Seminar: Ausgewählte Themen des Web 2.0

Veranstalter: Prof. Dr. Andreas Hotho, Martin Becker, Alexander Dallmann, Lena Hettinger, Daniel Schlör und Daniel Zoller.

WueCampus-Kurs: TBA

Vorbesprechung: 16.10.2015 um 14:00 in B002 (Informatikgebäude)

Vorbesprechung

Bei der Vorbesprechung werden die Themen - so nicht schon vergeben - an die anwesenden Seminarteilnehmer verteilt. Grundsätzlich besteht Anwesenheitspflicht. Teilnehmer, die bereits ein Thema bearbeiten, aber ohne Absprache fehlen, verlieren ihren Anspruch auf ihr Thema. Die Literatur zu einem Thema besteht zumeist aus einem Buchkapitel oder einem Satz wissenschaftlicher Aufsätze (Paper). In der Vorbesprechung werden auch die Termine für die Vorträge festgelegt.

Ablauf des Seminars

Es handelt sich um ein Blockseminar, d.h. Vorträge werde nicht über das Semester verteilt, sondern in einigen wenigen Sitzungen gegen Ende der Vorlesungszeit gehalten. Jeder Votrag behandelt ein spezifisches Thema mit zugehöriger Literatur und erstreckt sich etwa über 30 Minuten. Zu jedem Vortrag muss eine Ausarbeitung, eine Zusammenfassung und ein Foliensatz erstellt werden.

Dannach setzt man sich in der Vorbereitungsphase mit dem gewählten Thema auseinander. Die gegebene Literatur wird aufarbeitet. Zuätzlich soll diese durch möglichst aktuelle Literatur ergänzt werden. Die Auswahl der weiteren Literatur trifft der Seminarteilnehmer letztendlich in Absprache mit dem Betreuer.

Nachdem die Literatur feststeht wird eine Ausarbeitung, eine einseitige Zusammenfassung und ein Foliensatz erstellt.

Zuletzt wird der Vortrag im Rahmen einer Blockveranstaltung gehalten. Bei jeder Blockveranstaltung gilt Anwesenheitspflicht.

Themen

Die Gundlage zu den Vorträgen bilden wissenschaftliche Aufsätze (Papers). Jedes hier aufgeführte Paper entspricht einem Thema.

themenrelevante Paper sind mit dem Tag thema versehen
die einzelnen Themen werden anhand der Tags thema:x, thema:y, etc. unterschieden

Sentiment Extraction from Tweets: Multilingual Challenges. Makrynioti, Nantia; Vassalos, Vasilis. In DaWaK, Bd. 9263 von Lecture Notes in Computer Science, S. Madria, T. Hara (Hrsg.), S. 136–148. Springer, 2015.
- [ BibTeX ]
- [ URL ]
@inproceedings{conf/dawak/MakryniotiV15, author = {Makrynioti, Nantia and Vassalos, Vasilis}, booktitle = {DaWaK}, crossref = {conf/dawak/2015}, editor = {Madria, Sanjay and Hara, Takahiro}, keywords = {multilingual}, pages = {136-148}, publisher = {Springer}, series = {Lecture Notes in Computer Science}, title = {Sentiment Extraction from Tweets: Multilingual Challenges.}, volume = 9263, year = 2015 }
Topological Properties and Temporal Dynamics of Place Networks in Urban Environments. Noulas, Anastasios; Shaw, Blake; Lambiotte, Renaud; Mascolo, Cecilia. In arXiv preprint arXiv:1502.07979. 2015.
- [ BibTeX ]
- [ URL ]
@article{noulas2015topological, author = {Noulas, Anastasios and Shaw, Blake and Lambiotte, Renaud and Mascolo, Cecilia}, journal = {arXiv preprint arXiv:1502.07979}, keywords = {thema}, title = {Topological Properties and Temporal Dynamics of Place Networks in Urban Environments}, year = 2015 }
DeepWalk: online learning of social representations. Perozzi, Bryan; Al-Rfou’, Rami; Skiena, Steven. In KDD, S. A. Macskassy, C. Perlich, J. Leskovec, W. W. 0010, R. Ghani (Hrsg.), S. 701–710. ACM, 2014.
- [ BibTeX ]
- [ URL ]
@inproceedings{conf/kdd/PerozziAS14, author = {Perozzi, Bryan and Al-Rfou', Rami and Skiena, Steven}, booktitle = {KDD}, crossref = {conf/kdd/2014}, editor = {Macskassy, Sofus A. and Perlich, Claudia and Leskovec, Jure and 0010, Wei Wang and Ghani, Rayid}, keywords = {web20}, pages = {701-710}, publisher = {ACM}, title = {DeepWalk: online learning of social representations.}, year = 2014 }
Mining the web to predict future events. Radinsky, Kira; Horvitz, Eric. In WSDM, S. Leonardi, A. Panconesi, P. Ferragina, A. Gionis (Hrsg.), S. 255–264. ACM, 2013.
- [ BibTeX ]
@inproceedings{radinsky2013mining, author = {Radinsky, Kira and Horvitz, Eric}, booktitle = {WSDM}, crossref = {conf/wsdm/2013}, editor = {Leonardi, Stefano and Panconesi, Alessandro and Ferragina, Paolo and Gionis, Aristides}, keywords = {events}, pages = {255-264}, publisher = {ACM}, title = {Mining the web to predict future events}, year = 2013 }
Personality and patterns of Facebook usage. Bachrach, Yoram; Kosinski, Michal; Graepel, Thore; Kohli, Pushmeet; Stillwell, David. In WebSci, N. S. Contractor, B. Uzzi, M. W. Macy, W. Nejdl (Hrsg.), S. 24–32. ACM, 2012.
- [ Abstract ]
- [ BibTeX ]
We show how users’ activity on Facebook relates to their personality, as measured by the standard Five Factor Model. Our dataset consists of the personality profiles and Facebook profile data of 180,000 users. We examine correlations between users’ personality and the properties of their Facebook profiles such as the size and density of their friendship net- work, number uploaded photos, number of events attended, number of group memberships, and number of times user has been tagged in photos. Our results show significant rela- tionships between personality traits and various features of Facebook profiles. We then show how multivariate regres- sion allows prediction of the personality traits of an individual user given their Facebook profile. The best accuracy of such predictions is achieved for Extraversion and Neuroticism, the lowest accuracy is obtained for Agreeableness, with Openness and Conscientiousness lying in the middle.

