Social Web Course @VU Amsterdam: Final Student Presentations

Social Web @VU���2015

Final Student PresentationsLecturer: Lora Aroyo

The Network InstituteVU University Amsterdam

Student housing in Amsterdam

Group 1

Menno Nelis Rutger van Gennep Melanie van der Velde

The idea

- Finding a home in a new city can be difficult for students

- Provide an overview of the largest student houses and

their surroundings - Preferably even filter on certain types of venues such as

restaurants, bars or fitness centers

The idea

Red circles = existing student housing Blue circles = student housing in development

Used data sources - Amsterdam Open Data - Dataset that provides

DUWO/de Key available housing for students - Foursquare - Data on public venues

- Facebook - Likes of the venues

Feature 1 - Amsterdam Open Data

- Dataset with stastic information about the student

houses - Name, address, coordinates

- Imported into MySQL database

Feature 2 - Information extracted from Foursquare

- Search for venues based on coordinates from

Amsterdam Open Data dataset - Result is exported into a JSON file

- JSON is displayed on website

via Javascript or PHP

Feature 3 - Retrieve extra information from Facebook pages of the

student houses - Likes - Other information possible

Screenshots

Screenshots

The GoTwi Grab BagJosephine Jessen, Roxane Lubbers & Arjan Vis

Group 2

Structure

The GoTwi Grab BagWorkflow of ApplicationSome screenshots AnalyseFutureIndividualQ&A

The GoTwi Grab Bag

The GoTwi Grab BagGoal → Get surprised by new interesting topics and subject!Method → Jump into a black box, with just little swing

Data Sources and services used next to our code:→ Twitter(API): tweets → Google+(API)→ Topia Term extractor→ NYtimes API → DBpedia → Wikipedia→ ‘Let me Google that for you’ server

Workflow of the Application- Query a subject of your interest

- Retrieve 20 Popular Tweets (if less than 6 popular tweets, go to 20 recent tweets) + 20 Google+ Posts → One list is formed out of the Tweets and Google+ Posts

- Choose random one Tweet or G+Post → If Twitter, the timeline of the users account will be retrieved → If Google, the activity of user will be retrieved

- Randomly choose one Tweet/activity→ mine subjects, take one randomly → result: one new ‘inspiring’ subject→ subject links to DBpedia, wikipedia and 10 articles from the New York Times

AnalyseCluster

◆ K means cluster ◆ Popular tweets ◆ Google+ posts◆ Topics

Link◆ Find similarities between tweets, posts, and subjects

Gain insight into interests of a Twitter and/or Google+ user

Individual: Josephine➔ Rationale: Surprise yourself with a new inspiring topic➔ Motivation: Let go of your structured search, through the

uncertainty you will invite new interesting topics out of your scope

➔ Evaluation: How ‘new’ are the subjects to the user ➔ Scoping: Is the new subject really new? Design of

application➔ Future work: Create a sort of loop, so the application

can see if the user is already familiar with the randomly chosen topic (and if so, take other topic etc.)

Individual: Roxane

➔ Rationale: Gain new insights with the topics you like

➔ Motivation: Meet new people with same interests

➔ Scoping: More user friendly ➔ Evaluation: How many users? ➔ Future work: Do clusters change over time?

Individual: Arjan

➔ Selecting random tweet / activity➔ Limitations: Number of tweets / activities that

can be retrieved➔ Evaluation: Usage of the app➔ Future work: Visualisation & tagging

Giada Binelli

Rio Essed Aydan Gasimova

Group 3

VU Amsterdam: Social Web 2015

Mobile ApplicationBrowse through potential roommates’ profiles, chat and schedule an appointment

Compatible roommates matched based on Social Web data:

Facebook Spotify In-app survey

LikesFriendsUser Profile infoAttended Events

Saved tracksPersonal PlaylistGenres

Personal preferencesToleranceBasic Information

Innovative feature: Using data on Music taste to determine compatibility

Added value:

➔ Makes search for a roommate a fun and social activity ➔ Better system for screening roommates, increases probability of finding the right person

Data retrieval

Rationale: Wide range of data already available

Motivation: Enhance usability

Scoping: Retrieve user related data (basic info, events, pictures, likes etc) and friendlist

