Sirris 
Data Innovation Group 
Elena(Tsiporkova(
Who are we? 
Collective centre 
of the technology industry 
 Non profit organisation 
 Industry owned 
MISSION 
Federation 
for the technology industry 
“To increase the competitiveness of companies of the Agoria 
sectors through technological innovations”
We have ~2500 member companies 
115 big companies (250 employees) 2485 SME’s (250 employees)
We offer services for 
customized technological innovation 
PRACTICAL ANSWERS 
to tangible technology 
problems 
INNOVATION PROJECTS 
realisation from idea 
to product 
TECHNOLOGICAL TRENDS 
ONLINE LIBRARY 
NEWSLETTERS 
THEMATIC SEMINARS  
workshops 
RESEARCH  DEVELOPMENT 
with industrial partners 
ACCESS TO HIGH-TECH 
infrastructure 
INDIVIDUAL SUPPORT 
SOURCE OF INSPIRATION 
SHARED RISKS 
SERVICES
We operate in 5 domains of technology 
METALS 
COMPOSITES 
PLASTICS  HYBRIDS 
COATINGS 
NANOMATERIALS 
FACTORIES OF THE FUTURE 
WORLD CLASS TECHNOLOGIES 
ADDITIVE MANUFACTURING 
ECO-MECHATRONICS 
SENSORIZED FUTURE 
MODEL BASED DESIGN 
SOFTWARE ENGINEERING 
CLOUD COMPUTING 
DATA INNOVATION
Industrial RD project track record 
TWIRL
Online recommendations for the consumer 
domain 
 Innovation goal 
Provide near-real-time recommendations for hyper-local  
hyper-personal content e.g. concerts, markets, restaurants, local 
events  happenings, … 
EVENTS 
RESTAURANTS 
DEALS 
TWIRL
Data requires handcrafted approaches 
 Challenging application context 
 No explicit user ratings 
 Extremely diverse set of items - low 
coverage 
 Noisy data - item profiling is difficult 
 Standard recommendation approach 
not applicable 
 Collaborative filtering - needs “coverage” 
 Content-based - needs item profiling 
 Hybrid recommendation strategy 
developed 
 Customized collaborative filtering  
content–based algorithms
Paradigm shift to wearable and embedded 
devices 
 Embedded and wearable devices allow to access and manipulate 
diverse information about people and their environment 
 Devices carried by humans or worn in/on the human body as smart phones, 
heart rate meters, Google Glass, ... 
 Sensors integrated into “dumb” objects, such as textiles, toothbrushes, 
mattresses, mirrors, thermostats, …
while HW/SW design remain important, 
it’s all about data 
algorithms 
trends  patterns 
privacy  
security 
uncovering 
insights 
storage
Pro-active decision support in data-intensive 
environments 
 Innovation goal 
Turn an overwhelming amount of data into user- and 
context-sensitive information to increase situational 
awareness and improve decision making 
 capture, model and fuse raw data about users  their 
environment from heterogeneous data sources 
 analyse, filter  pro-actively deliver relevant information 
using appropriate modalities 
h9p://www.astuteproject.eu/(
Emergency dispatching prototype 
Pro-active decision support 
! Aid commanders, superiors,... by 
providing suggestions for actions 
! Show relevant data only 
Server 
Multi-modal communication 
! Integrated audio, video, voice, 
haptic, messaging, map 
annotations,… 
Adaptive map-centric HMI 
! Embedded on-sleeve, in-vehicle, etc. 
! Updated in real-time with tracking 
feeds, annotations,... 
Allbasedonuser 
stateandcontext
User state modeling is still challenging 
 Accurate user state recognition requires 
 real-time user state monitoring 
 robust user state models 
 Real-time continuous data capturing of user state is not fully 
feasible 
 still obtrusive and stigmatizing 
 accuracy vs. low energy (battery) consumption 
 scalable data storage and transfer 
 User state modeling is extremely data-intensive and requires 
 availability of high quality historical data 
 strategies for model personalization 
 domain-specific user state interpretation
The European Platform to Promote Healthy 
Lifestyle and improve care through a Personal 
Persuasive Assistant 
 Innovation goal 
Build a digital coaching platform that continuously 
monitors, advises and interacts with the user to help 
him/her acquire and maintain a healthier lifestyle 
h9p://www.withmeproject.eu/(
User profile modelling 
• lifestyle 
• health 
• personality 
• context
User profile modelling 
• lifestyle 
• health 
• personality 
• context 
• eating  drinking 
• activities 
• frequent locations 
• ...
