A multimedia indexing and retrieval framework for multimedia database systems

The main challenges of multimedia data retrieval lie in the effective mapping between low-level features and high-level concepts, and in the individual users' subjective perceptions of multimedia content. ^ The objectives of this dissertation are to develop an integrated multimedia indexing and retrieval framework with the aim to bridge the gap between semantic concepts and low-level features. To achieve this goal, a set of core techniques have been developed, including image segmentation, content-based image retrieval, object tracking, video indexing, and video event detection. These core techniques are integrated in a systematic way to enable the semantic search for images/videos, and can be tailored to solve the problems in other multimedia related domains. In image retrieval, two new methods of bridging the semantic gap are proposed: (1) for general content-based image retrieval, a stochastic mechanism is utilized to enable the long-term learning of high-level concepts from a set of training data, such as user access frequencies and access patterns of images. (2) In addition to whole-image retrieval, a novel multiple instance learning framework is proposed for object-based image retrieval, by which a user is allowed to more effectively search for images that contain multiple objects of interest. An enhanced image segmentation algorithm is developed to extract the object information from images. This segmentation algorithm is further used in video indexing and retrieval, by which a robust video shot/scene segmentation method is developed based on low-level visual feature comparison, object tracking, and audio analysis. Based on shot boundaries, a novel data mining framework is further proposed to detect events in soccer videos, while fully utilizing the multi-modality features and object information obtained through video shot/scene detection. ^ Another contribution of this dissertation is the potential of the above techniques to be tailored and applied to other multimedia applications. This is demonstrated by their utilization in traffic video surveillance applications. The enhanced image segmentation algorithm, coupled with an adaptive background learning algorithm, improves the performance of vehicle identification. A sophisticated object tracking algorithm is proposed to track individual vehicles, while the spatial and temporal relationships of vehicle objects are modeled by an abstract semantic model. ^

An incremental multilinear system for human face learning and recognition

This dissertation establishes a novel system for human face learning and recognition based on incremental multilinear Principal Component Analysis (PCA). Most of the existing face recognition systems need training data during the learning process. The system as proposed in this dissertation utilizes an unsupervised or weakly supervised learning approach, in which the learning phase requires a minimal amount of training data. It also overcomes the inability of traditional systems to adapt to the testing phase as the decision process for the newly acquired images continues to rely on that same old training data set. Consequently when a new training set is to be used, the traditional approach will require that the entire eigensystem will have to be generated again. However, as a means to speed up this computational process, the proposed method uses the eigensystem generated from the old training set together with the new images to generate more effectively the new eigensystem in a so-called incremental learning process. In the empirical evaluation phase, there are two key factors that are essential in evaluating the performance of the proposed method: (1) recognition accuracy and (2) computational complexity. In order to establish the most suitable algorithm for this research, a comparative analysis of the best performing methods has been carried out first. The results of the comparative analysis advocated for the initial utilization of the multilinear PCA in our research. As for the consideration of the issue of computational complexity for the subspace update procedure, a novel incremental algorithm, which combines the traditional sequential Karhunen-Loeve (SKL) algorithm with the newly developed incremental modified fast PCA algorithm, was established. In order to utilize the multilinear PCA in the incremental process, a new unfolding method was developed to affix the newly added data at the end of the previous data. The results of the incremental process based on these two methods were obtained to bear out these new theoretical improvements. Some object tracking results using video images are also provided as another challenging task to prove the soundness of this incremental multilinear learning method.

Text analytics of social media: Sentiment analysis, event detection and summarization

In the last decade, large numbers of social media services have emerged and been widely used in people's daily life as important information sharing and acquisition tools. With a substantial amount of user-contributed text data on social media, it becomes a necessity to develop methods and tools for text analysis for this emerging data, in order to better utilize it to deliver meaningful information to users. ^ Previous work on text analytics in last several decades is mainly focused on traditional types of text like emails, news and academic literatures, and several critical issues to text data on social media have not been well explored: 1) how to detect sentiment from text on social media; 2) how to make use of social media's real-time nature; 3) how to address information overload for flexible information needs. ^ In this dissertation, we focus on these three problems. First, to detect sentiment of text on social media, we propose a non-negative matrix tri-factorization (tri-NMF) based dual active supervision method to minimize human labeling efforts for the new type of data. Second, to make use of social media's real-time nature, we propose approaches to detect events from text streams on social media. Third, to address information overload for flexible information needs, we propose two summarization framework, dominating set based summarization framework and learning-to-rank based summarization framework. The dominating set based summarization framework can be applied for different types of summarization problems, while the learning-to-rank based summarization framework helps utilize the existing training data to guild the new summarization tasks. In addition, we integrate these techneques in an application study of event summarization for sports games as an example of how to better utilize social media data. ^