@inproceedings{bachrach2012personality, abstract = {We show how users’ activity on Facebook relates to their personality, as measured by the standard Five Factor Model. Our dataset consists of the personality profiles and Facebook profile data of 180,000 users. We examine correlations between users’ personality and the properties of their Facebook profiles such as the size and density of their friendship net- work, number uploaded photos, number of events attended, number of group memberships, and number of times user has been tagged in photos. Our results show significant rela- tionships between personality traits and various features of Facebook profiles. We then show how multivariate regres- sion allows prediction of the personality traits of an individual user given their Facebook profile. The best accuracy of such predictions is achieved for Extraversion and Neuroticism, the lowest accuracy is obtained for Agreeableness, with Openness and Conscientiousness lying in the middle.}, author = {Bachrach, Yoram and Kosinski, Michal and Graepel, Thore and Kohli, Pushmeet and Stillwell, David}, booktitle = {WebSci}, crossref = {conf/websci/2012}, editor = {Contractor, Noshir S. and Uzzi, Brian and Macy, Michael W. and Nejdl, Wolfgang}, keywords = {five}, pages = {24-32}, publisher = {ACM}, title = {Personality and patterns of Facebook usage}, year = 2012 }
Evolution of Experts in Question Answering Communities. Pal, Aditya; Chang, Shuo; Konstan, Joseph A. In ICWSM, J. G. Breslin, N. B. Ellison, J. G. Shanahan, Z. Tufekci (Hrsg.). The AAAI Press, 2012.
- [ BibTeX ]
- [ URL ]
@inproceedings{pal2012evolution, author = {Pal, Aditya and Chang, Shuo and Konstan, Joseph A.}, booktitle = {ICWSM}, crossref = {conf/icwsm/2012}, editor = {Breslin, John G. and Ellison, Nicole B. and Shanahan, James G. and Tufekci, Zeynep}, keywords = {qa}, publisher = {The AAAI Press}, title = {Evolution of Experts in Question Answering Communities}, year = 2012 }
Friendship and mobility: user movement in location-based social networks. Cho, Eunjoon; Myers, Seth A; Leskovec, Jure. In International Conference on Knowledge Discovery and Data Mining (SIGKDD), C. Apte, J. Ghosh, P. Smyth (Hrsg.), S. 1082–1090. ACM, 2011.
- [ BibTeX ]
- [ URL ]
@inproceedings{cho2011friendship, author = {Cho, Eunjoon and Myers, Seth A and Leskovec, Jure}, booktitle = {International Conference on Knowledge Discovery and Data Mining (SIGKDD)}, editor = {Apte, Chid and Ghosh, Joydeep and Smyth, Padhraic}, keywords = {citedby:scholar:timestamp:2017-12-19}, organization = {ACM}, pages = {1082–1090}, publisher = {ACM}, title = {Friendship and mobility: user movement in location-based social networks}, year = 2011 }
Supervised random walks: predicting and recommending links in social networks. Backstrom, Lars; Leskovec, Jure. In Proceedings of the fourth ACM international conference on Web search and data mining, von WSDM ’11, S. 635–644. ACM, Hong Kong, China, 2011.
- [ Abstract ]
- [ BibTeX ]
- [ URL ]
Predicting the occurrence of links is a fundamental problem in networks. In the link prediction problem we are given a snapshot of a network and would like to infer which interactions among existing members are likely to occur in the near future or which existing interactions are we missing. Although this problem has been extensively studied, the challenge of how to effectively combine the information from the network structure with rich node and edge attribute data remains largely open. We develop an algorithm based on Supervised Random Walks that naturally combines the information from the network structure with node and edge level attributes. We achieve this by using these attributes to guide a random walk on the graph. We formulate a supervised learning task where the goal is to learn a function that assigns strengths to edges in the network such that a random walker is more likely to visit the nodes to which new links will be created in the future. We develop an efficient training algorithm to directly learn the edge strength estimation function. Our experiments on the Facebook social graph and large collaboration networks show that our approach outperforms state-of-the-art unsupervised approaches as well as approaches that are based on feature extraction.

