ainoqualia

[ainoqualia]

Campaign diffusion: The case of Manos Sergios

Piraeus Bank created the campaign “Greece by your side” featuring the key persona of Manos Sergios. A fictional football player who was summoned by the National teams’ manager to join the team in the FIFA World Cup 2014 Brazil. Everyone was talking, searching and writing about him. The campaign has been very successful.

YouTube video snapshot

Qualia’s goal was to maximize the spread of the campaign in the social networks by computing a small group of influential opinion leaders that would accelerate the adoption of the initial rumour. We started with a seed list of journalists, athletes and celebrities who frequently post about sports on Twitter. We enriched the initial set, in an iterative manner. In each iteration candidates were re-ranked according to their importance in the community, that was measured as a function of followers, influence and network centrality. We filtered the candidates that were strongly biased in favor of a specific Greek football club. Finally, we applied sentiment analysis in order to filter candidates that expressed negative emotions towards our national team. Twitter reach: 70% of the Greek domain until the rumour was revealed. Exposure = 3,43 messages per user. We have practically covered the active Greek domain 3,43 times.

Campaign diffusion

YouTube video views

YouTube video engagement on Facebook

[ainoqualia]

Greek users sharing video links

Video sharing is one of the primary user activities on Facebook. People share inspirational and funny videos, music and sports videos. And since YouTube is the number one video sharing site in the world, people create tons of links from Facebook to YouTube videos.

Which YouTube videos have Greek users shared on Facebook, in 2013? We have calculated the relative percentages of selected performers in different domains.

Greek singers

Classical music composers

Pop music singers

Rock music legends

Highest paid models

Sports teams

8 o’clock evening news presenters

TV program presenters

Actors

Seasons of the year

Should I stay or should I go ?

[ainoqualia]

Social media emotion analysis for Athens Biennale 2013

Happiness is not just a function of personal experience. It is also the property of groups. Emotions are a collective phenomenon, as is sentiment, i.e. the expression of emotions, opinions, and attitudes. Other peoples’ happiness affects our own. If someone smiles at you, you instinctively smile back. Often our emotions are more intelligent than we are.  And perhaps because the unconscious brain has the capacity to process enormous amounts of information at lightning speed, emotions seem to have the power of foresight.  

With this work, we wished to calculate the collective emotions/sentiments from a very large data sample, extracting information about how people in Greece feel, observing the way they express themselves in social media. Aided by a sound methodology involving the recognition of expressive means that describe emotions and emotional situations, and with the help of natural language processing techniques, we calculated 51 basic emotions/sentiments that are expressed by users of Greek social media within a 3-month period, January 1st to March 31st 2013. Among these are: insecurity, optimism, fear, dissatisfaction, enthusiasm, fatigue and expectation. Calculating the emotions/sentiments on a large data sample and using advanced technology, provides us with a profound insight into what a large number of the people in this country feel and impulsively express.

Emotion shares appear on the Athens Biennale 2013 page: http://athensbiennale.org/en/agora_en/ 

In the following graphs we present the timelines of three emotions, together with remarks on some of the observed peaks. The x-axis depicts the date while the y-axis depicts the percentage of posts containing an expression of the sentiment.

Optimism

A peak on Νew Υear’s day !

Love

A peak on Valentine’s day on the 14th of February. Otherwise love is always present !

Fear

A peak on the 15th of February, the International Childhood Cancer Day, accompanied by a variety of comments and shares on the subject. Also on the same day a meteor streaked across the sky above Russia's Ural Mountains. It was the biggest blast in a hundred years. It resulted in a mass panic and UFO fears.

[ainoqualia]

TEDxAthens 2011 insights

The TEDxAthens 2011 theme was “the Art of Disruption”. The conference took place in Athens, on  December 3rd 2011 with great success. Together with the TEDxAthens team and the Civitas team we have analyzed the 9323 social media references to TEDxAthens 2011 covering the entire week of the event.

