agenda
14, 15, 16 November 2019
9:00 - 10:00
Registration / Networking
10:00 - 10:20
Main Stage
Welcome
10:20 - 10:40
Main Stage
Using data for social good
Antonio Campello
Data scientist, Wellcome Trust & Datakind
The data-for-good community is all about making an impact, with enthusiastic data scientists using data to
drive innovation, and a vibrant charitable sector willing to embrace data to advance its work. In this
talk, Antonio will share his experience of working with Datakind UK volunteers and social change
organisations. He will discuss how to assess readiness for a data project (data maturity), as well as
several impactful projects that Datakind UK has run over the past years, including engagements with
charities Global Witness (mapping anonymous company owners to uncover corruption) and the Welcome Centre
(identifying people in need of food banks).
10:40 - 11:00
Main Stage
The Opportunity Atlas
Joey Cherdarchuk
Co-founder and Design Lead, Darkhorse Analytics
The Opportunity Atlas visualizes research done at Harvard and Brown Universities tracing
the roots of poverty and incarceration back to the neighbourhoods in which children grew up. Its goal is
to enable policy solutions that will empower families rise out of poverty. In this talk, Joey will discuss
some of the design challenges involved in creating a tool that shares insights found in the 75 million
data points being mapped.
11:00 - 11:20
Main Stage
DFRLab’s approach to raise awareness about disinformation
Anna Pellegatta
Associate Director, Digital Forensic Research Lab
In recent years more and more governments and enterprises have been opening up data they
aggregate. According to the open data index published by the Open Knowledge Foundation, Taiwan is leading
the list followed by the UK and US. Germany's open data rank is currently only 24. As we are entering the
data-driven future, organizations and governments realize that opening up data is less of a choice and
more of a necessity. While opening up data can significantly enhance efficiency, raise trust and add
social value, we are facing a set of new challenges and risks than ever before - starting from the ethical
issues in AI and ending with the absence of data-driven education. This talk will focus on the current
state of open data, touch upon the importance and challenges of data ethics and offer entry points for
institutions for a critical assessment of their data policy as well as guidelines on how to open up data
consciously and consistently.
11:20 - 11:40
Main Stage
Fake Anything: "The Art of Deep Learning"
Eyal Gruss
Independent consultant
The age of creative machines is afoot. Eyal will review the recent state of the art
applications of generative deep learning algorithms in image processing, language modeling, and media
arts. He will also exhibit his digital artwork and perform live demos of the new generation of deep
learning algorithms. He will also present his latest work, "The Art of Deep Learning", which is a
collection of short videos and slides that demonstrate the power of deep learning.
11:40 - 12:00
Short Break
12:00 - 12:20
Main Stage
How to build your own project without big fuckups
Jane Klepa
Executive Director, 1991 Open Data Incubator
What challenges do we have? And how we can solve it with proven methodologies.
12:20 - 13:00
Main Stage
How data impacts decisions and why activists need to know about it
Mihai Avram
Founder, Enable
Did you ever feel that data is interesting, but a bit too complex? If the answer is yes,
then this is the workshop for you. The workshop will present 3 cases in which data was used to improve the
lives of festival-goers, people with asthma and Moldovan cyclists. The lessons extracted from these
examples will prove that working with data can be fun and less complex than it seems at first.
12:00 - 13:00
The EU for Civil Society Activists Room
Introduction to data journalism and an overview of some of the free tools to begin the data journey with
Marek Miller
Teaching Fellow, Google News Lab
This workshop is a brief introduction to data journalism. By looking at the global
examples you will learn what data journalism is in practice and how it is used throughout the world. You
will learn where to look for and how to find the data. You will see how you can nearly automatically
scrape the data from a webpage into Google Sheets. Finally, you will see how ease it is to show the data
by using simple and free tools such as Google GIF Maker, Google MyMaps and Flourish.
12:00 - 13:00
ODIHR Human Rights Room
How to set up a successful micro-tasking project
Milena Marin
Senior Advisor for Tactical Research, Amnesty International
Join this session to learn how to set up a micro-tasking project to engage your audiences
while processing large amounts of unstructured data. Learn how Amnesty International engaged tens of
thousands of digital volunteers to sift through satellite images, documents, pictures and social media
streams. The workshop will explore and document the technology, process, and practical lessons to set up
micro-tasking projects in your organisation.
12:00 - 13:00
ForSet DataViz Room
Convert you "itch" for sharing professional experience into social capital (Part I)
Helen Gerasimova
Head of Data Science Faculty, Netology
How does one transfer their knowledge to someone else? In VUCA world simple apprenticeship
does not work anymore. Now you power a strong educational product with your professional experience and
help your peers to set foot on a new career path. With Helen Gerasimova, Head of Data Science and
analytics faculty at Netology online university and fellow edu-producer Polina Pastovenskaya, you will go
through the essential steps of designing edu-project from both sides – producer and expert.
12:00 - 13:00
BoG Fintech Room
Introduction to Attribution Modeling
Rudolf Eremyan
Data Scientist, Windsor.AI
During the workshop Rudolf will talk about attribution modeling, show examples and how it
helps to optimize marketing campaign budgets. First part of the lecture will be about theoretical side of
the problem.The second part will be dedicated for implementing attribution model using hidden markov
models.