Evaluation: Check for similarity in data of attributes that co-occur in the different sources

Analysing & Processing

● Frequency analysis of genres listened to by users & rank => matching the highest ranks

● Identify overlaps in tracks listened by users & score and rank matches

● Encode user preference/tolerance weighed based on assigned importance=> Calculate a satisfaction score of each match=> Calculate the Match percentage using geometric mean

● Filter match suggestions by clustering user-data on gender, age, and location

Matching algorithm

Rationale: “Similarity Theory” in Psychology;Matching based on (Music) taste

Target users: Younger demographic;Roommate seekers likely to use Spotify

Survey Data => (Preference vs. Tolerance)*Importance

Limitations: “Type 1 vs. Type 2 errors”;Not using a parameter of relevance VS, using an irrelevant one

Evaluating success: Collect and analyse data on actual matches as opposed to Match percentages

Future work: Use association rules;Look at more than overlaps in taste but also at similarities (would require a similarity score)

Learning & Optimizing

● Learning as an optimization problem● Backpropagation algorithm● Update weights● Minimize loss function

➔ Improved matches

Thank you for your attention!

Group 4: Galen, Isabelle & Vigo19 March 2015

S

ProblemSmartphones have become a daily requisite, but battery life is not keeping up with our demand.

Crowdsource the locations of the numerous yet hard to find power sockets in public places.

Solution

S

S

S

S

S

S

S

S 2 sockets.Inside Starbucks.Upon entering on your right.Next to a table.In the corner.Close to the floor.

Features & Functionalitieso Crowdsourcing sockets

o Google (Indoor) Maps

o Socket-community

o Link to social networks

o Gamification:

Point-system, rankings & unlockables

Screenshot of actual app

Research: Similar appsTomTom Waze PulsePoint Healthmap

Research: Questionnaire (1)o Heterogeneous group

o n = 100

o ♂/♀ = 51/49

o Mean age = 31

Research: Questionnaire (2)Functionality appreciation:

+ + Google Maps

+ + Location finder

+ Search

- Profile

- Signup

Motivators for (non-)contributors to contribute to crowdsourcing:

Point system

Social benefits

Curiosity

Conscience

Planned data analyses• Summarization: visualize and create report on data

• For the developer as well as the user (in-app)

• Clustering: group and structure the data• Detect if users that make use of the app also contribute to the community

• Association rule learning: show relationships between variables• Track users’ frequent locations and suggest favorite sockets or near sockets

Event ExplorerTSW 2015, Group 6

Problem-statement

● Hard to get overview about local events

● Lack of good recommendation for events

● Facebook often used to search for events

● Not possible to search for events on a map

Event Cluster 1

Event Cluster 2

Event Cluster 3

ApplicationLe

xica

l Si

mila

rity

FeaturesFacebook-Login● Rationale: Personalization● Motivation: Find interesting Events● Scoping: Analyze previous Event attendances● Limitation:

○ No additional Profile information○ Change of user taste○ Not the full picture

● Evaluation: Self-Evaluation, Analysis on past Events● Future work: Evaluate facebook likes

FeaturesFind Local Facebook Events● Rationale: Search/Recommendation● Motivation: Missing local search● Scoping:

○ Events which are present on external sources ○ Public Events○ Lexical Similarity Search + Geographic filtering

● Evaluation: Compare Map with manual FB-crawling ● Future Work:

○ Expand the amount of external sources○ Search for events by (facebook) place’s

FeaturesMap View for Events● Rationale: Visualization● Motivation: Replace Lists● Scoping:

○ Show events for one day/week○ Show limited number of events○ Show only one event at one location

● Evaluation: Usability & Performance● Future Work:

○ Display more Event Information○ Make timeframe configurable○ Advanced Clustering

Technology

http://eventexplorer.openshift.com/

http://graph.facebook.com/

http://api.meetup.com/open_eventshttp://api.eventful.com/json/events/search

https://www.eventbriteapi.com/v3/events/search/

http://ws.audioscrobbler.com/2.0/?method=geo.getevents

GuideMeNowYour Social Tourist Guide

Ali Harrak: Front-end Yassin el Aajati: Business Case Abdelilah Mounir: Back-end

Team 8

PROBLEM STATEMENTDestinations compete for tourists in a very competitive environment (Kevin K.F Wong, 2001)

It is observed that major tourist problems are deviation in the arrangements made for their stay, visit, transport and unexpected expenses. (Chockalingam & Ganesh, 2010).