The analysis has produced a variety of insights, such as:

Calculation of buzz; segmentation of the conference to time periods corresponding to the speakers; detection of the top quotes for each user, as well as the top quotes in general, as these were tweeted or re-tweeded by the audience; computation of the links to web sites, blogs and photo sharing sites that were shared by users; identification of the most popular nodes of the network that was formed around the event; and computation of the aspects of the discussion, that is the key-words, phrases and terms with the most discriminative capacity and which better characterized the discussion about TEDxAthens.

We have linguistically processed the “About” fields that are included in the profiles of the 100 most important nodes of the network that was developed around the event and we have created the following word cloud.

A beautiful video of the insights that characterized the event can be found here: http://blog.tedxathens.com/tedx-athens-2011-social-media-video-infographic  

The report can be found here: http://dl.dropbox.com/u/153498/TEDxAth/2011/TEDx%20Athens%20social%20media%20buzz%20analysis.pdf

Our warmest thanks to the TEDx Athens team and the Civitas team for this wonderful collaboration ! 

[ainoqualia]

Twitter engaged community: sharing links to social media

This is the last post in the series on community engagement. 

People use Twitter to share links. In June 2011 almost 200 million tweets were being sent per day and 25% of these tweets contained links. It seems that content sharing is a secret to Twitter success. 

The theme of our last infographic is sharing links to social media. Thank you for your comments, remarks and ideas. There is more to come !

[ainoqualia]

Twitter engaged community: sharing links to world media sites

Which are the most popular world media sites? And how many unique users post links to those sites?

[ainoqualia]

Twitter engaged community: sharing links to Greek media sites

Which are the most popular media sites? And how many unique users post links to those sites?

[ainoqualia]

Twitter engaged community: sharing links to lifestyle sites

Which are the most popular lifestyle sites? And how many unique users post links to those sites?

[ainoqualia]

Twitter engaged community: sharing links to sport sites

Which are the most popular sport sites? And how many unique users post links to those sites?

[ainoqualia]

Twitter engaged community: sharing links to government sites

Which are the most popular government sites? And how many unique users post links to those sites?

[ainoqualia]

Twitter engaged community: sharing links to sites of political parties

People often tweet links to content on web pages. Which are the most popular pages? And how many unique users post links to those pages?

We have computed the number of unique Greek users that post links to web sites and we have ranked the sites accordingly. The engaged community of a web site is a strong indicator of the site's impact. This is part of the deeper analysis that Qualia performs on the characteristics of such communities.

Period: 01 January - 15 September 2011.

[ainoqualia]

Opinion Aspects: the case of President Barack Obama

July 2011 has been a difficult month for President Obama and his administration, with many crucial issues running in parallel. In what concerns Greece, the President has welcomed the important steps taken to stabilize the euro zone under the new Greek debt deal. Also, regarding the american debt crisis, Barack Obama declared on July 15 that “we are not Greece”. Moreover, Hillary Clinton, the US Secretary of State, expressed her support for Greece during a visit to Athens, saying: "We stand by the people and the government of Greece". We could also mention the ads that appeared at various stations and bus stops in Washington, D.C. that discredited Greece with the slogan “Next Stop: Greece. Raising the debt without cutting spending is our own Greek Tragedy”.

What is the opinion on Barack Obama among Greek bloggers? Opinion analysis is a new feature included in the latest release of aino. We have analyzed opinion for President Obama on Greek blogs, for the period July 1 – July 28, 2011. We start our presentation of the findings by showing the number of mentions on a daily basis:

Here is in short the procedure that we have implemented in order to compute opinion within texts. We start by creating the entity “Barack Obama” and searching for its mentions in very large corpora from Greek news sites and blogs. Next, we train our model in order to form a conceptual cloud around the entity that will make possible the linking between, for instance, Barack Obama, President Obama and the President of the U.S. We have manually created sentiment resources in Greek and we match our dictionaries against the input documents. Next, we apply natural language processing techniques in order to compute whether sentiment expressions actually refer to the entity “Barack Obama”. Once we find associations, we apply rules such as negation in order to determine the intensity and polarity of the expression. The next diagram depicts the opinion over time on President Obama, in the range [-1…+1] (very negative to very positive).

We can see that opinion on President Obama in Greek blogs has been negative during July 2011. The main events that contributed to the negative opinion were:

On July 8, the President blamed Greece, Japan, high gas prices, uncertainty over the debt-ceiling issue, and natural disasters across the globe for the rising US unemployment.