Text Analytics of Social Media: Sentiment Analysis, Event Detection and Summarization

In the last decade, large numbers of social media services have emerged and been widely used in people's daily life as important information sharing and acquisition tools. With a substantial amount of user-contributed text data on social media, it becomes a necessity to develop methods and tools for text analysis for this emerging data, in order to better utilize it to deliver meaningful information to users. Previous work on text analytics in last several decades is mainly focused on traditional types of text like emails, news and academic literatures, and several critical issues to text data on social media have not been well explored: 1) how to detect sentiment from text on social media; 2) how to make use of social media's real-time nature; 3) how to address information overload for flexible information needs. In this dissertation, we focus on these three problems. First, to detect sentiment of text on social media, we propose a non-negative matrix tri-factorization (tri-NMF) based dual active supervision method to minimize human labeling efforts for the new type of data. Second, to make use of social media's real-time nature, we propose approaches to detect events from text streams on social media. Third, to address information overload for flexible information needs, we propose two summarization framework, dominating set based summarization framework and learning-to-rank based summarization framework. The dominating set based summarization framework can be applied for different types of summarization problems, while the learning-to-rank based summarization framework helps utilize the existing training data to guild the new summarization tasks. In addition, we integrate these techneques in an application study of event summarization for sports games as an example of how to better utilize social media data.

New media data to identify student training needs

The main objective is to exhibit how usage data from new media can be used to assess areas where students need more help in creating their ETDs. After attending this session, attendees will be able to use usage data from new media, in conjunction with traditional assessment data, to identify strengths and weaknesses in ETD training and resources. The burgeoning ETD program at Florida International University (FIU) has provided many opportunities to experiment with assessment strategies and new media. The usage statistics from YouTube and the ETD LibGuide revealed areas of strength and weakness in the training resources and the overall ETD training initiative. With the ability to assess these materials, they have been updated to better meet student needs. In addition to these assessment tools, there are opportunities to connect these statistics with data from a common error checklist, student feedback from ETD workshops, and final ETD submission surveys to create a full-fledged outcome based assessment program for the ETD initiative.

An evaluation of child welfare inservice training: The relative effectiveness of substance abuse content in improving worker knowledge and attitudes

Given the seriousness of substance abuse as a child welfare problem, the purpose of this study was to examine the relative effectiveness of an inservice training curriculum for child welfare workers. The training was designed to improve worker knowledge and attitudes in working with substance abusing families. Seventy (70) child welfare workers from public and private agencies in two south Florida counties participated in a pretest/posttest control group design that also trained and retested the control group. The literature review supports that the general preparedness of child welfare workers for the issues presented by substance abusing families is in question. Confounding this problem is a lack of understanding of substance abuse dynamics, worker biases, and predispositions. The two research hypotheses focused on whether inservice training could increase worker knowledge and improve worker attitudes in working with this population. Training delivery was in the form of a five-day inservice focusing on an array of substance abuse knowledge and attitudinal topics. Separate knowledge and attitude instruments were developed for the research and were administered, before and after training, to a purposive sample of participants that were randomly assigned to the experimental and control groups. The data analysis supported the research hypotheses but raised a question. Specifically, the experimental group demonstrated significant improvement in posttest scores on both instruments after receiving the training; whereas the control group, with training withheld, also demonstrated a significant improvement at posttest, but only on the knowledge instrument. Although the question was unanswered, when examined at a more critical significance level, only the experimental group remained significant. The hypotheses were reconfirmed when, after training and retesting, the control group also displayed significant improvement on both instruments. The findings support the conclusion that this substance abuse inservice was effective in improving worker knowledge and attitudes regarding working with substance abusing families. As an implication for social work practice, it suggests that similar inservice training can be a viable training resource when formal substance abuse training is unavailable. Additional research is suggested regarding to what degree increased substance abuse knowledge and improved worker attitudes correlate with improved practice.