TOURIST INDUSTRY IS VITAL

one in 11 jobs globally 9 % of the world’s economy

World Travel & Tourism Council (WTTC), 2012

API’S

Data Analysis

● Cluster Analysis

● Classification Analysis

WHO MIGHT BE INTERESTED?

OUR AIM

\

1 2 3

FEATURES Collect ideas for your trips and getaways

● Nearby recommendations at-the-spot

● Discover the experiences from the social community

REFERENCES Ap, J., & Wong, K. K. (2001). Case study on tour guiding: Professionalism, issues and problems. Tourism Management, 22(5), 551-563.

Ganesh, A. A., & Chockalingam, M. (2010). Problems encountered by tourists. Business and Economic Horizons, (03), 68-72.

QUESTIONS

Routist

Group 9Bas Sijtsma, Hayo Bart & Stefan Paap

March 19th, 2015

✓ Idea○ Visualizing movements between points of interested

throughout Amsterdam○ Based on Flickr and Foursquare data

✓ Purpose○ Describe movements of people throughout the city○ Display interaction of venues with one another

✓ Target Group○ Businesses (e.g. marketing)○ Tourist (e.g. provide insight, discover hot spots,

assist trip planning)

Concept

Visualisation✓ Venue exploration

✓ Movement exploration

✓ Filtering○ On time period○ On tourist nationality

Data✓ Foursquare

○ Points of Interest: ~ 800 venues

✓ Flickr○ Photos containing geo-tags: ~82.000○ Movements between points of interest

➢ Issue○ Flickr users ‘location’ field (country of origin) is

unstructured○ Solution: Geocode Location Lookup API

Data Analysis

✓ Clustering○ Soft assignment of photos to venues○ Aggregation of individual movements into

aggregated movements✓ Patterns

○ Identification of movements of individual tourists between venues

Group Effort

✓ Bas○ Back-end development; data processing

✓ Hayo & Stefan○ Front-end development; visualizations in d3

TripReco!

Martin Altmann Sebastian Hoffmann

Hsu-Young Ho  

Agenda!•  Introduc*on  •  Model  of  TripReco  •  Demo  •  Discussion  – Limita*on  – Future  feature    

About TripReco!•  TripReco  aims  to  help  users  to  find  the  popular  places  

•  Using  the  friends´  photos  which  shared  on  thier  Facebook  and  Instagram  

•  TripReco  displays  the  overview  in  the  whole  map  (which  Facebook  &  Instagram  do  not  do)    

§  Find  the  paFerns  of  different  genders  

Model!

Data  processing  Data  analyzing  

TOP  10  places  

Data  Visualizing  Data  gathering  

•  Facebook  API  ✓  Friends  

✓  Gender  ✓  Photos  

✓  ID  ✓  Created_*me  ✓  Name_tags  ✓  Place  ✓  Source  ✓  Link  

•  Instagram  API  ✓  Rela*onships  

✓  Follows  ✓  Media  

•  Geo  informa*on  •  genderize.io    

✓  Gender  of  Instagrams’  followers  

•  Top  popular  places  ranking  by  friends  •  Categories  of  gender  •  Map  the  geotagged  photoz    

Model!•  Facebook  API  

✓  Friends  ✓  Gender  ✓  Photos  

✓  ID  ✓  Created_*me  ✓  Name_tags  ✓  Place  ✓  Source  ✓  Link  

•  Instagram  API  ✓  Rela*onships  

✓  Follows  ✓  Media  

Model!

•  Geo  informaAon  

•  genderize.io    ✓  Gender  of  Instagrams’  

followers  

Model!

•  Top  popular  places  ranking  by  friends  

•  Map  the  geotagged  photos  on  the  map  

•  Categories  of  gender    

Demo!

Demo!

Demo!

Demo!

Demo!

Demo!

Demo!

Limitation!•  Facebook  API  ✓ The  endpoints  of    friends´  photos  only  allowed  the  photos  which  friends  were  tagged  in.  