@inproceedings{Backstrom:2011:SRW:1935826.1935914, abstract = {Predicting the occurrence of links is a fundamental problem in networks. In the link prediction problem we are given a snapshot of a network and would like to infer which interactions among existing members are likely to occur in the near future or which existing interactions are we missing. Although this problem has been extensively studied, the challenge of how to effectively combine the information from the network structure with rich node and edge attribute data remains largely open. We develop an algorithm based on Supervised Random Walks that naturally combines the information from the network structure with node and edge level attributes. We achieve this by using these attributes to guide a random walk on the graph. We formulate a supervised learning task where the goal is to learn a function that assigns strengths to edges in the network such that a random walker is more likely to visit the nodes to which new links will be created in the future. We develop an efficient training algorithm to directly learn the edge strength estimation function. Our experiments on the Facebook social graph and large collaboration networks show that our approach outperforms state-of-the-art unsupervised approaches as well as approaches that are based on feature extraction.}, address = {New York, NY, USA}, author = {Backstrom, Lars and Leskovec, Jure}, booktitle = {Proceedings of the fourth ACM international conference on Web search and data mining}, keywords = {web20}, pages = {635–644}, publisher = {ACM}, series = {WSDM '11}, title = {Supervised random walks: predicting and recommending links in social networks}, year = 2011 }
Folks in Folksonomies: social link prediction from shared metadata. Schifanella, Rossano; Barrat, Alain; Cattuto, Ciro; Markines, Benjamin; Menczer, Filippo. In Proceedings of the third ACM international conference on Web search and data mining, von WSDM ’10, S. 271–280. ACM, New York, New York, USA, 2010.
- [ Abstract ]
- [ BibTeX ]
- [ URL ]
Web 2.0 applications have attracted a considerable amount of attention because their open-ended nature allows users to create lightweight semantic scaffolding to organize and share content. To date, the interplay of the social and semantic components of social media has been only partially explored. Here we focus on Flickr and Last.fm, two social media systems in which we can relate the tagging activity of the users with an explicit representation of their social network. We show that a substantial level of local lexical and topical alignment is observable among users who lie close to each other in the social network. We introduce a null model that preserves user activity while removing local correlations, allowing us to disentangle the actual local alignment between users from statistical effects due to the assortative mixing of user activity and centrality in the social network. This analysis suggests that users with similar topical interests are more likely to be friends, and therefore semantic similarity measures among users based solely on their annotation metadata should be predictive of social links. We test this hypothesis on the Last.fm data set, confirming that the social network constructed from semantic similarity captures actual friendship more accurately than Last.fm's suggestions based on listening patterns.