Knowledge of Python is required for this workshop.
12:00 - 13:00
Maxin AI Room
Churn Prediction With Machine Learning; Approaches and Challenges
Davit Abgaryan
Data Scientist, DISQO
Davit will walk you through the process of problem setting and goal defining; data
preprocessing and feature engineering. He will look at models, metrics, hyperparameter tuning and feature
importances; ensemble models multidimensionality, sparsity and parallel processing
This workshop's requirement is at least basic level of Python and Machine Learning.
13:00 - 14:00
Networking Break
14:00 - 14:20
Main Stage
I am telling you 3 things about Chatbot
Cheuk Ting Ho
Data scientist, Santander
Chatbot is cool! Now we can talk to our gadgets as if it’s a real human, right? It also makes you wonder
when you chat with the “customer service” online, is that a real person or a robot on the other side?
Cheuk's boss wanted her to build a chatbot, and here’s 3 things that she discovered.
14:20 - 14:40
Main Stage
Painting by Numbers: The Art of Data Visualisation
Sophie Shawdon
Senior Data Analyst, ClearScore
In this talk, Sophie will talk through the power of good data design, and show how to cut
through the noise to create powerful, impactful visualisations.
14:40 - 15:00
Main Stage
Making email campaigns more effective: Send time optimization
Ildiko Czeller
Lead Data Scientist, Emarsys
At Emarsys their goal is to make marketing communication truly personal. A significant part of this is
sending messages at the right time. They developed a machine learning algorithm resulting in increased
user engagement and shorter times between sending and reading messages. In this talk Ildi will highlight
the most important milestones of this journey: what a successful machine learning project looks like from
the birth of a new idea through implementation to measuring results. She will show how they proved the
significance of sending times with a simulation based on past data to avoid costly experiments, how they
prototyped several algorithms and what they learned from piloting with a few clients before releasing to
all of them.
15:00 - 15:20
Main Stage
Let’s talk emotions: Unlocking the power of financial data
Stefi Gual
Product Manager for Digital Channels at Strands, Strands
Can we use financial data in order to satisfy the growing, and changing, needs of our customers? This talk
will shed light on how banks can utilize data to incorporate financial wellbeing into their core strategy.
Emotional banking is no longer an empty buzzword, and Stefi will share her firsthand experience in making
it a reality.
14:00 - 15:20
The EU for Civil Society Activists Room
How to produce impactful dataviz
Jacopo Ottaviani
Head of Data, Code for Africa
The workshop will go through the entire dataviz process, from research, refine, visualization and
interpretation. The participants will be able to produce their own dataviz with free and/or open source
tools such as Datawrapper and Flourish Studio.
To attend this workshop you'll need familiarity with spreadsheets and online tools, such as Google Drive. A Google account will be needed to use Google Spreadsheet.
14:00 - 15:20
ODIHR Human Rights Room
Designing for impact (Part I)
DATA4CHANGE
Researcher, Data4Change
This workshop introduces participants to the concept of rapid visual prototyping through a
series of practical, hands-on exercises using free online tools and open data. Participants will explore
the data and formulate a research question, which they will then use to create a data driven story.
You'll need to have a laptop, Google Chrome and a gmail or GSuite email address to attend this workshop.
14:00 - 15:20
ForSet DataViz Room
Convert you "itch" for sharing professional experience into social capital (Part II)
Helen Gerasimova
Head of Data Science Faculty, Netology
14:00 - 15:20
BoG Fintech Room
Introduction to Deep Generative Modeling for Computer Vision
Sandro Barnabishvili
Senior Machine Learning Engineer, MaxinAI
Generative machine learning algorithms learn patterns from data in order to create new data samples rather
than just predicting something from it as majority of conventional algorithms do. Imagine a neural network
that can learn how to paint a portrait, generate a speech or simply take your image and put a smile on it.
This workshop will be a tutorial on some of the popular deep learning techniques for generative modeling
including hands-on illustrations and examples.
For this workshop you'll need a laptop, familiarity with basics of Python and Machine Learning.
14:00 - 15:20
Maxin AI Room
Semi-supervised learning
Hrant Khachatrian
Director, YerevaNN
For many real-world problems it is hard to obtain large labeled datasets, but it is
relatively easy to get access to many samples without labels. Semi-supervised learning is a branch of
machine learning which attempts to leverage the unlabeled data to get better performance on the task.
Recently there has been remarkable progress in semi-supervised learning for deep learning architectures.
In this workshop Hrant will discuss the main papers that shaped the research in semi-supervised learning
in the past 4 years.
This workshop require basics of supervised learning: training/test sets, loss functions, regularizers, gradients.
15:20 - 16:00
Break
16:00 - 16:20
Main Stage
The potential of a machine that listens
Markus Lippus
Co-founder & Data Science Lead, MindTitan
So we know that machines can understand humans relatively well in large languages and good conditions. But
how does it behave when conditions aren't that favourable? In this talk Markus will introduce a case study
of using machine learning to assist people over the phone in a low resource language in any situation.