On July 12, Le Monde in its editorial with the title “Ces gamins qui nous gouvernent” (those kids that govern us) declared: la situation requiert des adultes - et on en manque, à Washington comme à Bruxelles http://www.lemonde.fr/idees/article/2011/07/12/ces-gamins-qui-nous-gouvernent_1547754_3232.html. The editorial was extensively reproduced in Greek on-line sources.

On July 15, the President said the US is not Greece. During the same period appeared the offensive campaign against Greece: Next Stop: Greece, Don’t raise the debt ceiling without cutting spending.

On July 22, the US debt limit talks failed.

The only positive event occurred on July 18, when the US Secretary of the State Hillary Clinton visited Greece and received positive media comments, as she declared the support of Washington in the efforts made by the Greek government to tackle the economic crisis.

The distribution of opinion in ranges of values is as follows:

 The majority of values fit in the range [-0,20…0]. Very positive values are associated with Mrs. Clinton’s visit in Athens and the final flight of NASA's space shuttle program, an event covered by Greek media.

There are two more dimensions of opinion analysis that are computed by aino: polarity and subjectivity. Positive and negative affect theories are based on the idea that positive and negative opinion should be separately tracked because they vary independently. An aggregate neutral opinion can be the result of equal antithetical opinions of low or high polarity. The evolution of polarity over time appears in the next diagram:

Some high polarity points are associated with the following events:

On July 4 a Fox news hacker announced that Obama is dead.

On July 8 the President blamed Greece for the spike in the US unemployment figures.

On July 23 several events took place: the Norway attack and the war on terrorism, House Speaker John Boehner’s decision to end debt talks with President Barack Obama, and an article signed by Paul Krugman that appeared in the Times and was reproduced in Greek media, entitled “letting bankers walk” http://www.nytimes.com/2011/07/18/opinion/18krugman.html.

The computational processing of subjectivity within texts is also a feature of the opinion analysis family. The idea is to compute the amount of subjective expressions within a document.

On July 11, Mrs. Lagarde, the new chief of IMF foresees "real nasty consequences" if the U.S. fails to raise its borrowing limit http://www.huffingtonpost.com/2011/07/10/imf-us-borrowing-limit_n_894044.html. Subjectivity exhibits local maxima on this date because Mrs. Lagarde’s statement has ignited a discussion with subjective arguments and critics that dismissed Greece’s butterfly effect role in the global crisis, since the US debt crisis is order of magnitudes more severe than the Greek debt crisis.

Finally, July 23 (see above) seems to be a day with very negative opinion, high polarity and high subjectivity as well.

[ainoqualia]

Tracking the Norway attack

The attack on a Norwegian island and the massive explosion in the heart of Oslo shocked the world. The devastating incidents took place on Friday July 22, and quickly climbed to the top of the news. The impact of the event as well as its share-of-voice are very high.

Topic detection and tracking is a technology that organizes a stream of constantly arriving news stories by the events that they discuss. Aino has detected the new event by using an advanced clustering algorithm that automatically creates a bundle, that is a multimedia topic aggregating information atoms from real-time multimedia streams regarding the new event. A topic is defined to be a seminal event or activity, along with all directly related events and activities.

The next picture taken from aino presents the details of the bundle by showing its summary, its term cloud and its mentions from the multimedia information sources that are followed by aino: mainstream news, blogs, twitter and broadcasts.

Major Greek TV channels interrupted normal programming with a breaking news report on the event. Aino applies automatic speech recognition in broadcast news in order to identify the mentions regarding the event. Moreover, aino applies video text recognition on overlay text and retrieves relevant video segments by searching the multimedia index. This methodology significantly improves the precision of the retrieval procedure.

Bloggers and twitters almost immediately covered the attacks.

  Once we have detected the topic and we have created its corresponding topic model, our aim is to map incoming information atoms to the predefined topic, so that we can track its evolution over time. Aino monitors the stream of arriving news stories and picks out those that discuss the Norway attacks. The topic model is updated with each new information atom that is assigned to the topic. The retrieval threshold can be defined by the user: if we drag the similarity button to the right, we track only the specific topic while if we drag it to the left, similar topics are retrieved as well.