Secondary school inclusion rates: The relationship between the training and beliefs of school site principals and the implementation of inclusion

Programs require strong support and guidance from those in leadership positions to ensure proper implementation (Fullen, 2001). Consequently, school site principals must rely on the training they have received to support them in making appropriate decisions. It is the school site principal’s leadership that is pivotal in the success of students with disabilities (DiPaola & Walther-Thomas, 2003; Monteith, 2000). In fact, the principal has a moral obligation to provide an environment that supports social justice in schools (Grogan & Andrews, 2002). The inclusion of students with disabilities does just that—it ensures that these students are not segregated to a “separate but equal” education. This study utilized a participant survey to collect data on principals’ beliefs and training in special education. This information was compared to the percentage of time students with disabilities spent with their non-disabled peers in the principals’ respective schools. An analysis was conducted to identify if a linear relationship exists between the selected variables and the inclusion percentages. Open-ended questions were included in the original survey which allowed for a thematic analysis of the responses. These responses were utilized to allow participants to further express their thoughts on the identified variables. Results indicated that there were no statistically significant relationships identified between the beliefs and training of secondary school site principals and the percentage of time that their students in special education spend with their non-disabled peers. Although the original research questions were not supported, further post hoc analysis indicated that the results obtained did support that the principals believed inclusion had a social benefit to students. Additional investigation into the academic benefits of inclusion is still needed. In addition, principals who indicated that they had some type of training in special education indicated a higher percentage that the individual student should be the focal point when making placement decisions. These results support the need for further research in the area of principal preparation programs and their relationships to the daily practice of school site principals.

An investigation of the strategies used by organizations participating in the welfare-to-work network program and correlation of the strategies with the retention data to determine best practices for job retention among former welfare recipients

This ex post facto study (N = 209) examined the relationships between employer job strategies and job retention among organizations participating in Florida welfare-to-work network programs and associated the strategies with job retention data to determine best practices. ^ An internet-based self-report survey battery was administered to a heterogeneous sampling of organizations participating in the Florida welfare-to-work network program. Hypotheses were tested through correlational and hierarchical regression analytic procedures. The partial correlation results linked each of the job retention strategies to job retention. Wages, benefits, training and supervision, communication, job growth, work/life balance, fairness and respect were all significantly related to job retention. Hierarchical regression results indicated that the training and supervision variable was the best predictor of job retention in the regression equation. ^ The size of the organization was also a significant predictor of job retention. Large organizations reported higher job retention rates than small organizations. There was no statistical difference between the types of organizations (profit-making and non-profit) and job retention. The standardized betas ranged from to .26 to .41 in the regression equation. Twenty percent of the variance in job retention was explained by the combination of demographic and job retention strategy predictors, supporting the theoretical, empirical, and practical relevance of understanding the association between employer job strategies and job retention outcomes. Implications for adult education and human resource development theory, research, and practice are highlighted as possible strategic leverage points for creating conditions that facilitate the development of job strategies as a means for improving former welfare workers’ job retention.^

High School Influences on the Selection of Athletic Training as a Career

Context: Research suggests internships, mentorship, and specialized school programs positively influence career selection; however, little data exists specific to athletic training. Objective: We identified high school (HS) experiences influencing career choice in college athletic training students (ATS). Design: Our survey included 35 Likert-type close-ended questions, which were reviewed by a panel of faculty and peers to establish content and construct validity. Setting: Participants completed an online questionnaire at their convenience. Participants: 217 college ATS (153 female, 64 male) from a random selection of accredited programs on the east coast. We excluded minors, freshmen, and undecided majors from the study. Informed consent was implied by proceeding to the questionnaire. Data Collection and Analysis: We used descriptive statistics to analyze the data collected via a secure website. Results: Mentors were most influential in the decision of career path (62.4%;n=131/210) with 85.2% (n=138/162) reporting mentors were readily available to answer questions regarding career options and 53.1% (n=86/162) counseled them regarding HS electives. Of participants involved in an internship (41.0%;n=86/210), most developed such opportunities independently (66.3%;n=57/86). Respondents who attended traditional HS suggested providing diverse electives (71.9%;n=133/185), additional internship (53.5%;n=99/185), and mentorship (33.0%;n=61/185) opportunities to effectively educate students regarding career options. Conclusions: College ATS that gained internship experience during HS report the opportunity positively influenced their career selection. Mentors support HS students by offering insight and expertise in guiding students’ career choices. Participants suggested HS afford diverse electives with internship and mentorship opportunities to positively influence interested students towards pursuing a career in athletic training.

