collaboration with Startit Center. Branko Kovač and I have been volunteering to teach this course since around 2015. Later on, I tried to align it more with the tidyverse approach to R programming and continued to volunteer to maintain it at Startit Centers in Belgrade and Novi Sad. This is a basic introduction to R programming for absolute beginners which covers the fundamentals of programming, emphasizes functional programming in R, and provides an introduction to tidyverse through ggplot2 and dplyr.
An old friend of mine from the early days of the Serbian Data Science scene, my CDO Ilija Šuša, a mathematician and Data Engineer, and I, were guests at Netokracija’s podcast focused mainly on the organization of data teams in companies and startups.
Branko, Nemanja, and I discussed the issues of the job market in Data Science, machine learning, and AI in 2023, focusing on changes caused by the economic crisis as much as the upcoming generative AI technologies in software engineering. We were hosted by the Inspira Hub in Subotica as part of the Machine Learning Conference Palić 2023. The discussion took place after Aleks Cvetković, Ilija Lazarević, and I held the “Day Zero” of the conference, providing a free Introduction to Machine Learning training in the Python programming language for a selected group of participants.
A recording from one of the first meet-ups in the Belgrade Data Science scene from 2016, Startit Center Belgrade, Savska 5. I spoke about the application of R and RStudio Shiny in the development of a complete Information Retrieval system and the use of the LDA topic model in analysis, prediction, and text classification in the IG Concepts system, which I developed for the international foundation DiploFoundation in 2015.
Another early recording from the Belgrade Data Science scene beginnings in Startit Center Belgrade, Savska 5, 2016. I spoke about the methods of distributional semantics in R, focusing on Latent Dirichlet Allocation.
My column dedicated to AI on the media portal Talas, in Serbian.
Anstasija Uspenski interviewed me for Netokracija on the most recent develepments in Generative AI and the consequences for the job market.
The Belgrade weekly magazine VREME interviewed me on the Facebook–Cambridge Analytica data scandal.
Published by DiploFoundation
In Partnership with:
As a Research Manager and Statistical Analyst for the Internet Governance Forum: Identifying the Impact (ISBN 978-99932-53-21-1) evaluation study, I have orchestrated a methodological and mainly quantitative approach to better understand the effects of this central diplomatic process in the Internet Governance arena.
Book promotion “Thinking, Fast and Slow” by Daniel Kahneman, SKC, Delfi Bookstore, Belgrade, 26. June 2015.
Speakers about the book included Dr. Goran S. Milovanović, Data Scientist, Dr. Dušan Pavlović, professor at the Faculty of Political Sciences in Belgrade, and Ana Imširović Đorđević, the translator.
I couldn’t find a better place to host my past academic work (1993 - 2013, mainly) than ResearchGate. Excluding books, most of the work should be available in full-text. As an academic researcher in Cognitive Science, I was mainly focused on Categorization and Concept Learning, Semantics, Causality, and Decision Theory.
Goran Milovanovic PR Data Kolektiv, Breza 4/7, ČUKARICA-BEOGRAD, 11000 Beograd, Republika Srbija, ID(APR):64498339, TIN:109890695