Aisling Kelliher

MetaFac: community discovery via relational hypergraph factorization

This paper aims at discovering community structure in rich media social networks, through analysis of time-varying, multi-relational data. Community structure represents the latent social context of user actions. It has important applications in information tasks such as search and recommendation. Social media has several unique challenges. (a) In social media, the context of user actions is constantly changing and co-evolving; hence the social context contains time-evolving multi-dimensional relations. (b) The social context is determined by the available system features and is unique in each social media website. In this paper we propose MetaFac (MetaGraph Factorization), a framework that extracts community structures from various social contexts and interactions. Our work has three key contributions: (1) metagraph, a novel relational hypergraph representation for modeling multi-relational and multi-dimensional social data; (2) an efficient factorization method for community extraction on a given metagraph; (3) an on-line method to handle time-varying relations through incremental metagraph factorization. Extensive experiments on real-world social data collected from the Digg social media website suggest that our technique is scalable and is able to extract meaningful communities based on the social media contexts. We illustrate the usefulness of our framework through prediction tasks. We outperform baseline methods (including aspect model and tensor analysis) by an order of magnitude.

Tell Me a Story

In general, multimedia is nothing without a story. The original storytellers - think Homers The Iliad - memorized long stories, with plenty of emotional content, and delivered them in a highly entertaining (and possibly interactive) manner. Music, movies, and video games are perhaps todays most popular forms of story-telling. The multimedia field is contributing to the evolution of these forms by summarizing long video sequences, discovering patterns in related content, or developing techniques for generating stories. The best question we can ask is, how can multimedia systems create better user experiences?

Community Discovery via Metagraph Factorization

This work aims at discovering community structure in rich media social networks through analysis of time-varying, multirelational data. Community structure represents the latent social context of user actions. It has important applications such as search and recommendation. The problem is particularly useful in the enterprise domain, where extracting emergent community structure on enterprise social media can help in forming new collaborative teams, in expertise discovery, and in the long term reorganization of enterprises based on collaboration patterns. There are several unique challenges: (a) In social media, the context of user actions is constantly changing and coevolving; hence the social context contains time-evolving multidimensional relations. (b) The social context is determined by the available system features and is unique in each social media platform; hence the analysis of such data needs to flexibly incorporate various system features. In this article we propose MetaFac (MetaGraph Factorization), a framework that extracts community structures from dynamic, multidimensional social contexts and interactions. Our work has three key contributions: (1) metagraph, a novel relational hypergraph representation for modeling multirelational and multidimensional social data; (2) an efficient multirelational factorization method for community extraction on a given metagraph; (3) an online method to handle time-varying relations through incremental metagraph factorization. Extensive experiments on real-world social data collected from an enterprise and the public Digg social media Web site suggest that our technique is scalable and is able to extract meaningful communities from social media contexts. We illustrate the usefulness of our framework through two prediction tasks: (1) in the enterprise dataset, the task is to predict users’ future interests on tag usage, and (2) in the Digg dataset, the task is to predict users’ future interests in voting and commenting on Digg stories. Our prediction significantly outperforms baseline methods (including aspect model and tensor analysis), indicating the promising direction of using metagraphs for handling time-varying social relational contexts.

NextSlidePlease: Authoring and delivering agile multimedia presentations

Presentation support tools, such as Microsoft PowerPoint, pose challenges both in terms of creating linear presentations from complex data and fluidly navigating such linear structures when presenting to diverse audiences. NextSlidePlease is a slideware application that addresses these challenges using a directed graph structure approach for authoring and delivering multimedia presentations. The application combines novel approaches for searching and analyzing presentation datasets, composing meaningfully structured presentations, and efficiently delivering material under a variety of time constraints. We introduce and evaluate a presentation analysis algorithm intended to simplify the process of authoring dynamic presentations, and a time management and path selection algorithm that assists users in prioritizing content during the presentation process. Results from two comparative user studies indicate that the directed graph approach promotes the creation of hyperlinks, the consideration of connections between content items, and a richer understanding of the time management consequences of including and selecting presentation material.

Eventory -- An Event Based Media Repository

This paper focuses on the development of an event driven media sharing repository to facilitate community awareness. In this paper, an event refers to a real-world occurrences that unfolds over space and time. Our event model implementation supports creation of events using the standard facets of who, where, when and what. A key novelty in this research lies in the support of arbitrary event-event semantic relationships. We facilitate global as well as personalized event relationships. Each relationship can be unary or binary and can be at multiple granularities. The relationships can exist between events, between media, and between media and events. We have implemented a web based media archive system that allows people to create, explore and mange events. We have implemented an RSS based notification system that promotes awareness of actions. The initial user feedback has been positive and we are in the process of conducting a longitudinal study.