@inproceedings{folks-rossano, abstract = {{Web 2.0 applications have attracted a considerable amount of attention because their open-ended nature allows users to create lightweight semantic scaffolding to organize and share content. To date, the interplay of the social and semantic components of social media has been only partially explored. Here we focus on Flickr and Last.fm, two social media systems in which we can relate the tagging activity of the users with an explicit representation of their social network. We show that a substantial level of local lexical and topical alignment is observable among users who lie close to each other in the social network. We introduce a null model that preserves user activity while removing local correlations, allowing us to disentangle the actual local alignment between users from statistical effects due to the assortative mixing of user activity and centrality in the social network. This analysis suggests that users with similar topical interests are more likely to be friends, and therefore semantic similarity measures among users based solely on their annotation metadata should be predictive of social links. We test this hypothesis on the Last.fm data set, confirming that the social network constructed from semantic similarity captures actual friendship more accurately than Last.fm's suggestions based on listening patterns.}}, address = {New York, NY, USA}, author = {Schifanella, Rossano and Barrat, Alain and Cattuto, Ciro and Markines, Benjamin and Menczer, Filippo}, booktitle = {Proceedings of the third ACM international conference on Web search and data mining}, keywords = {web20}, month = {03}, pages = {271–280}, publisher = {ACM}, series = {WSDM '10}, title = {{Folks in Folksonomies: social link prediction from shared metadata}}, year = 2010 }
Predicting the Future with Social Media. Asur, Sitaram; Huberman, Bernardo A. In CoRR. 2010.
- [ Abstract ]
- [ BibTeX ]
- [ URL ]
In recent years, social media has become ubiquitous and important for socialnetworking and content sharing. And yet, the content that is generated fromthese websites remains largely untapped. In this paper, we demonstrate howsocial media content can be used to predict real-world outcomes. In particular,we use the chatter from Twitter.com to forecast box-office revenues for movies.We show that a simple model built from the rate at which tweets are createdabout particular topics can outperform market-based predictors. We furtherdemonstrate how sentiments extracted from Twitter can be further utilized toimprove the forecasting power of social media.

@article{asur2010predicting, abstract = {In recent years, social media has become ubiquitous and important for socialnetworking and content sharing. And yet, the content that is generated fromthese websites remains largely untapped. In this paper, we demonstrate howsocial media content can be used to predict real-world outcomes. In particular,we use the chatter from Twitter.com to forecast box-office revenues for movies.We show that a simple model built from the rate at which tweets are createdabout particular topics can outperform market-based predictors. We furtherdemonstrate how sentiments extracted from Twitter can be further utilized toimprove the forecasting power of social media.}, author = {Asur, Sitaram and Huberman, Bernardo A.}, journal = {CoRR}, keywords = {media}, note = {cite arxiv:1003.5699}, title = {Predicting the Future with Social Media}, year = 2010 }
Labeled LDA: A Supervised Topic Model for Credit Attribution in Multi-labeled Corpora. Ramage, Daniel; Hall, David; Nallapati, Ramesh; Manning, Christopher D. In Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing: Volume 1 - Volume 1, von EMNLP ’09, S. 248–256. Association for Computational Linguistics, Singapore, 2009.
- [ Abstract ]
- [ BibTeX ]
- [ URL ]
A significant portion of the world's text is tagged by readers on social bookmarking websites. Credit attribution is an inherent problem in these corpora because most pages have multiple tags, but the tags do not always apply with equal specificity across the whole document. Solving the credit attribution problem requires associating each word in a document with the most appropriate tags and vice versa. This paper introduces Labeled LDA, a topic model that constrains Latent Dirichlet Allocation by defining a one-to-one correspondence between LDA's latent topics and user tags. This allows Labeled LDA to directly learn word-tag correspondences. We demonstrate Labeled LDA's improved expressiveness over traditional LDA with visualizations of a corpus of tagged web pages from del.icio.us. Labeled LDA outperforms SVMs by more than 3 to 1 when extracting tag-specific document snippets. As a multi-label text classifier, our model is competitive with a discriminative baseline on a variety of datasets.

@inproceedings{ramage2009labeled, abstract = {A significant portion of the world's text is tagged by readers on social bookmarking websites. Credit attribution is an inherent problem in these corpora because most pages have multiple tags, but the tags do not always apply with equal specificity across the whole document. Solving the credit attribution problem requires associating each word in a document with the most appropriate tags and vice versa. This paper introduces Labeled LDA, a topic model that constrains Latent Dirichlet Allocation by defining a one-to-one correspondence between LDA's latent topics and user tags. This allows Labeled LDA to directly learn word-tag correspondences. We demonstrate Labeled LDA's improved expressiveness over traditional LDA with visualizations of a corpus of tagged web pages from del.icio.us. Labeled LDA outperforms SVMs by more than 3 to 1 when extracting tag-specific document snippets. As a multi-label text classifier, our model is competitive with a discriminative baseline on a variety of datasets.}, address = {Stroudsburg, PA, USA}, author = {Ramage, Daniel and Hall, David and Nallapati, Ramesh and Manning, Christopher D.}, booktitle = {Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing: Volume 1 - Volume 1}, keywords = {classification}, pages = {248–256}, publisher = {Association for Computational Linguistics}, series = {EMNLP '09}, title = {Labeled LDA: A Supervised Topic Model for Credit Attribution in Multi-labeled Corpora}, year = 2009 }
Why We Search: Visualizing and Predicting User Behavior. Adar, E.; Weld, D.; Bershad, B.; Gribble, S. In Proceedings of the 16th International World Wide Web Conference. 2007.
- [ Abstract ]
- [ BibTeX ]
The aggregation and comparison of behavioral patterns on the WWW represent a tremendous opportunity for understanding past behaviors and predicting future behaviors. In this paper, we take a first step at achieving this goal. We present a large scale study correlating the behaviors of Internet users on multiple systems ranging in size from 27 million queries to 14 million blog posts to 20,000 news articles. We formalize a model for events in these time-varying datasets and study their correlation. We have created an interface for analyzing the datasets, which includes a novel visual artifact, the DTWRadar, for summarizing differences between time series. Using our tool we identify a number of behavioral properties that allow us to understand the predictive power of patterns of use.

