The research in this dissertation consists of two parts: An active learning algorithm for hierarchical labels and an embedding-based retrieval algorithm. In the first part, we present a new approach for learning hierarchically decomposable concepts. The approach learns a high-level classifier (e.g., location vs. non-location) by separately…

Bag-of-words is a commonly used text representation method for many text classification applications. However, bag-of-words representation fails to consider the context of the text because it only examines text documents based on the presence of individual words and explores relationships between texts with similar word choices (Bengfort, 2018).…

While access to images in general has improved in the last 20 years, due to both advances in electronic storage and dissemination and to improvements in the intellectual provisions of them, access to editorial cartoons lags behind access to other types of images. While there have been piecemeal or ad hoc efforts to organize large cartoon…

The inverted index is the main data structure used by all the major search engines. Search engines build an inverted index on their collection to speed up query processing. As the size of the web grows, the length of the inverted list structures, which can easily grow to hundreds of MBs or even GBs for common terms (roughly linear in the size of…

This dissertation focuses on investigating whether users will locate desired images more efficiently and effectively when they are provided with information descriptors from both experts and the general public. This study develops a way to support image finding through a human-computer interface by providing subject headings and social tags about…

Background: According to Salo (2010), the metadata entered into repositories are "disorganized" and metadata schemes underlying repositories are "arcane". This creates a challenging repository environment in regards to personal information management (PIM) and knowledge organization systems (KOSs). This dissertation research is…

Quickly accessing the contents of a video is challenging for users, particularly for unstructured video, which contains no intentional shot boundaries, no chapters, and no apparent edited format. We approach this problem in the domain of lecture videos though the use of machine learning, to gather semantic information about the videos; and through…

Surveillance video recording is becoming ubiquitous in daily life for public areas such as supermarkets, banks, and airports. The rate at which surveillance video is being generated has accelerated demand for machine understanding to enable better content-based search capabilities. Analyzing human activity is one of the key tasks to understand and…

Recognition and separation of sounds played by various instruments is very useful in labeling audio files with semantic information. This is a non-trivial task requiring sound analysis, but the results can aid automatic indexing and browsing music data when searching for melodies played by user specified instruments. Melody match based on pitch…

This study investigates the relationships between the use of a zoom tool, the terms they supply to describe the image, and the type of image being viewed. Participants were assigned to two groups, one with access to the tool and one without, and were asked to supply terms to describe forty images, divided into four categories: landscape, portrait,…