✓ Retrieved  the  photos  from  the  albums  do  not  provide  the  informa*on  of  la*tude  &  longitude.  

•  Instagram  API  ✓ The  endpoints  of  follower  do  not  include  the  informa*on  of  media  

✓ The  endpoints  of  media  only  can  retrieve  the  public  user  ID  even  though  it  it  is  follower  

•  Gender  ü The  informa*on  of  instagram’s  first  name  does  not  match  genderize.io    

Demo video!

Work!•  Mar*n  Altmann  

–  Data  analysis  –  Web  applica*on  

•  Sebas*an  Hoffmann  –  Data  collec*on  

•  Hsu-­‐Young  Ho  –  Data  collec*on  –  Slides  

The Social MuseumJohan AssinkMarc JacobsNicky van Oorschot

Team 11

Problem

- Museum Guestbook- Overview of museums in NL

- Combination of both problems result in:

The Social Museum

Solution● Museums overview sort by county.● Additional information for each museum.● Tweets as guestbook notes.● Slider that shows popularity over time.● Museum recommendations.● Analyse popularity and social Phenomena.

Datasets

The Social Museum

DBpedia

WikipediaTwitter

Yelp

Google

Demo time!

https://www.youtube.com/watch?v=BuKDCQXyooI&feature=youtu.be

AcknowledgmentsJohan Marc Nicky

Timeline Tweet Crawler Counties

Switching leaflet maps Twitter integration Museum Description (DBPedia)

Coffee provider Yelp Integration Museum Recommendation

DBPedia integration

Social Web

Group 12Baraa Alnawakil

Amir Hossein ZadehIoannis Markopoulos

Goals- Help users find real-time events based on location of their preferences - Provide information about events to help users decide on which

events to go to

Real-time events finder app

● Facebook API○ Event Name○ Description○ End time○ Place○ attendees

● Twitter API○ tweets about the event

Data Used

● Ranking Algorithm based on tweet texts analysis.○ categorise tweet words in 5 groups ( very good, good , normal, bad, very bad).○ assign weight measure for each group.○ calculate the frequency of words in each group.○ calculate the rank based on frequency and weight.

● Classification based on location of attendees○ classify locations based on the cities in which the attendees currently live.○ find number of attendees per city and calculate percentage.

Type of data analysis

The  Social  Playlist  

Group  14  Eric,  Peter,  Lara  &  Paul  

Group 14The Social Playlist

Eric, Peter, Lara & Paul

+

The  Social  Playlist  

Group  14  Eric,  Peter,  Lara  &  Paul  

The Social Playlist

The Social Playlist

Party  9me!  

Genre  Age

Artist   Tracks  

Event:     SocWeb2015  

Party  9me!  Home  

Age  

Unknown   1  

Till  16   0  

16  9ll  18   0  

18  9ll  25   1  

25  9ll  30   1  

30  9ll  40   0  

40  and  older   1  

The Social Playlist

Party  9me!  

Genre  Age

Artist   Tracks  

The Social Playlist

Party  9me!  Home  

Genre  

Pop   10  

House   8  

Electro  House   6  

Edm   3  

Permanent  Wave   2  

R  &  B   2  

Neo  Soul   2  

Dutch  House   2  

New  Wave   1  

The Social Playlist

Party  9me!  

Genre  Age

Artist   Tracks  

The Social Playlist

Party  9me!  Home  

Ar9st  

Bakermat   4  

Mar9n  Garrix   4  

ATer  House  FlicFlac   3  

Serif  Chase   3  

John  Legend   3  

Nicky  Romero   1  

One  Direc9on   1  

Jus9n  Bieber   1  

The Social Playlist

Party  9me!  

Genre  Age

Artist   Tracks  

The Social Playlist

Party  9me!  Home  

Event  

Ar9sts  in  this  selec9on  

Include  top  

Number  of  tracks  per  ar9st  (1  –  10)  

25  

20  2  

The Social Playlist

Party  9me!  