Comparing novice designer performance using a rule -based approach versus a pattern -based approach in conceptual data modeling tasks

Conceptual database design is an unusually difficult and error-prone task for novice designers. This study examined how two training approaches---rule-based and pattern-based---might improve performance on database design tasks. A rule-based approach prescribes a sequence of rules for modeling conceptual constructs, and the action to be taken at various stages while developing a conceptual model. A pattern-based approach presents data modeling structures that occur frequently in practice, and prescribes guidelines on how to recognize and use these structures. This study describes the conceptual framework, experimental design, and results of a laboratory experiment that employed novice designers to compare the effectiveness of the two training approaches (between-subjects) at three levels of task complexity (within subjects). Results indicate an interaction effect between treatment and task complexity. The rule-based approach was significantly better in the low-complexity and the high-complexity cases; there was no statistical difference in the medium-complexity case. Designer performance fell significantly as complexity increased. Overall, though the rule-based approach was not significantly superior to the pattern-based approach in all instances, it out-performed the pattern-based approach at two out of three complexity levels. The primary contributions of the study are (1) the operationalization of the complexity construct to a degree not addressed in previous studies; (2) the development of a pattern-based instructional approach to database design; and (3) the finding that the effectiveness of a particular training approach may depend on the complexity of the task.

Ensemble Stream Model for Data-Cleaning in Sensor Networks

Ensemble Stream Modeling and Data-cleaning are sensor information processing systems have different training and testing methods by which their goals are cross-validated. This research examines a mechanism, which seeks to extract novel patterns by generating ensembles from data. The main goal of label-less stream processing is to process the sensed events to eliminate the noises that are uncorrelated, and choose the most likely model without over fitting thus obtaining higher model confidence. Higher quality streams can be realized by combining many short streams into an ensemble which has the desired quality. The framework for the investigation is an existing data mining tool. First, to accommodate feature extraction such as a bush or natural forest-fire event we make an assumption of the burnt area (BA*), sensed ground truth as our target variable obtained from logs. Even though this is an obvious model choice the results are disappointing. The reasons for this are two: One, the histogram of fire activity is highly skewed. Two, the measured sensor parameters are highly correlated. Since using non descriptive features does not yield good results, we resort to temporal features. By doing so we carefully eliminate the averaging effects; the resulting histogram is more satisfactory and conceptual knowledge is learned from sensor streams. Second is the process of feature induction by cross-validating attributes with single or multi-target variables to minimize training error. We use F-measure score, which combines precision and accuracy to determine the false alarm rate of fire events. The multi-target data-cleaning trees use information purity of the target leaf-nodes to learn higher order features. A sensitive variance measure such as ƒ-test is performed during each node's split to select the best attribute. Ensemble stream model approach proved to improve when using complicated features with a simpler tree classifier. The ensemble framework for data-cleaning and the enhancements to quantify quality of fitness (30% spatial, 10% temporal, and 90% mobility reduction) of sensor led to the formation of streams for sensor-enabled applications. Which further motivates the novelty of stream quality labeling and its importance in solving vast amounts of real-time mobile streams generated today.

A State and Territorial Survey Regarding Utilization of Environmental Health Shelter Assessments during Disasters, and a Secondary Analysis of Available Shelter Assessment Data

Disasters are complex events characterized by damage to key infrastructure and population displacements into disaster shelters. Assessing the living environment in shelters during disasters is a crucial health security concern. Until now, jurisdictional knowledge and preparedness on those assessment methods, or deficiencies found in shelters is limited. A cross-sectional survey (STUSA survey) ascertained knowledge and preparedness for those assessments in all 50 states, DC, and 5 US territories. Descriptive analysis of overall knowledge and preparedness was performed. Fisher’s exact statistics analyzed differences between two groups: jurisdiction type and population size. Two logistic regression models analyzed earthquakes and hurricane risks as predictors of knowledge and preparedness. A convenience sample of state shelter assessments records (n=116) was analyzed to describe environmental health deficiencies found during selected events. Overall, 55 (98%) of jurisdictions responded (states and territories) and appeared to be knowledgeable of these assessments (states 92%, territories 100%, p = 1.000), and engaged in disaster planning with shelter partners (states 96%, territories 83%, p = 0.564). Few had shelter assessment procedures (states 53%, territories 50%, p = 1.000); or training in disaster shelter assessments (states 41%, 60% territories, p = 0.638). Knowledge or preparedness was not predicted by disaster risks, population size, and jurisdiction type in neither model. Knowledge: hurricane (Adjusted OR 0.69, 95% C.I. 0.06-7.88); earthquake (OR 0.82, 95% C.I. 0.17-4.06); and both risks (OR 1.44, 95% C.I. 0.24-8.63); preparedness model: hurricane (OR 1.91, 95% C.I. 0.06-20.69); earthquake (OR 0.47, 95% C.I. 0.7-3.17); and both risks (OR 0.50, 95% C.I. 0.06-3.94). Environmental health deficiencies documented in shelter assessments occurred mostly in: sanitation (30%); facility (17%); food (15%); and sleeping areas (12%); and during ice storms and tornadoes. More research is needed in the area of environmental health assessments of disaster shelters, particularly, in those areas that may provide better insight into the living environment of all shelter occupants and potential effects in disaster morbidity and mortality. Also, to evaluate the effectiveness and usefulness of these assessments methods and the data available on environmental health deficiencies in risk management to protect those at greater risk in shelter facilities during disasters.