[ainoqualia]

Buzz vs. Like, Malou vs. Tsakalakos, Greek Idol Final

Η μεγάλη ώρα έφτασε! Μαλού και Τσακαλάκος προετοιμάζονται για την τελευταία τους παρουσία στο μεγάλο τελικό του greek idol. Είχαν και οι δύο πρωταγωνιστές ιδιαίτερα ενεργή παρουσία στα social media, όπου και πραγματοποιήθηκε η συντριπτική πλειοψηφία των συζητήσεων για το σόου. Με την ευκαιρία λοιπόν αυτή, επιχειρούμε μέσα από το Facebook και το Twitter να μετρήσουμε τη δημοφιλία τους, το engagement των fans τους και να ανακηρύξουμε τελικά τον μεγάλο νικητή!

Μετρήσαμε το buzz (αναφορές) των δύο φιναλίστ συνολικά, αλλά και κατά τη διάρκεια της τελευταίας εβδομάδας. Η Μαλού υπερισχύει στο σύνολο με μεγάλη διαφορά του Τσακαλάκου, αλλά την τελευταία εβδομάδα η ψαλίδα δείχνει να κλείνει.

    Στη συνέχεια αναλύσαμε τα fan pages των δύο φιναλίστ και συγκρίναμε το engagement του κοινού τους μετρώντας τα expressions {expression: post, comment, like}. Ο Τσακαλάκος ξεκίνησε με περισσότερους fans από την αρχή του σόου και γρήγορα απέκτησε μεγάλο προβάδισμα από την Μαλού (και τους υπόλοιπους παίκτες πλην Πλασκασοβίτη). Συνέχισε να ανεβάζει τον αριθμό των fans του με μεγαλύτερο ρυθμό από την Μαλού μέχρι και την τελευταία εβδομάδα. Η Μαλού έχει οριακά περισσότερα expressions, αλλά οι unique users που έχουν κάνει expression είναι περισσότεροι για τον Τσακαλάκο (62% - 38%).

Μελετώντας το φύλο των fans του greek idol, βλέπουμε ότι συνολικά στο greek idol οι γυναίκες είναι περισσότερο ενεργές από τους άντρες. Αυτό φυσικά ισχύει και για καθέναν από τους φιναλίστ, με τον Τσακαλάκο να έχει βέβαια μεγαλύτερη διείσδυση στον γυναικείο πληθυσμό από ότι η Μαλού, κάτι που του δίνει ένα επιπλέον προβάδισμα.

  Συνεπώς οι δυο φιναλίστ μπαίνουν στον τελικό με την Μαλού να κυριαρχεί στο buzz και ειδικά στο Twitter και τον Τσακαλάκο να συγκεντρώνει πολύ μεγαλύτερο κοινό (fans και engaged users) στο Facebook. Οι μετρήσεις στο εξωτερικό δείχνουν ότι επικρατεί πάντοτε η επονομαζόμενη κατάρα του Twitter (“Twitter curse”)  και ο παίκτης με το μεγαλύτερο buzz χάνει. Η ψηφοφορία στην Ελλάδα θα γίνει με sms. Θα επιβεβαιωθεί το Twitter curse;

[ainoqualia]

social metrics comparison: #dwts2 vs. #greekidol2

«Dancing with the Stars 2» εναντίον «Greek Idol 2». Η κόντρα έχει ανάψει, μπαίνουμε στην τελική ευθεία των δύο τηλεοπτικών σόου και είναι μια καλή στιγμή για μια μικρή social σύγκριση!

Μετρήσαμε το buzz των σόου στα social media (Twitter, Facebook, YouTube, Google Buzz) και το αποτυπώσαμε στο επόμενο γράφημα. Βλέπουμε πως το dwts2 είναι ο κυρίαρχος! Το greekidol2 κατάφερε μόνο στο πρώτο του LIVE να ξεπεράσει τον μεγάλο του αντίπαλο, αλλά έκτοτε οι social επιδόσεις του δείχνουν ελαφρά πτωτικές.