Genre  Age

Artist   Tracks  

The Social Playlist

The  Social  Playlist  

Group  14  Eric,  Peter,  Lara  &  Paul  

The Social PlaylistGroup 14

Features: •  Select age group •  Select genre •  Select artists •  Select tracks and create playlist

TrendngGambling on the current trends on Twitter

Florian Golemo Marjeta Markovic Kevin Wezeman

19-03-2015

Idea

Twitter Trends + Stock Market + Game

(think binary option trading)

10 twitter trends

10 twitter trends

60s to place a bet

42$

10 twitter trends

60s to place a bet

then cashout& next round

Idea - Incentive (Moneh moneh)

● now: just virtual money

● future: bitcoins?

Datasources

● Twitter:○ REST: Top 10 worldwide trends every 5min

○ Streaming: all tweets for those 10 hashtags

● What The Trend:○ Description for trends

The Interaction

● Every minute 1 bet

● Up/down

● Higher payout for higher risk

● Tweet link for each hashtag

The Implementation

● Server-side:○ NodeJS

○ Socket.io

○ Twitter OAuth

● Client-side:○ AngularJS

Extra Ideas

● add map, chose regional trends

Who does what

● The game: Florian & Marjeta

● The report: Marjeta & Kevin

Team 16

Erik Lubbers

Christian Heymans

Juan Manuel Bedregal

FAN FAVORITE

ACTORS

The Social Web - 2015

IDEA

• Present relevant general and social information related to a specific actor‟s and

his/her movies

• Genre

• Year

• Rating

• Likes

• People Talking

about them

TARGET USER

• Anyone interested in viewing the social

information related to the movies of a specific

actor

• Anyone interested in viewing a movie from a

specific actor.

USE

• Navigate through the movies of a specific

actor, explore movies with similar:

rating/point in time/likes/people talking

about

• The information and visualizations provided

by the application, can be used to help the

user choose a movie from an actor

DATA SOURCES

• ranker.com API

• (top 250 voted actors)

• dbpedia.org API

• (actor information and movie list)

• facebook API

• (likes and people talking about it)

• omdb API

• (IMDB rating and genre of the movies)

HOW IT WORKS 1/2

1. Retrieve the actor from our Database and list it on the applications searchbox, ranker.com

2. Search for the actor in dbpedia and retrieve:

• photo

• summary

• movie list

3. Query Facebook

• query movies => Facebook return list of pages

• Clean and Filter Data: search for „movie‟ category, query Facebook for more information, retrieve actor list and compare

• get likes and people talking about the movies

HOW IT WORKS 2/2

4. Query OMDB

• query movies => IMDB return list of movies

• Clean and Filter Data: search list of movies for actor list and compare

• get first genre and rating

5. Create a new dataset with the summarized information:

• Movie rating X Years X Genre

• Genre X Rating X Number of movies

• Movies X Likes X People talking X Rating X Genre

6. Draw the visualizations

find-it.nl/ffactors

LIMITATIONS

• Inconsistency of the data

• Different spelling of Movies

• Incomplete information

• Missing: category, cast info, additional

information

• This affected the overall reliability of the

results

INDIVIDUAL WORK

• CHRISTIAN

• Overall Design

• Styling of the application

• Integration and coding of the Facebook and OMDB api‟s

• ERIK

• Overall Design

• Finding and integrating the ranker actor list

• Integration and coding of the DBPedia querying

• Tuning the complete project to work as a whole

• JUAN

• Overall Design

• Prepare the summarized data for the Visualizations

• Design and Code the Visualizations

CONCLUSIONS

• It is feasible to gather social Information from different sources, and analyze in

order to create a bigger understanding of a particular subject.

• Important to define first the “problem to solve” and then search for the Data

Sources.

• Difficulty to Mine information from Social API‟s

• Inconsistency and Incompleteness of Information

THANK YOU

QUESTIONS ?

visit us @:

find-it.nl/ffactors

#Tweebay

Group 17: Wojciech Sidor, Janusz Kukla, Elinesofie Dolhain

Added value

• Connect buyers / sellers on Twitter • Opportunity to compare via eBay

•  Search by ratio

2

Demo

3

Opportunities

• Number of tweets with geolocation ▫  # Tweebay ▫  Other ways to find location on tweets

•  Streaming instead of search API •  Search by date / period of time •  SPAM filter