Mand – to Tact Training Transfer Acquisition Rates in Children Diagnosed with Autism Spectrum Disorder with Limited Verbal Skills

Autism Spectrum Disorder () is defined as “the presence of severe and pervasive impairments in reciprocal social interaction and in verbal and nonverbal communication skills” (Diagnostic & Statistical Manual, 2000). It is estimated that 1 in 68 children across the United States are diagnosed with ASD. One of the most common delays that children diagnosed with ASD experience are language delays. Children with ASD that have a language delay will often develop maladaptive behaviors as a result of poor communication skills (Carr & Durand, 1985). The failure to develop mand acquisition in typical fashion results in behaviors ranging from social withdrawal to self-injurious behaviors (Cooper et. al, 2007). A lack of a strong tact repertoire can further impede and complicate the learning of other necessary components of language due to the inability to successfully label items and events in the physical environment of the child. The purpose of this study is to replicate with a reversal in verbal operant training of the procedures described in Wallace et al. (2006) in which two children with ASD underwent tact training to facilitate the formation of mands; essentially this study aims to accomplish mand training first to establish as tact. It is hypothesized that mand training will result in a greater repertoire of tacts due to strength of the relationship between mands and the control over the social environment (Cooper et al., 2007). The two children in the study will be taught to mand items that will be ranked in order of preference via stimulus preference assessment. This study is of great importance due to the indispensable value of effective social communication skills. Data gathered on improving communication skills is of great value to the ASD community as the implications for functional skills result in better communication with family and greater control of individual functioning.

The Effects of Attention Bias Modification Training on Cognitive Biases Perceived Threats in Children with Anxiety

It has been found in research that children and adults with anxiety have a bias toward interpreting ambiguous situations as threatening. This bias is thought to consequently maintain many symptoms of anxiety. An emergent computer treatment system called Attention Bias Modification Training (ABMT) has been used to try to reduce this bias. It is essential to understand whether this bias can be reduced with ABMT because of its feasibility and cost effective nature of treatment. In the current study, interpretation bias is measured using the Children's Opinions of Everyday Life Events (COELE). The ABMT treatment is given to children once a week for an hour and their answers to the COELE are recorded before and after treatment. The recorded procedures are transcribed by undergraduate students working at the Child Anxiety and Phobia lab, and then scored. Each of the situations of the COELE are rated 0 being neutral or 1 threatening interpretation of the situation. The hypothesis is that ABMT will reduce the negative interpretation bias in children over the course of 4 weeks of treatment. The study is still in the collection and transcription of data phase, and will expect to have analytical conclusions in the start of spring 2015.

Mand – to Tact Training Transfer Acquisition Rates in Children Diagnosed with Autism Spectrum Disorder with Limited Verbal Skills

Autism Spectrum Disorder () is defined as “the presence of severe and pervasive impairments in reciprocal social interaction and in verbal and nonverbal communication skills” (Diagnostic & Statistical Manual, 2000). It is estimated that 1 in 68 children across the United States are diagnosed with ASD. One of the most common delays that children diagnosed with ASD experience are language delays. Children with ASD that have a language delay will often develop maladaptive behaviors as a result of poor communication skills (Carr & Durand, 1985). The failure to develop mand acquisition in typical fashion results in behaviors ranging from social withdrawal to self-injurious behaviors (Cooper et. al, 2007). A lack of a strong tact repertoire can further impede and complicate the learning of other necessary components of language due to the inability to successfully label items and events in the physical environment of the child. The purpose of this study is to replicate with a reversal in verbal operant training of the procedures described in Wallace et al. (2006) in which two children with ASD underwent tact training to facilitate the formation of mands; essentially this study aims to accomplish mand training first to establish as tact. It is hypothesized that mand training will result in a greater repertoire of tacts due to strength of the relationship between mands and the control over the social environment (Cooper et al., 2007). The two children in the study will be taught to mand items that will be ranked in order of preference via stimulus preference assessment. This study is of great importance due to the indispensable value of effective social communication skills. Data gathered on improving communication skills is of great value to the ASD community as the implications for functional skills result in better communication with family and greater control of individual functioning.