3. The
Reincarnation App
Filip Ilievski
Sylvia van Schie
Wouter Stuifmeel
The Reincarnation App • A social web application to let you see who you were in previous lives • The Social Web • March 20th, 2014, VU University Amsterdam
GROUP
16
4. Introduction / Approach
TARGET Users of Facebook
GOAL Find the reincarnation chain of the user (all predecessors)
METHOD We hook up the user's date of birth to a person who deceased around the same date.
(margin up to three weeks)
VALUE The app is unique for it links data from Facebook to DBpedia and shows info about deceased people
It furthermore has a high entertainment value
The Reincarnation App • A social web application to let you see who you were in previous lives • The Social Web • March 20th, 2014, VU University Amsterdam
5. Data flow
The Reincarnation App • A social web application to let you see who you were in previous lives • The Social Web • March 20th, 2014, VU University Amsterdam
6. Data clustering
The Reincarnation App • A social web application to let you see who you were in previous lives • The Social Web • March 20th, 2014, VU University Amsterdam
Clustering is based on the
8 results we received when
entering the initial birth date
of 20 March 2007.
7. User interface
The Reincarnation App • A social web application to let you see who you were in previous lives • The Social Web • March 20th, 2014, VU University Amsterdam
9. Social Web - Facebook Activity Checker
Thomas Berger, Kristoffer Lie Braathen, Timo Nijhof
Vrije Universiteit Amsterdam
March 20th, 2014
10. What?
How?
Who?
What does it look like?
1 What?
2 How?
3 Who?
4 What does it look like?
Thomas Berger, Kristoffer Lie Braathen, Timo Nijhof Social Web - Facebook Activity Checker
11. What?
How?
Who?
What does it look like?
Aim of the application
The overall goal is to explore when (ie: which days? what time?)
users are the most active on Facebook.
Thomas Berger, Kristoffer Lie Braathen, Timo Nijhof Social Web - Facebook Activity Checker
12. What?
How?
Who?
What does it look like?
Data mining
Pyhon script using Python SDK for Facebook;
Getting friends statuses and likes;
Categorize them, sum up all likes and statuses within a
category;
For the last week, sum up friends activities for each day.
Thomas Berger, Kristoffer Lie Braathen, Timo Nijhof Social Web - Facebook Activity Checker
13. What?
How?
Who?
What does it look like?
Individual work
Thomas: Python code, documentation;
Timo: Charts, design of app, data integration and
documentation;
Kristoffer: Charts, design of app, data integration and
documentation.
Thomas Berger, Kristoffer Lie Braathen, Timo Nijhof Social Web - Facebook Activity Checker
14. What?
How?
Who?
What does it look like?
Thomas Berger, Kristoffer Lie Braathen, Timo Nijhof Social Web - Facebook Activity Checker
15. What?
How?
Who?
What does it look like?
Thomas Berger, Kristoffer Lie Braathen, Timo Nijhof Social Web - Facebook Activity Checker
16. What?
How?
Who?
What does it look like?
Thomas Berger, Kristoffer Lie Braathen, Timo Nijhof Social Web - Facebook Activity Checker
17. What?
How?
Who?
What does it look like?
Thomas Berger, Kristoffer Lie Braathen, Timo Nijhof Social Web - Facebook Activity Checker
18. What?
How?
Who?
What does it look like?
Thomas Berger, Kristoffer Lie Braathen, Timo Nijhof Social Web - Facebook Activity Checker
22. Feature I: Clustering genres
blues-rock > blues & rock
viking metal > metal
synthpop > electronic & pop
Use pre-determined general tags
blues, classical, country, dance, electronic, folk, hip-hop,
indie, jazz, latin, metal, musicals, pop, reggae, rnb, rock
23. Feature II: Clustering
venues
Classification: most occurring word (genre)
Bag of Words model based on clustered tags
{'blues':
0.0,
'classical':
0.0,
'country':
25.476190476190474,
'dance':
7.1428571428571423,
'electronic':
7.1428571428571423,
'folk':
25.476190476190474,
'hip-‐hop':
0.0,
'indie':
2.8571428571428572
...
24. Feature III: Filtering &
Visualizing of venues
| JSON data
| venue name
| latitude, longitude
| genre classification
| genre vector
25.
26.
27.
28.
29.
30.
31. | Limitations
{ Cold Start Problem/Popularity bias,
ad hoc annotation behavior,
weak labeling
{ Non-agreement on taxonomies
{ Ill-defined genre labels
{ Scalability of genre taxonomies
| Scope
{ From taxonomies to folksonomies?
32. | Evaluation
{ User Qualitative Evaluation of the
Application
| Future Work / Developments
{ Hierarchical Classification
{ Alternative Autotagging Models
{ Combining data sources:
hybrid context - content systems for
music classification & recommendation
34. Exploiting Twitter data for the
automatic annotation of Dutch
Public TV Programming
Group 22
Guido van Bruggen, Jochem Havermans, Shailin
Mohan & Frank Schurgers
The Social Web 2014, VU Amsterdam | Final Presentation Social Web App
35. Goal:
Helping the Netherlands Institute of
Sound and Vision to
automatically annotate TV shows
Target user group:
Clients of Sound and Vision
36. Advantages:
● Use of existing data;
● inexpensive;
● instantaneous annotation;
● easy to implement;
● automatic detection of hot topics;
● less need for tedious human annotation.
much advantage
wow
such cheap
very automatic
amaze
so data
39. Additions
● Words per minute annotation
● fragmentation of episodes;
● access to tweet content of all tweets;
● normalizing tweet count with viewer
count;
42. "Footprint - Where I've Been"
is a designed map app for
adding notes and marking
places
“Flickr”
capture, create, and share
photos
“Instagram”
way to capture and share the
world's moments
“Worldcam”
easy way of finding photos
from a specific venue
43.