@inproceedings{adar2007, abstract = {The aggregation and comparison of behavioral patterns on the WWW represent a tremendous opportunity for understanding past behaviors and predicting future behaviors. In this paper, we take a first step at achieving this goal. We present a large scale study correlating the behaviors of Internet users on multiple systems ranging in size from 27 million queries to 14 million blog posts to 20,000 news articles. We formalize a model for events in these time-varying datasets and study their correlation. We have created an interface for analyzing the datasets, which includes a novel visual artifact, the DTWRadar, for summarizing differences between time series. Using our tool we identify a number of behavioral properties that allow us to understand the predictive power of patterns of use.}, author = {Adar, E. and Weld, D. and Bershad, B. and Gribble, S.}, booktitle = {Proceedings of the 16th International World Wide Web Conference}, keywords = {behavior}, title = {Why We Search: Visualizing and Predicting User Behavior}, year = 2007 }

Leistungsnachweis

Ein (1!) Tag vor dem ersten geplanten Vortrag sind per Email als PDF dem Betreuer abzugeben:

eine 12-seitige Ausarbeitung im hier beschriebenen Format (die Vorlage für LaTeX findet sich hier)
eine einseitige Zusammenfassung (die allen Teilnehmern ausgehändigt wird),
der Foliensatz

Die einseitige Zusammenfassung wird von dem Betreuer drei Tage vor dem Seminarvortrag an alle Teilnehmer gesendet, so dass sich die Teilnehmer auf die vorgetragenen Themen einstellen können.

Vorherige Absprachen mit dem Betreuer sind ausdrücklich erlaubt. Alle verwendeten Referenzen sind zusätzlich zum Literaturverzeichnis der Ausarbeitung in http://www.bibsonomy.org mit den folgenden Tags einzugeben:

"seminar", "ws2015", "talk", den Tags des jeweiligen Themas und weiteren sinnvollen Tags.

Die Vortragsdauer beträgt verbindlich 30 Minuten, nach dem Vortrag besteht Gelegenheit zur Diskussion. Der Vortrag (inkl. Folien und Abstract) geht mit 40% in die Endnote ein, die Ausarbeitung ebenfalls mit 40%. Die letzten 20% bewerten die Eigenständigkeit der Ausarbeitung, Einhaltung von Terminen, Beteiligung am Seminar etc.

Das Thema kann bis zu zwei Wochen nach der Vorbesprechung noch abgelegt werden. Dies muss schriftlich erfolgen. Geschieht dies nicht, muss auf jeden Fall eine Ausarbeitung abgegeben und ein Vortrag gehalten werden. Wird dies nicht getan, so zählt das Seminar als nicht bestanden.

Bearbeitungshinweise

Als Richtlinie für die Erstellung einer guten Seminararbeit (inkl. Vortrag und Ausarbeitung) wird das Buch

Markus Deininger and Horst Lichter and Jochen Ludewig and Kurt Schneider. Studien-Arbeiten: ein Leitfaden zur Vorbereitung, Durchführung und Betreuung von Studien-, Diplom- Abschluss- und Doktorarbeiten am Beispiel Informatik. 5. Auflage. vdf Hochschulverlag, Zürich, 2005.

empfohlen, welches beim Betreuer des Seminars (Prof. Dr. Andreas Hotho) ausliegt. Wir empfehlen die Anschaffung dieses Buchs (9,50 €), da es Sie bis zur Masterarbeit (und weiter) begleiten kann. Die Benotung der Seminararbeit erfolgt in Anlehnung an das dort auf Seite 77 angegebene Schema, angepasst auf die Erfordernisse einer Seminararbeit.

Seminar: Ausgewählte Themen des Web 2.0