You'll hear about the performance you can expect, the elements that won't work as you'd think they will
and some surprising use cases for a machine that can eavesdrop on calls and scan archives.
16:20 - 16:40
Main Stage
Innovations in Hate Speech Measurement: Deep Learning + Item Response Theory
Chris Kennedy
Data Scientist, Berkeley Institute for Data Science
This talk will describe a newly developed AI methodology for measuring hate speech in social media, which
is a major innovation above existing technology. This method integrates item response theory with a deep
learning-based natural language processing system. This allows us to estimate the bias of human reviewers
who create the training data, and eliminate that bias from our measurement of hate speech - a first in
machine learning. We are further able to measure hate speech at a granular level that incorporates
magnitude: we can track a spectrum of hate speech ranging from genocidal on one extreme to counterspeech
on the other extreme. By tracking hate speech in such an accurate fashion Chris hopes to empower
organizations to now focus on interventions to reduce hate speech over time.
16:40 - 17:00
Main Stage
Engaging conversations
Piero Zagami
Co-founder, Market Cafe Magazine & Encode: Data Journeys
As data visualization practitioners, we all aspire to do always great work. It’s only natural that when we
see amazing projects in the field, we wonder how they did it, or what skills did they master, and lots of
other questions about methodology, struggles, success, and failures. The reason for making a data
visualization magazine in the first place was to get up close and personal with our 'dataviz heroes' and
have a chance to speak with them about their work. As editors of Market Cafe Magazine, we had the
privilege to interview more than 30 amazing contributors, and we asked them the ins-and-outs of their
practice, whether it was project-specific or more process-led. It turns out they are not shy! In our talk,
we will share what we learned during the engaging in conversations we had with our contributors over the
three years of zine-making, lessons learned along the way, precious tips and tricks, and of course, some
weird anecdotes.
16:40 - 17:00
Main Stage
Computational inference of first impressions
Laurent Nguyen
Data Scientist, Eyeware
First impressions play an imprtant role in our day-to-day lives. Be it people we meet for
the first time, appartments we book for a long week-end, or job interviews we conduct or take as
applicants, first impressions are an important component of how we interact with and make judgments on
others. Interestingly, computational models can be trained to infer first impressions up to a certain
level and within specified contexts. In this workshop, Laurent will present you the research he did on the
automatic inference of first impressions in the context of recruitment and apartment rentals
16:00 - 17:40
The EU for Civil Society Activists Room
Data Walking (continuation after the city walk part)
David Hunter
Associate Senior Lecturer, Ravensbourne
In this workshop you will go out and explore the streets of Tbilisi and using smartphones
gather data on what you see and what interests you. You will then go back to create online maps of your
data using Mapbox.
No specific knowledge/experience needed for this workshop,
but a smartphone, smartphone cable to connect to your laptop
and a laptop would be useful. People can work in pairs if they do not all have a laptop.
16:00 - 17:40
ODIHR Human Rights Room
Designing for impact (Part II)
DATA4CHANGE
Researcher, Data4Change
This workshop introduces participants to the concept of rapid visual prototyping through a
series of practical, hands-on exercises using free online tools and open data. Participants will explore
the data and formulate a research question, which they will then use to create a data driven story.
You'll need to have a laptop, Google Chrome and a gmail or GSuite email address to attend this workshop.
16:00 - 17:40
ForSet DataViz Room
POC in a day and demo ElasticSearch + Kibana for viz
Benjamin Auffarth & Elena Nemtseva
Lead Data Scientist, Oleeo
Ben and Elena will show, from start to finish, an NLP analytics project with search and
dashboarding implemented in python and using Kibana and Elasticsearch.
Bringing a laptop to this workshop would be helpful. Also, it'd be good,
but not necessary, if attendees have python installed. The talk is oriented towards people who have some
(at least basic) experience with python, but are not expert level yet.
16:00 - 17:00
BoG Fintech Room
Machine Learning for Beginners: Hands-On Text Classification with R
Anton Prokopyev
Former Data Scientist, The World Bank
The availability of text data has exploded in recent times, and so has the demand for
analysis of that data. This workshop introduces beginners to machine learning applied to text with no
prior experience required to start. The course is applied in nature, and Anton will explain some concepts
in a simplified manner, but the primary aim is to help attendees understand that machine learning is easy
when you know where to start.
You'll need a laptop, familiarity with spreadsheets and scripting
16:00 - 17:00
Maxin AI Room
Semi-supervised learning (Part II)
Hrant Khachatrian
Head, YerevaNN
For many real-world problems it is hard to obtain large labeled datasets, but it is
relatively to get access to many samples without labels. Semi-supervised learning is a branch of machine
learning which attempts to leverage the unlabeled data to get better performance on the task. Recently
there has been remarkable progress in semi-supervised learning for deep learning architectures. In this
workshop we will discuss the main papers that shaped the research in semi-supervised learning in the past
4 years.
This workshop require basics of supervised learning: training/test sets, loss functions, regularizers, gradients.
17:40 - 18:00
Main Stage
Closing
18:00 - 20:00