44. WhatsAround
FAST GLOBAL FUN
Way to share your life and travels with friends, family and community
Easiest way of finding out what's happening right now, anywhere and
everywhere
67. GROUP 26-APP INTRODUCTION
• Goal: Find top popular places among my friends
• Function:
• 1. Filter different type of places
• 2. Find people from different countries like to go where?
• 3. Analyze our data
68.
69. INDIVIDUAL WORK
• Annan Cheng: 1. Retrieve friends data from Facebook; 2.
Extract where friends have been from data; 3. Get friends
nationality through Google Geocoding API
• Jiahui Chen: Design the demo, realize the map, filter and
sequencing functions through Google map API, d3 and
JavaScript
• Ziyan Zong: Realize analyze data function by using different
visualization methods like histogram and pie chart, etc. Find
out meaningful and useful facts behind those data.
73. Approach
● Search for query
○ Select most popular tweets
○ Find most relevant terms
○ Look for synonyms
● Select relevant terms
○ Refine the query
○ Update the results
77. Predicting attendance and outcome
of local elections - A web
application
Group 29:
Remco Draijer
Renee Vaessen
Lily Martinez Ugaz
78. Description & Goal
• A web application
▫ Local elections on March 19th in the Netherlands
• Predicting outcome and attendance
▫ Based on number of tweets
• Functions as an informative app
• Investigates correlation between tweets and:
▫ Outcome (number of seats in municipal council)
▫ Attendance (turnout)
79. Approach
• Mining Twitter for location-based tweets
▫ Amsterdam, Den Haag, Rotterdam and Utrecht
• Nine parties selected
▫ All occur in four cities
• Extract tweets with party-name in text, user-
name or user-description
▫ Preprocess data for duplicates and party-
generated tweets
80. Data
• Twitter streaming API
▫ Fetch tweets during 24 hours
• Interactive visualizations with d3.js
• Results: correlation between % of seats in
council and % of tweets about parties
▫ For Amsterdam, Den Haag, Rotterdam, Utrecht
• Distribution of tweets over mining period
84. Motivation & Purpose
• information
• easy, visual way of getting up to speed
• easier than reading a newspaper or an RSS reader
• like a more visual, web-based, Flipboard
85. Data Sources
• media agencies around the world (list from
Wikipedia)
• Twitter
• Bing image search
• Sentiment140
86. Data Flow
• get the trending stories from the news agencies
• find tweets about the stories
• perform sentiment analysis on the tweets
• get images (and videos) related to the story
• export via JSON API
87. Data Flow
• get the trending stories from the news agencies
• find tweets about the stories
• perform sentiment analysis on the tweets
• get images (and videos) related to the story
• export via JSON API
filtered by region and/or tag
88. Display
• Web app
• Mobile friendly
• Responsive gallery
• See a prototype at http://www.mihneadb.net/news_timeline/
92. The Social Web 2014 – Final assignment – G31: Jeroen Wever JWR970, Jeffrey Bruijntjes JBS257, Marije ten Brink MBK262
APPSPIRE!
What are his/her true aspirations?
An entertaining way to discover dreams and desires of
influential people… or your friends!
93. The Social Web 2014 – Final assignment – G31: Jeroen Wever JWR970, Jeffrey Bruijntjes JBS257, Marije ten Brink MBK262
Goal
a. Discover aspirations of a person of your interest
b. Compare your aspirations with others
c. Visualize aspirations in tag clouds or photo clouds
d. Find people who have specific aspirations
e. Follow trends in our shared aspirations
Related work:
Twitter Account Showdown
94. The Social Web 2014 – Final assignment – G31: Jeroen Wever JWR970, Jeffrey Bruijntjes JBS257, Marije ten Brink MBK262
Approach: discover
1. Retreive all tweets of user with Greptweet.com
2. Clean up tweets, remove punctuation, stop words, links
3. Use stemmer: matches verbs and nouns with same stem
4. Use word list with aspirational verbs to extract tweets
with aspirations
5. Count occurrences of meaningful word
Other possible data sources: FB, blogs, Pinterest, Google+, ...
95. The Social Web 2014 – Final assignment – G31: Jeroen Wever JWR970, Jeffrey Bruijntjes JBS257, Marije ten Brink MBK262
Approach: compare
Determine similarity
between 2 users, based
on amount of matching
words: correlation
96. The Social Web 2014 – Final assignment – G31: Jeroen Wever JWR970, Jeffrey Bruijntjes JBS257, Marije ten Brink MBK262
Approach: visualize
Tag cloud
Image cloud
Timeline
107. Song Recommendation
■ Based on the selected genre and on the
likes of followed people
■ Based on the recent likes of the followed
artists
■ Visualized search results