4

Individual slide (1) Wojciech

•  Free eBay

• Geodata – not only a threat

•  Twitter is omnipotent

5

Individual slide (2) Elinesofie

• Connect users via geolocation (real-time update) for selling / buying products

• Easy to sell home-made products

• #Tweebay as evaluation

•  Improvements on user evaluation

6

Individual slide (3) Janusz

•  Pre-attentive human analysis of the data

• Harder to draw any conclusions from the raw data

•  The possibility to add more data in order to help users to get in touch makes a new type of society – Internet society

7

Group  18  

Meet

In

Middle

Alsjeblieft

Social Web 2015

Anthony Nwosisi

Aron de Vries

Roberto Floris

Group 18

19 March 2015

Application Introduction

• Humans are social beings

• Socializing involves meeting

• Meeting could be very hectic

• Friends, family or colleagues might be reluctant to travel

• MIMA proposes a solution – the Dutch Solution

• MIMA is based on shared distance between two points

• The application is using the Google Maps Javascript API

Application works with: • Full address• A postcode• City• Country• Objects

• Two locations: Amsterdam and Alkmaar• App calculates distance using a straight line (Purple

in the Diagram)• The App also depicts the road landmark between

the two points (the blue line)• The middle point is the point with the bouncing

object.• Makes a circle at the middle with a radius of 1

kilometer

http://garagejackbakker.nl/sw

Read A Movie

A Book & Movie Recommendation Application

Image from : http://www.fanpop.com/clubs/reading/images/27819134/title/read-book-photo Group 19

Idea

• Recommend books based on the movies and books that users and their Facebook friends like + Goodreads ratings

• Recommend movies based on the movies and books that users and their Fecebook friends like + Goodreads ratings

Motivation

• Movies are “fast” (we can see more films than read books):o “Cold start”

o Broader exploration - easier to identify new fields of interest

• Books are “deep”o Time investment is greater and people choose carefully what to read

o There are good films without stars and famous directors, there are hardly any good films without a good story

• Naturally connected

Many movies are based on books and many screenwriters write books

Resources

• Facebook API

Likes about movies and books

• Goodreads API, Listopia

Book ratings

• imdbapi

Genre and screenwriter properties

Under Construction Issues

• Privacy - not possible to harvest friends likes without Facebook’s explicit permission

• Incompatibility - Goodreads API works only with XML format (not even DOM)

• Listopia - blocked

• Only one way queries allowed genre => author

Data Analysis

Book Ratings Similarity

Demo

https://www.youtube.com/watch?v=SgLTt2V2kcg&feature=youtu.be

Who did what

• Sander – development

• Aneta – concept and research

• Sergio – XML to JSON parser development

1

Company Proprietary and Confidential Copyright Info Goes Here Just Like This

Hay fever application

Group 20Sietse HuismanDavid Lopez MejiaGert‐Jan de Graaf

Vrije Universiteit – the social web

2

Company Proprietary and Confidential Copyright Info Goes Here Just Like This

• Extract tweets related to hay fever

• Convert number of tweets to heat map of the Netherlands

Issues

• 1% of tweets has geo location available

• Number of tweets is scarce

Initial idea

3

Company Proprietary and Confidential Copyright Info Goes Here Just Like This

Improved idea - HooikoortsBot

• Flow– Extract hay fever tweets with key word extraction– Use sentiment analysis

• Positive• Negative

– Tweet back to the person “Solution / Preparation”– Data visualization

4

Company Proprietary and Confidential Copyright Info Goes Here Just Like This

Data analysis – key word extraction

• Alchemyapi

5

Company Proprietary and Confidential Copyright Info Goes Here Just Like This

Data analysis - Sentiment Analysis

Positive sentiment

Negative sentiment

6

Company Proprietary and Confidential Copyright Info Goes Here Just Like This

Presenting the HooikoortsBot!

