Nathan Good

From context to content: leveraging context to infer media metadata

The recent popularity of mobile camera phones allows for new opportunities to gather important metadata at the point of capture. This paper describes a method for generating metadata for photos using spatial, temporal, and social context. We describe a system we implemented for inferring location information for pictures taken with camera phones and its performance evaluation. We propose that leveraging contextual metadata at the point of capture can address the problems of the semantic and sensory gaps. In particular, combining and sharing spatial, temporal, and social contextual metadata from a given user and across users allows us to make inferences about media content.

Towards context-aware face recognition

In this paper, we focus on the use of context-aware, collaborative filtering, machine-learning techniques that leverage automatically sensed and inferred contextual metadata together with computer vision analysis of image content to make accurate predictions about the human subjects depicted in cameraphone photos. We apply Sparse-Factor Analysis (SFA) to both the contextual metadata gathered in the MMM2 system and the results of PCA (Principal Components Analysis) of the photo content to achieve a 60% face recognition accuracy of people depicted in our cameraphone photos, which is 40% better than media analysis alone. In short, we use context-aware media analysis to solve the face recognition problem for cameraphone photos.

Effects of content and time of delivery on receptivity to mobile interruptions

In this paper we investigate effects of the content of interruptions and of the time of interruption delivery on mobile phones. We review related work and report on a naturalistic quasi-experiment using experience-sampling that showed that the receptivity to an interruption is influenced by its content rather than by its time of delivery in the employed modality of delivery - SMS. We also examined the underlying variables that increase the perceived quality of content and found that the factors interest, entertainment, relevance and actionability influence peoples receptivity significantly. Our findings inform system design that seeks to provide context-sensitive information or to predict interruptibility and suggest the consideration of receptivity as an extension to the way we think and reason about interruptibility.

How users use access control

Existing technologies for file sharing differ widely in the granularity of control they give users over who can access their data; achieving finer-grained control generally requires more user effort. We want to understand what level of control users need over their data, by examining what sorts of access policies users actually create in practice. We used automated data mining techniques to examine the real-world use of access control features present in standard document sharing systems in a corporate environment as used over a long (> 10 year) time span. We find that while users rarely need to change access policies, the policies they do express are actually quite complex. We also find that users participate in larger numbers of access control and email sharing groups than measured by self-report in previous studies. We hypothesize that much of this complexity might be reduced by considering these policies as examples of simpler access control patterns. From our analysis of what access control features are used and where errors are made, we propose a set of design guidelines for access control systems themselves and the tools used to manage them, intended to increase usability and decrease error.

Using Context and Similarity for Face and Location Identification

This paper describes a new approach to the automatic detection of human faces and places depicted in photographs taken on cameraphones. Cameraphones offer a unique opportunity to pursue new approaches to media analysis and management: namely to combine the analysis of automatically gathered contextual metadata with media content analysis to fundamentally improve image content recognition and retrieval. Current approaches to content-based image analysis are not sufficient to enable retrieval of cameraphone photos by high-level semantic concepts, such as who is in the photo or what the photo is actually depicting. In this paper, new methods for determining image similarity are combined with analysis of automatically acquired contextual metadata to substantially improve the performance of face and place recognition algorithms. For faces, we apply Sparse-Factor Analysis (SFA) to both the automatically captured contextual metadata and the results of PCA (Principal Components Analysis) of the photo content to achieve a 60% face recognition accuracy of people depicted in our database of photos, which is 40% better than media analysis alone. For location, grouping visually similar photos using a model of Cognitive Visual Attention (CVA) in conjunction with contextual metadata analysis yields a significant improvement over color histogram and CVA methods alone. We achieve an improvement in location retrieval precision from 30% precision for color histogram and CVA image analysis, to 55% precision using contextual metadata alone, to 67% precision achieved by combining contextual metadata with CVA image analysis. The combination of context and content analysis produces results that can indicate the faces and places depicted in cameraphone photos significantly better than image analysis or context analysis alone. We believe these results indicate the possibilities of a new context-aware paradigm for image analysis.

Empirical studies on software notices to inform policy makers and usability designers

We evaluate the usability of End User License Agreements (EULAs) of popular consumer programs. Results from an empirical evaluation of 50 popular programs show the lack of accessibility and readability of notices. Our data from a recent study with 64 users involved in installation tasks confirms the public perception that notice to and consent by the user is not achieved.

Radio frequency Id and privacy with information goods

1. NORMS AND LAW This paper examines the privacy impacts of using radio frequency identification (RFID) to tag information goods such as books, music, and video. Individuals have strong expectations of privacy in their choice of information goods. These expectations are supported by both social norms and law. As a matter of practice, people may generally purchase and browse information goods without identifying themselves or the subject of their inquiry. People may pay in cash and avoid creating records that provide opportunities for third parties to learn of their information habits. Information providers that maintain records, such as libraries and bookstores, have staunchly defended their patrons’ privacy, and indeed are often bound legally to demand due process of law before disclosing those records. Data holders can examine subpoenas for authenticity and cause, and challenge them in court before disclosing private information. Bookstores have done so in recent high-profile cases. [6][9] Libraries have developed elaborate policy mechanisms to ensure records are kept private, [1] and lobbied for laws protecting library records.