Ph.D. Thesis — Mining Social Media: Tracking Content and Predicting Behavior

Manos Tsagkias

University of Amsterdam
08 November 2012
Keywords: phd thesis

From the back cover—

The advent of social media has established a symbiotic relationship between social media and online news. This relationship can be leveraged for tracking news content, and predicting behavior with tangible real-world applications, e.g., online reputation management, ad pricing, news ranking, and media analysis. In this thesis we focus on tracking news content in social media, and predicting user behavior.

In the first part, we develop methods for tracking content which build upon, and extend practices in Information Retrieval. We begin with discovering social media posts that discuss a news article yet they do not provide a hyperlink to it. Our methods model news articles using several channels of information, either endogenous or exogenous to the article. These models are then used to query an index of social media posts. During this process we found that the query models are close in size to the documents to be retrieved, violating a standard assumption of language modeling. We correct for this discrepancy by introducing two hypergeometric language models for modeling both queries, and documents to be retrieved.

In the second part, we focus on predicting behavior. First we look at predicting listeners’ preference in spoken user generated content, namely, podcasts. Then, we predict popularity of news articles from several news agents in terms of the volume of comments they receive. We develop models for predicting the popularity of an article for both before and after it is published. Finally, we look at a different aspect of news impact: how reading a news article affects future user browsing behavior. In each setting, we find patterns that characterize the underlying behavior and extract features that we then use to establish models for predicting online behavior.

I will defend my Ph.D. thesis on Wednesday, 5 December 2012, at 14:00 (GMT+1), at Agnietenkapel, Amsterdam. You are most welcome to join.

In the meantime, please feel free to grab a copy, and cite the book:

@phdthesis{tsagkias2012-thesis,
Title = {Mining Social Media: Tracking Content and Predicting Behavior},
Author = {Manos Tsagkias},
Year = {2012},
School = {University of Amsterdam}
}

Related Posts

  1. Much of what is discussed in social media is inspired by events in the news and, vice versa, social media provide us with a handle on the impact of news events. We address the following linking task: given a news article, find social media utterances that implicitly reference it. We follow a three-step approach: we derive multiple query models from a given source news article, which are then used to retrieve utterances from a target social media index, resulting in multiple ranked lists that we then merge using data fusion techniques. Query models are created by exploiting the structure of the source article and by using explicitly linked social media utterances that discuss the source article. To combat query drift resulting from the large volume of text, either in the source news article itself or in social media utterances explicitly linked to it, we introduce a graph-based method for selecting discriminative terms.  

    54.20% similar — Paper at WSDM 2011 — Linking online news and social media

  2. Abstract

    Modeling user browsing behavior is an active research area with tangible real-world applications, e.g., organizations can adapt their online presence to their visitors browsing behavior with positive effects in user engagement, and revenue. We concentrate on online news agents, and present a semi-supervised method for predicting news articles that a user will visit after reading an initial article. Our method tackles the problem using language intent models trained on historical data which can cope with unseen articles. We evaluate our method on a large set of articles and in several experimental settings. Our results demonstrate the utility of language intent models for predicting user browsing behavior within online news sites.

    47.23% similar — Paper at SIGIR 2012 — Language Intent Models for Inferring User Browsing Behavior

  3. Republished article finding is the task of identifying instances of articles that have been published in one source and republished more or less verbatim in another source, which is often a social media source. We address this task as an ad hoc retrieval problem, using the source article as a query. Our approach is based on language modeling. We revisit the assumptions underlying the unigram language model taking into account the fact that in our setup queries are as long as complete news articles. We argue that in this case, the underlying generative assumption of sampling words from a document with replacement, i.e., the multinomial modeling of documents, produces less accurate query likelihood estimates.  

    43.81% similar — Paper at SIGIR 2011 — Hypergeometric Language Models for Republished Article Finding

  4. We describe the participation of the University of Amsterdam’s ILPS group in the web, blog, web, entity, and relevance feedback track at TREC 2009. Our main preliminary conclusions are as follows. For the Blog track we find that for top stories identification a blogs to news approach outperforms a simple news to blogs approach. This is interesting, as this approach starts with no input except for a date, whereas the news to blogs approach also has news headlines as input. For the web track, we find that spam is an important issue in the ad hoc task and that Wikipedia- based heuristic optimization approaches help to boost the retrieval performance, which is assumed to potentially reduce the spam in top ranked documents. As for the diversity task, we explored different methods. Initial results show that clustering and a topic model-based approach have similar performance, which are relatively better than a query log based approach. Our performance in the Entity track was downright

    38.94% similar — Paper at TREC 2009 — The University of Amsterdam at TREC 2009

  5. Abstract

    We propose a retrieval model for searching microblog posts for a given topic of interest. We develop a language modeling approach tailored to microblogging characteristics, where redundancy-based IR methods cannot be used in a straightforward manner. We enhance this model with two groups of quality indicators: textual and microblog specific. Additionally, we propose a dynamic query expansion model for microblog post retrieval. Experimental results on Twitter data reveal the usefulness of boolean search, and demonstrate the utility of quality indicators and query expansion in microblog search.

    References

    37.89% similar — Paper at ECIR 2011 — Incorporating Query Expansion and Quality Indicators in Searching Microblog Posts

  6. Recent years have witnessed a persistent interest in generating pseudo test collections, both for training and evaluation purposes. We describe a method for generating queries and relevance judgments for microblog search in an unsupervised way. Our starting point is this intuition: tweets with a hashtag are relevant to the topic covered by the hashtag and hence to a suitable query derived from the hashtag. Our baseline method selects all commonly used hashtags, and all associated tweets as relevance judgments; we then generate a query from these tweets. Next, we generate a timestamp for each query, allowing us to use temporal information in the training process. We then enrich the generation process with knowledge derived from an editorial test collection for microblog search.  

    37.40% similar — Paper at SIGIR 2013 — Pseudo Test Collections for Training and Tuning Microblog Rankers

  7. Abstract

    Offering access to information in microblog posts requires successful language identification. Language identification on sparse and noisy data can be challenging. In this paper we explore the performance of a state-of-the-art n-gram-based language identifier, and we introduce two semi-supervised priors to enhance performance at microblog post level: (i) blogger-basedprior, using previous posts by the same blogger, and (ii) link-based prior, using the pages linked to from the post. We test our models on five languages (Dutch, English, French, German, and Spanish), and a set of 1,000 tweets per language. Results show that our priors improve accuracy, but that there is still room for improvement.

    References

    36.41% similar — Paper at DIR 2011 — Semi-Supervised Priors for Microblog Language Identification