  Μετρήσαμε με την βοήθεια του aino το social engagement και το influence των χρηστών που συμμετέχουν στην συζήτηση στα κοινωνικά δίκτυα Twitter και Facebook. Αξιοποιούμε για τον σκοπό αυτό όλα τα σχετικά expressions των χρηστών στα δίκτυα (tweet, post, comment, Like), την θέση του κάθε χρήστη ως κόμβου του συνολικού δικτύου, καθώς και τον βαθμό επιρροής του (influence) στους υπόλοιπους χρήστες και κάνουμε τις μετρήσεις. Ορισμένα ενδιαφέροντα στοιχεία της ανάλυσης παρουσιάζονται στη συνέχεια.

Engagement

Η ανάλυση των δεδομένων από όλα τα user expressions έδειξε μια υπεροχή του dwts2.

Είναι όμως σημαντικό να αναφερθεί ότι μεγάλο κομμάτι της συζήτησης στο Facebook για το greekidol2 δεν γίνεται στο επίσημο fan page, αλλά στο ανεπίσημο. Αν συμπεριλάβουμε τα δεδομένα και από αυτό, η εικόνα όπως φαίνεται στα γραφήματα αλλάζει και φαίνεται το greekidol2 να υπερέχει σε συνολικό αριθμό user expressions.

Παρότι όμως οι χρήστες του greekidol2 είναι πιο engaged, σε απόλυτο αριθμό χρηστών υπάρχει ελαφρά υπεροχή του dwts2.

  Κλείνουμε το post με την ανάλυση του γένους του κοινού των δύο σόου.

Δημογραφική σύγκριση

Μελετώντας το φύλο των engaged χρηστών στα δύο σόου στο Facebook η κυριαρχία του «ασθενούς» φύλου είναι ξεκάθαρη!

  Με τέτοια σύνθεση κοινού θα είναι δύσκολο για την Μαλού να φτάσει στην κορυφή. Φαινομενικά οι Πλασκασοβίτης και Τσακαλάκος έχουν το πάνω χέρι, αλλά... θα επανέλθουμε σε αυτό το θέμα με περισσότερη λεπτομέρεια σύντομα!

[ainoqualia]

Ανάλυση γνώμης για το φράκτη στον Έβρο

Στην αρχή του χρόνου η κυβέρνηση ανακοίνωσε την “εγκατάσταση αποτρεπτικών μέσων ανεξέλεγκτα εισερχομένων παράνομων μεταναστών κατά μήκος των 12,5 χιλιομέτρων των χερσαίων συνόρων στον Έβρο”. Η ανακοίνωση προκάλεσε ποικίλες αντιδράσεις, προβληματισμό και έντονη συζήτηση. Μελετήσαμε τις συζητήσεις στις online πηγές και υπολογίσαμε τα διάφορα ποιοτικά και ποσοτικά στοιχεία για τις γνώμες που εκφράστηκαν σε αυτές.

 Εξαγωγή γνώμης από διαδικτυακές πηγές

Συγκεντρώσαμε τα άρθρα και τα posts που αναφέρονταν στο θέμα του φράκτη από 500 sites (νέα, ειδήσεις, επίκαιρα, αρθρογραφία), 35.000 ελληνικά blogs και από Facebook, Twitter, Google Buzz. Τα κείμενα καλύπτουν το χρονικό διάστημα 23/12/10 - 13/02/11. Εντοπίσαμε με επεξεργασία της γλώσσας τα σημεία στα οποία εκφράζεται γνώμη για το θέμα και υπολογίσαμε αν η συνολική γνώμη που απορρέει από κάθε κείμενο είναι θετική ή αρνητική και κατά πόσο. Η ανάλυση περιλαμβάνει μορφολογική επεξεργασία, εκφράσεις υποκειμενικότητας (subjectivity) και συναισθήματος (sentiment), καθώς και ένα σύνολο από συντακτικούς κανόνες για την αναγνώριση φαινομένων της γλώσσας που παίζουν ρόλο στον υπολογισμό της γνώμης (πχ. άρνηση). Η συνολική γνώμη όπως προκύπτει από την ανάλυση, αποτυπώνεται στα παρακάτω διαγράμματα:

Τι φανερώνουν οι δηλώσεις για το θέμα

Στόχος μας ήταν ο εντοπισμός των δηλώσεων για το θέμα και η κατηγοριοποίησή τους βάσει της πόλωσης που αυτές οι δηλώσεις φέρουν. Ακολουθήσαμε την παρακάτω μεθοδολογία:

1. Αυτόματη εξαγωγή των δηλώσεων σχετικά με την κατασκευή του φράκτη.