7

Company Proprietary and Confidential Copyright Info Goes Here Just Like This

Example

8

Company Proprietary and Confidential Copyright Info Goes Here Just Like This

Visualization

Sentiment of tweets

Positive Neutral Negative

02468

101214

runny nose itchy eyes dry throat clogged ears groggyness

Symptoms described by users with sentiment analysis

Positive Neurtal Negative

9

Company Proprietary and Confidential Copyright Info Goes Here Just Like This

Future work

• Heat map• Connect twitter users with similar

symptoms

Linked movie location

Group 21:Fleur, Elmar & Arnold

Planning● Introduction - the initial idea● Our application

○ The idea○ Screenshots

● Acknowledgements● Conclusion● Questions

Introduction - the initial idea● Movies from the Facebook API

○ Description & Actor from DBpedia

● Maybe add:○ Recommendations based on genre + director○ Film location from Linkedmdb plotted on Google

maps with Sgvizler

Our application - the idea● Lookup actor information

○ DBpedia + Sgvizler on HTML/PHP

● Film information from Assignment 3 is used

● Login with facebook

AcknowledgementsArnold: Presentation, Application, DocumentationElmar: Documentation, ApplicationFleur: Presentation, Documentation

Conclusion● Linkedmdb is buggy and the right

information was difficult to retrieve.

● We are not developers, so building an application was hard.

MovieVisVisualization of opinions and sentiment of movie reviews

Group 22: Adriatik BedjetiBoris de GrootEdgar Weidema

MovieVis: Why?!Motivation● Movie ratings are not

enough● Movie reviews are long

and countless● Movie reviews contains

spoilers

Goal● Analyse sentiment in

movie reviews● Analyse movies reviews

for subjective terms that describe, but do not give away, the movie

● Get user opinions and compare them to public data

What data

Crawl user reviews from IMDb and Metacritic

Sentiment Analysis● Analyse sentiment in user

reviews

● Categorize sentiment

● Visualize the results to the user

Gather User Opinions

● Data-preprocessing: clearing the data

● Get the most used descriptive terms used

● Visualize the results to the user Can you guess this 2015 movie?

Individual Work● Adriatik: Project idea, visual design, theoretical

background● Boris: Sentiment analysis & User opinion programming● Edgar: IMDB Crawling, MetaCritic Crawling, Data

preprocessing

Duration of trending topics

Group 23Felicia Hotie

Mark van der LaanMarc Went

Introduction

Aim: Gain insight in topic trending between different social media platforms and a historic social media source

Find answers to quenstions like:Are trending topics of shorter relevance in comparison to 20 years ago?

Clustered by social media source

● The Digital City (1995)

● Reddit

● Twitter

Approach

● How it works:

- Retrieve posts from the 3 sources with a search query- Sort by date- Count frequency of posts per date- Visualise in graph

● Clustered by topic (query) and social media source

Interface

● Mine data using the Reddit API and Python● Mine data from DDS using a self created parser● Mine data from Twitter, first prototype is manual labor● Interface created with HTML5● Graph plotted with JavaScript

Screenshots

Evaluation / improvements

● Limitations:- Reddit API allows only 100 search results

● Future work:- Add more data to compare. E.g. compare word frequencies per topic and

source- Add more social media sources

Group  24  

RoasTV David

Carlo

Priscilla

The app

!  Its a forum that shows info of a TV show, where people can give their opinions about it

!  To see a discussion you must have an opinion toward the particular TV show

!  People get more points when they make more posts, and when they become more influencial

Data !  Create an RDF store using Tomcat

!  Storing all the stuctured used inthe application for future reference

!  Data sources !  DBpedia

!  Information about TV show for reference

!  Twitter ! Positive and negative tweets related to the show

Analysis !  Co-occurrence of words on posts

!  Influences Patterns !  Track users attitude towards the different shows !  Track general opinion and discussion intensity !  Track general opinion on Twitter periodically

!  *Possible in another social network

Visualization

!  Visualize influence related to others !  Array which shows position towards each TV show by

the user --- compare it with general position array, and twitter array

!  Array with frequent words ---- compared with general frequent

!  Track appearance of new words

Visualization

Group 26 Recommendation App

Emmanouil Pavlidakis

eps780

Jaideep Khandelwal

jkl650

Andreas Manios

ams620

Course: Social Web 2015

Group : 26

1

What is our app? It is a Facebook Application.

Its purpose to provide personalized recommendations for movies based on the similarity score.

Differences between our application and other movie recommendation sites.

The rating for each movie is unique for a particular user and that ratingwill be based on his preferences and his personal profile.

The recommended movies will be a selection of movies based on the similarity score of two or more users.

