data - driven journalism 1

Data Driven Journalism
Giulia Dezi, Giorgio Dimino, Maurizio Mazzoneschi,
Alberto Messina, Sabino Metta, Giuseppe Mondelli, Maurizio Montagnuolo
RAI – Radiotelevisione Italiana
Centre for Research and Technological Innovation
FIAT/IFTA World Conference 2016
“Rethink the future of AV”
October 12 to October 15, Warsaw

FIAT/IFTA World Conference 2016, October 12-15, Warsaw Data Driven Journalism @ RAI
Agenda
Definitions
Best practises
Our approach
The data team
Tools & architectures
Some preliminary results
Development ideas
Conclusions

Role of Data in Journalism
 as origin of the news and of related content
(data driven journalism)
 as accurate and verifiable description of inherent
semantic aspects
(precision journalism)
 Basic building block for presentation / usage of news
content
(visual journalism)

Extracting sense out of data to create newsworthy
stories
This implies
Having data
Analysing data
Identifying “sense”, or “sense the news”
Presenting data
Story Publication

some insight
Story Publication
Data Harvesting
Which data?
Which formats?
Which time window?
Data Analysis:
Content Analysis
Semantic Analysis
Statistic Analysis
Data Classification:
Ontologies
Automated categorisation
Team
Collaboration
Editorial Process
Detecting the Story
Modelling the Story
Developing the Story
Platform Adaptation
Second Screen
Interactivity

One example
Distribution of feminicides in Italy between Jan 2012 and Aug 2015
From Federica Quaglia Msc Thesis - 2015

Wanting a 100% digital
product
Huge investments in
integrated tech. resources
Data Journalism as a
public service
Visual Journalism and ad
hoc apps
Continuous improvement
of product quality
Perfectly integratesd desks
(journalists & techies)
Data as source for
narration
Notable technical
competence of journalists
Open Source and Open
Data Journalism
Crowdsourcing as a
resource
Investigative Journalism
through data
International collaborations
for data exchange
Expanding scope from
local to international
Experimental editorial
techniques
Editorial objectives Production & org.
approach
4
Best practises

Our Approach
A combination of the main features detected in best
practise
RAI wants to foster an approach at Data Journalism
orientated to
Increase information trustworthiness, based on harvested,
analysed and verifiable data
Improve user experience through visual presentation of data
Obtain more and more “full digital” products exploitable on a
variety of platforms

 Define a workflow model for the editorial staffs,
identifying organisational impacts
 Identify the most appropriate practises and
approaches at Data Journalism among the many
available
 Design and implement a toolbox and an
integrated platform supporting the workflow
Objectives

Architectural Overview
Project Management
(team management, collaboration)
Story Modelling
(FreeMind)
Sources
(Concept Book)
Data Platform
(CKAN)
Visualization
(DataWrapper)
Multiplatform Publishing
(Es. Wordpress)
«Data Team»
«Customer»

The Data Team
A group of people with multidisciplinary skills
Journalistic skills
Technology skills
Agile interaction towards the objectives
Technical members support journalists in finding / harvesting
/processing data
Journalists set the editorial line, develop the story, assess data
relevance
Result is that
Efficiency increased due to sinergy and delegation
Cross-semination gives birth to unexpected explorations

Modelling the Story
https://webbrain.com/brainpage/brain/434E72FE-3EED-7B13-2F44-561D8F294F28

Which Information Sources?
Internet
News
Media
News
News
in
RSS
RAI
Programmes
Open Data
National
TV News
(incl. RAI)
Information Domain of Interest
Input Google
RAI CMS
EVN
News
Agen
cies
Other subscr.
Infota
in

So what?
No one alone covers everything is needed for data
journalism
Neither Google
We need a wider approach than simply “search on
the web”
Solution:
Integrated and flexible search and analysis of heterogeneous
sources at enterprise level

Which Information Sources?
Internet
News
Media
News
News
in
RSS
RAI
Programmes
Open Data
National
TV News
(incl. RAI)
Information Domain of Interest
Input Concept Book
Input Google
RAI CMS
EVN
News
Agen
cies
Other subscr.
Infota
in
Thematic Aggregations
(Hyper Media News)

RAI Concept Book
Rai Concept Book is a portal for professional
information services that addresses DDJ tasks with
a uniform and holistic approach
Artificial intelligence and advanced statistical tools
are used to automate tasks such as information
extraction and multimedia content analysis
The system allows to define customized search
profiles that are automatically and dynamically
updated with the relevant contents found in the
monitored information sources

1. Ingest
• RSS feeds
• Blogs
• DTT streams
• EBU Eurovision News
• Rai Archives
2. Process
• Speech to Text
• Natural Language
Processing
• Document Classification
• Named Entity Recognition
3. Understand
• News aggregation
• Topic identification
• Data warehousing
4. Archive & Access
• Indexing
• Search & Retrieval
• Browsing & filtering
• Recommendation
• Exportation
Processing Pipeline

Available
profiles
New profile
registration
Personalised Dashboard

Sources
Open Data
Infotain
News
in
RSS
News
Agen
cies
EVN
National
TV News
(incl. RAI)
Thematic Aggregations

Filter
& sort
Example: TV Content

Interactive charts of semantic entities

Exploring facts

Localising in space

… and time

A platform for the Data Value Chain
Comprehensive Knowledge Archive Network (CKAN)
Open Source used by many organisations as a
platform for open data publication
Used in the project as a platform for data journalism
production
Editorial staffs  CKAN organisations
Easy integration with other CKAN-based open data
repositories
Provides updates of data automatically
Extensible via plugins

Visualisation
Many many options available
Default choice for Datawrapper in this phase of the
project
Open Source
Many graphics & many options
Extensible via plugins

Integration
Source data selected from the RAI Concept Book
become datasets in CKAN
Datasets harvested by CKAN are classified and
analysed by RAI Concept Book toolbox
Integration between CKAN datasets and
Datawrapper
SSO between CKAN and Concept Book

Some initial products α

Conclusions
Finding the “right” approach at Data Journalism
taking into account RAI’s peculiarities
Multidisciplinarity is key
Integration of proprietary and SotA tools works
Still much work to do at all levels
Integration
Workflow
Processes and skills
Very good feedback from RAI top level mgmt

(some) Future R&I activity
Tighter integration with Semantic Data
Open data in RDF
Developing new automatic classification
technologies
On more relevant taxonomies than those currently used
Improve model for “concepts”
Evoluting towards more semantic structuring
Does Visual Search play a role in Data Journalism?
Second screen & data journalism
…

Data is the new soil
- David McCandless -
Sabino Metta
RAI – Radiotelevisione Italiana
Centre for Research and Technological Innovation
sabino.metta@rai.it

data - driven journalism 1

More Related Content

What's hot

Viewers also liked

Similar to data - driven journalism 1

More from FIAT/IFTA

Recently uploaded

data - driven journalism 1