Life editing: third-party perspectives on lifelog content

Lifelog collections digitally capture and preserve personal experiences and can be mined to reveal insights and understandings of individual significance. These rich data sources also offer opportunities for learning and discovery by motivated third parties. We employ a custom-designed storytelling application in constructing meaningful lifelog summaries from third-party perspectives. This storytelling initiative was implemented as a core component in a university media-editing course. We present promising results from a preliminary study conducted to evaluate the utility and potential of our approach in creatively interpreting a unique experiential dataset.

Understanding Community Dynamics in Online Social Networks: A multidisciplinary review

Social network systems are significant scaffolds for political, economic, and sociocultural change. This is in part due to the widespread availability of sophisticated network technologies and the concurrent emergence of rich media Web sites. Social network sites provide new opportunities for social-technological research. Since we can inexpensively collect electronic records (over extended periods) of social data spanning diverse populations, it is now possible to study social processes on a scale of tens of million individuals. To understand the large-scale dynamics of interpersonal interaction and its outcome, this article links the perspectives in the humanities for analysis of social networks to recent developments in data intensive computational approaches. With special emphasis on social communities mediated by network technologies, we review the historical research arc of community analysis as well as methods applicable to community discovery in social media.

Temporal patterns in social media streams: Theme discovery and evolution using joint analysis of content and context

Online social networking sites such as Flickr and Facebook provide a diverse range of functionalities that foster online communities to create and share media content. In particular, Flickr groups are increasingly used to aggregate and share photos about a wide array of topics or themes. Unlike photo repositories where images are typically organized with respect to static topics, the photo sharing process as in Flickr often results in complex time-evolving social and visual patterns. Characterizing such time-evolving patterns can enrich media exploring experience in a social media repository. In this paper, we propose a novel framework that characterizes distinct timeevolving patterns of group photo streams. We use a nonnegative joint matrix factorization approach to incorporate image content features and contextual information, including associated tags, photo owners and post times. In our framework, we consider a group as a mixture of themes — each theme exhibits similar patterns of image content and context. The theme extraction is to best explain the observed image content features and associations with tags, users and times. Extensive experiments on a Flickr dataset suggest that our approach is able to extract meaningful evolutionary patterns from group photo streams. We evaluate our method through a tag prediction task. Our prediction results outperform baseline methods, which indicate the utility of our theme based joint analysis.

Discovering multirelational structure in social media streams

In this article, we present a novel algorithm to discover multirelational structures from social media streams. A media item such as a photograph exists as part of a meaningful interrelationship among several attributes, including time, visual content, users, and actions. Discovery of such relational structures enables us to understand the semantics of human activity and has applications in content organization, recommendation algorithms, and exploratory social network analysis. We are proposing a novel nonnegative matrix factorization framework to characterize relational structures of group photo streams. The factorization incorporates image content features and contextual information. The idea is to consider a cluster as having similar relational patterns; each cluster consists of photos relating to similar content or context. Relations represent different aspects of the photo stream data, including visual content, associated tags, photo owners, and post times. The extracted structures minimize the mutual information of the predicted joint distribution. We also introduce a relational modularity function to determine the structure cost penalty, and hence determine the number of clusters. Extensive experiments on a large Flickr dataset suggest that our approach is able to extract meaningful relational patterns from group photo streams. We evaluate the utility of the discovered structures through a tag prediction task and through a user study. Our results show that our method based on relational structures, outperforms baseline methods, including feature and tag frequency based techniques, by 35%--420%. We have conducted a qualitative user study to evaluate the benefits of our framework in exploring group photo streams. The study indicates that users found the extracted clustering results clearly represent major themes in a group; the clustering results not only reflect how users describe the group data but often lead the users to discover the evolution of the group activity.

Summarization of social activity over time: people, actions and concepts in dynamic networks

We present a framework for automatically summarizing social group activity over time. The problem is important in understanding large scale online social networks, which have diverse social interactions and exhibit temporal dynamics. In this work we construct summarization by extracting activity themes. We propose a novel unified temporal multi-graph framework for extracting activity themes over time. We use non-negative matrix factorization (NMF) approach to derive two interrelated latent spaces for users and concepts. Activity themes are extracted from the derived latent spaces to construct group activity summary. Experiments on real-world Flickr datasets demonstrate that our technique outperforms baseline algorithms such as LSI, and is additionally able to extract temporally representative activities to construct meaningful group activity summary.