The user can register and login with his Facebook account.

At the registration the app asks for permissions and it imports some personal information and the movies he has liked.

After the registration the user is asked to select genre/s .

Based on the genre he/she has selected and the movies from FB, a list of movies is displayed to him. This list contains :

1. Movies that the user likes in his FB account.

2. The most popular and highly rated movies that belong to the genres that he/she has selected.

Finally, the user is asked to rate this list of movies.

2

What is the similarity score? From these information (movies ratings) a personal profile has been created for each user.

By the use of it the App is able to calculate the similarity score between users.

After the registration and the creation of the Personal profile, the user is always able to rate new movies that he/she sees in his timeline.

As a result the similarity score becomes more accurate.

The similarity score is a percentage that displays, in which degree the user that has logged-in, has similar or identical preferences in movies with

other users.

It is calculated based on the movies that users has rated.

And it will be different for each user or for a pair of users .

E.g. if user_a has similarity score 80% with user_b, user_b may have 60% similarity score with user_a.

In the case that the similarity score of the logged-in user is 0 (worst case), the system recommends to that user the most popular movies based

on the selection of genre.

As a result the user has more involvement with the system.

3

Social Aspect The logged-in user :

1. Sees a list of the users based on high similarity score.

He/she is able to follow those users, see their profile and create one way relationship.

Also, the logged-in user is able to see the movies that these users has ratted.

2. Following users.

The logged-in user is able to follow other users.

As a result the user :

Can get notifications if these users rate a new movie.

Can send them messages.

Can ask them for recommendations.

3. Get Recommendations about new movies that other users has rated.

These recommendations will be based on the users that the logged-in user has high

similarity score.

4

APP OVERVIEW 5

Screen Casts of the App 1/2 6

Login and Registration by FB account

The user after the registration selects genres

The user is asked to rate movies based on the genres that he has selected

7Screen Casts of the App 2/2After the completion of the movie profile the user is able to get recommendations for movies.

See users with high

similarity score

Follow them and

view their ratings Get recommendations

based on the similarity

score

Change the rating

See the users that

he follows

Contact users that

he follows

Individual Work Andreas Manios

Front End

Emmanouil Pavlidakis

The theoretical part of the report and the creation of the graphs.

Jaideep Khandelwal

Back End

Links

Back end code : https://github.com/jdk2588/socialweb

Application Link : http://dessad.altervista.org/yars/main.html

8

COMPARING USAGE OF TAGS PER COUNTRY IN A SPECIFIC TIME PERIOD

Group 27: Annelore Franke, Daniel Gallo, Lars Rouvoet and Reza Mahmood KhalesiThe Social Web 2015

COMPARING USAGE OF TAGS PER COUNTRY IN A SPECIFIC TIME PERIOD

#happy

worldwide

period of time

1) Location Clustering:

Hashtag behaviour by

country

2) Hashtag behavior

worldwide over time analysis

#happy

#sad

The Netherlands

#galaxy

#iphone

#Bavaria

#Heineken

3) Comparisons between

hashtags by country

period of time

Web InterfaceUser

Twitter API

Web Framework

APPLICATION ARCHITECTURE

… and other sources

DEMO SCREENCAST NO.1Hashtag behaviour by country:Web Interface

DEMO SCREENCAST NO.1Hashtag behaviour by country:Results of #happy visualizedby country

DEMO SCREENCAST NO.2Comparison of 2 hashtags worldwide:Web Interface

DEMO SCREENCAST NO.2Comparison of 2 hashtags worldwide:Results #bavaria and #heineken worldwide

Bar chart Line chart

DEMO SCREENCAST NO.3Comparison of 2 hashtags in a specific country:Web Interface

DEMO SCREENCAST NO.3Comparison of 2 hashtags in a specific country:Results of ‘happy’ and ‘sad’ in The Netherlands

Bar chart Line chart

THANK YOU FOR LISTENINGARE THERE ANY QUESTIONS?

Group 27: Annelore Franke, Daniel Gallo, Lars Rouvoet and Reza Mahmood KhalesiThe Social Web 2015

Social Web @VU���2015

Final Student PresentationsLecturer: Lora Aroyo

The Network InstituteVU University Amsterdam