User profile modelling 
• lifestyle 
• health 
• personality 
• context 
• medical history 
• weight 
• body composition 
• ...
User profile modelling 
• lifestyle 
• health 
• personality 
• context 
• psychological 
• motivational 
• ...
User profile modelling 
• lifestyle 
• health 
• personality 
• context • demographics 
• social 
• professional 
• ...
capture objective  subjective data 
user-friendly questionnaires 
 feedback gathering 
indoor  outdoor 
localisation 
continuous capturing in 
the background 
refining rudimentary 
activity recognition
combine, analyse and enrich data 
• frequent locations 
• most active hours of 
the day 
• ...
Condition monitoring for solar energy production 
 Innovation Goal 
Reduce the maintenance costs of solar plants by means of real-time 
failure prediction  operation anomaly detection through 
pattern detection in historical data 
h9p://www.arrowhead.eu/(
Mediation  querying of 
non-standardized event log data 
through an ontology
Characterization of 
production behavior
Clustering based on 
daily operational dynamics
Data innovation is not … 
 About volume 
 Even relatively small datasets can be very complex to understand and handle 
 Data innovations can already be realized with relatively limited datasets 
 Having access to the right data with the right quality is essential 
 About installing  configuring platforms, libraries or tools 
 It’s about statistics, machine learning, experimentation, trial  error, ... 
 Just data crunching 
 Even the most intelligent algorithm will be useless without some prior knowledge of 
the application domain 
 Gaining initial understanding of the nature, type and quality of data under study is a 
must 
 Ideally the data scientists need to be involved already during the design of the data 
capturing phase 
 Easily scalable in terms of effort and human resources 
 Few data problems can be solved with off-the-shelf products and IT generalists 
 Most data problems require a handcrafted “artisanal” approach and data scientists 
with deep expertise in statistics, machine learning, …
At the core of data innovation are ... 
 The algorithms that interpret data collected from online 
sources, sensors and embedded devices e.g. 
 data integration from heterogeneous sources 
 real time information processing and event recognition 
 inference of additional knowledge from data 
 The availability of the right amount of the right data with the right 
quality 
 The presence of skilled professionals that combine deep mathematical 
skills with practical domain knowledge to exploit data in creative and 
innovative ways
Agenda 
 13:30 - 14:00 : Introduction to the project 
 14:00 - 15:00 : Project partner presentations 
 15:00 - 15:20 : User group involvement 
 15:20 - 15:45 : Break 
 15:45 - 16:45 : Brainstorm on potential demonstrators 
 16:45 - 17:00 : Next steps  closing remarks 
 17:00 - … : Networking drink

2 partners ed_kickoff_sirris

  • 1.
    Sirris Data InnovationGroup Elena(Tsiporkova(
  • 2.
    Who are we? Collective centre of the technology industry Non profit organisation Industry owned MISSION Federation for the technology industry “To increase the competitiveness of companies of the Agoria sectors through technological innovations”
  • 3.
    We have ~2500member companies 115 big companies (250 employees) 2485 SME’s (250 employees)
  • 4.
    We offer servicesfor customized technological innovation PRACTICAL ANSWERS to tangible technology problems INNOVATION PROJECTS realisation from idea to product TECHNOLOGICAL TRENDS ONLINE LIBRARY NEWSLETTERS THEMATIC SEMINARS workshops RESEARCH DEVELOPMENT with industrial partners ACCESS TO HIGH-TECH infrastructure INDIVIDUAL SUPPORT SOURCE OF INSPIRATION SHARED RISKS SERVICES
  • 5.
    We operate in5 domains of technology METALS COMPOSITES PLASTICS HYBRIDS COATINGS NANOMATERIALS FACTORIES OF THE FUTURE WORLD CLASS TECHNOLOGIES ADDITIVE MANUFACTURING ECO-MECHATRONICS SENSORIZED FUTURE MODEL BASED DESIGN SOFTWARE ENGINEERING CLOUD COMPUTING DATA INNOVATION
  • 6.
    Industrial RD projecttrack record TWIRL
  • 7.
    Online recommendations forthe consumer domain Innovation goal Provide near-real-time recommendations for hyper-local hyper-personal content e.g. concerts, markets, restaurants, local events happenings, … EVENTS RESTAURANTS DEALS TWIRL
  • 8.
    Data requires handcraftedapproaches Challenging application context No explicit user ratings Extremely diverse set of items - low coverage Noisy data - item profiling is difficult Standard recommendation approach not applicable Collaborative filtering - needs “coverage” Content-based - needs item profiling Hybrid recommendation strategy developed Customized collaborative filtering content–based algorithms
  • 9.