2. Κατηγοριοποίηση αυτών σε θετικές και αρνητικές.

3. Εξαγωγή των σημαντικότερων λέξεων-κλειδιών που περιέχονται στις δηλώσεις.

4. Ανάλυση του υλικού. Για το σκοπό αυτό, δημιουργήσαμε το παρακάτω βίντεο. Ξεκινώντας από αριστερά παρακολουθούμε σημαντικές εκφράσεις-λέξεις που συνδέθηκαν με αρνητικές δηλώσεις για το θέμα. Προχωράμε στο κέντρο και έχουμε λέξεις ή απόψεις που μοιράστηκαν στις δύο κατηγορίες. Τέλος δεξιά εμφανίζονται λέξεις που συνδέθηκαν με θετικές ως προς την κατασκευή του φράκτη δηλώσεις (κυρίως δηλώσεις της κυβέρνησης και τμήματο�� της τοπικής κοινωνίας).

Σημείωση: το κομμάτι λέγεται Quiet Riot, είναι του Alex Dimou και ευχαριστούμε πολύ την KLIK Records (http://www.klikrecords.gr) για την ευγενική παραχώρηση.

Δημοσκοπήσεις

Οι δημοσκοπήσεις για τη δημιουργία του φράκτη που δημοσιεύθηκαν σε έντυπα μέσα, είχαν σημαντική απόκλιση σε σχέση με τις μετρήσεις που έγιναν στο online περιεχόμενο και αυτό παρουσιάζει ενδιαφέρον. Επιβεβαιώνει αρχικά πως η διαδικτυακή κοινότητα δεν έχει τα χαρακτηριστικά του σώματος των ερωτηθέντων. Επίσης το διαδίκτυο είναι διαδραστικό μέσο, με τους χρήστες να εκφράζουν άποψη ενώ οι δημοσκοπήσεις όχι, μιας και οι δυνατές απαντήσεις είναι συγκεκριμένες. Η τελευταία αυτή διαφορά ίσως εξηγεί γιατί ενώ ένα σημαντικό ποσοστό των ερωτηθέντων (59%) θεωρεί το φράκτη αναποτελεσματικό μέτρο, τελικά απαντά στο βασικό ερώτημα μάλλον συμφωνώ/συμφωνώ. Τα αποτελέσματα των δημοσκοπήσεων έχουν ως εξής:

[ainoqualia]

The dynamics of the “I do not pay” meme

The “I do not pay” movement slowly started when drivers refused to pay in advance toll charges for roads that were not yet built, but it seems that lately it has diffused to other public transportations services (such as the metro) as well. It is a very controversial and highly criticised movement and it has gained further publicity these last days because the Ministry of Transport is passing a bill that proposes a serious fine for violators. This element of behavior – a meme – has very interesting physical properties, as it propagates through our culture. The word meme comes from the ancient Greek word mimeme (μίμημα) that means imitate. And by the force of imitation, the “I do not pay” meme has gained significant momentum. We present here measurements for some of the meme’s physical properties. The measurements were taken on three corpora that cover the period from November 1, 2010 to February 13, 2011 and contain respectively web news from 500 sites, blog posts from 35K blogs and tweets mentioning the meme. After two “warnings” by the movement in mid-November and mid-December that quickly faded out, a third attempt to gain momentum in the new year has been successful. For three weeks it has built-up mass and has started to spread. It attained a new height around January 20 and slowed down again for a few days. It should have been clear by then that the “I do not pay” movement had a significant public image and what started as a local raising of toll cash-box barriers could replicate through the population as a generalized protest against financial measures. After a few days the meme took off, gained momentum and the effort to take measures in order to reverse its momentum will be orders of magnitude greater.