    Paradigm shift towearable and embedded devices Embedded and wearable devices allow to access and manipulate diverse information about people and their environment Devices carried by humans or worn in/on the human body as smart phones, heart rate meters, Google Glass, ... Sensors integrated into “dumb” objects, such as textiles, toothbrushes, mattresses, mirrors, thermostats, …
  • 10.
    while HW/SW designremain important, it’s all about data algorithms trends patterns privacy security uncovering insights storage
  • 11.
    Pro-active decision supportin data-intensive environments Innovation goal Turn an overwhelming amount of data into user- and context-sensitive information to increase situational awareness and improve decision making capture, model and fuse raw data about users their environment from heterogeneous data sources analyse, filter pro-actively deliver relevant information using appropriate modalities h9p://www.astuteproject.eu/(
  • 12.
    Emergency dispatching prototype Pro-active decision support ! Aid commanders, superiors,... by providing suggestions for actions ! Show relevant data only Server Multi-modal communication ! Integrated audio, video, voice, haptic, messaging, map annotations,… Adaptive map-centric HMI ! Embedded on-sleeve, in-vehicle, etc. ! Updated in real-time with tracking feeds, annotations,... Allbasedonuser stateandcontext
  • 13.
    User state modelingis still challenging Accurate user state recognition requires real-time user state monitoring robust user state models Real-time continuous data capturing of user state is not fully feasible still obtrusive and stigmatizing accuracy vs. low energy (battery) consumption scalable data storage and transfer User state modeling is extremely data-intensive and requires availability of high quality historical data strategies for model personalization domain-specific user state interpretation
  • 14.
    The European Platformto Promote Healthy Lifestyle and improve care through a Personal Persuasive Assistant Innovation goal Build a digital coaching platform that continuously monitors, advises and interacts with the user to help him/her acquire and maintain a healthier lifestyle h9p://www.withmeproject.eu/(
  • 15.
    User profile modelling • lifestyle • health • personality • context
  • 16.
    User profile modelling • lifestyle • health • personality • context • eating drinking • activities • frequent locations • ...
  • 17.
    User profile modelling • lifestyle • health • personality • context • medical history • weight • body composition • ...
  • 18.
    User profile modelling • lifestyle • health • personality • context • psychological • motivational • ...
  • 19.
    User profile modelling • lifestyle • health • personality • context • demographics • social • professional • ...
  • 20.
    capture objective subjective data user-friendly questionnaires feedback gathering indoor outdoor localisation continuous capturing in the background refining rudimentary activity recognition
  • 21.
    combine, analyse andenrich data • frequent locations • most active hours of the day • ...
  • 22.
    Condition monitoring forsolar energy production Innovation Goal Reduce the maintenance costs of solar plants by means of real-time failure prediction operation anomaly detection through pattern detection in historical data h9p://www.arrowhead.eu/(
  • 23.
    Mediation queryingof non-standardized event log data through an ontology
  • 24.
  • 25.
    Clustering based on daily operational dynamics
  • 26.
    Data innovation isnot … About volume Even relatively small datasets can be very complex to understand and handle Data innovations can already be realized with relatively limited datasets Having access to the right data with the right quality is essential About installing configuring platforms, libraries or tools It’s about statistics, machine learning, experimentation, trial error, ... Just data crunching Even the most intelligent algorithm will be useless without some prior knowledge of the application domain Gaining initial understanding of the nature, type and quality of data under study is a must Ideally the data scientists need to be involved already during the design of the data capturing phase Easily scalable in terms of effort and human resources Few data problems can be solved with off-the-shelf products and IT generalists Most data problems require a handcrafted “artisanal” approach and data scientists with deep expertise in statistics, machine learning, …
  • 27.
    At the coreof data innovation are ... The algorithms that interpret data collected from online sources, sensors and embedded devices e.g. data integration from heterogeneous sources real time information processing and event recognition inference of additional knowledge from data The availability of the right amount of the right data with the right quality The presence of skilled professionals that combine deep mathematical skills with practical domain knowledge to exploit data in creative and innovative ways
  • 28.
    Agenda 13:30- 14:00 : Introduction to the project 14:00 - 15:00 : Project partner presentations 15:00 - 15:20 : User group involvement 15:20 - 15:45 : Break 15:45 - 16:45 : Brainstorm on potential demonstrators 16:45 - 17:00 : Next steps closing remarks 17:00 - … : Networking drink