Benefit from the recent advancements of artificial intelligence (AI) methods, industrial automation has replaced human labors in many tasks. However, humans are still placed in the central role when visual searching tasks are highly involved for manufacturing decision-making. For example, highly customized products fabricated by additive manufacturing processes have posed significant challenges to AI methods in terms of their performance and generalizability. As a result, in practice, human visual searching tasks are still widely involved in manufacturing contexts (e.g., human resource management, quality inspection, etc.) based on various visualization techniques. Quantitatively modeling the visual searching behaviors and performance will not only contribute to the understanding of decision-making process in a visualization system, but also advance AI methods by incubating them with human expertise. In general, visual searching can be quantitatively understood from multiple scales, namely, 1) the population scale to treat individuals equally and model the general relationship between individual's physiological signals with visual searching decisions; 2) the individual scale to model the relationship between individual differences and visual searching decisions; and 3) the attention scale to model the relationship between individuals' attention in visual searching and visual searching decisions. The advancements of wearable sensing techniques enable such multiscale quantitative analytics of human visual searching performance. For example, by equipping human users with electroencephalogram (EEG) device, eye tracker, and logging system, the multiscale quantitative relationships among human physiological signals, behaviors and performance can be readily established. This dissertation attempts to quantify visual searching process from multiple scales by proposing (1) a data-fusion method to model the quantitative relationship between physiological signals and human's perceived task complexities (population scale, Chapter 2); (2) a recommender system to quantify and decompose the individual differences into explicit and implicit differences via personalized recommender system-based sensor analytics (individual scale, Chapter 3); and (3) a visual language processing modeling framework to identify and correlate visual cues (i.e., identified from fixations) with humans' quality inspection decisions in human visual searching tasks (attention scale, Chapter 4). Finally, Chapter 5 summarizes the contributions and proposes future research directions. The proposed methodologies can be readily extended to other applications and research studies to support multi-scale quantitative analytics. Besides, the quantitative understanding of human visual searching behaviors performance can also generate insights to further incubate AI methods with human expertise. Merits of the proposed methodologies are demonstrated in a visualization evaluation user study, and a cognitive hacking user study. Detailed notes to guide the implementation and deployment are provided for practitioners and researchers in each chapter. / Doctor of Philosophy / Existing industrial automation is limited by the performance and generalizability of artificial intelligence (AI) methods. Therefore, various human visual searching tasks are still widely involved in manufacturing contexts based on many visualization techniques, e.g., to searching for specific information, and to make decisions based on sequentially gathered information. Quantitatively modeling the visual searching performance will not only contribute to the understanding of human behaviors in a visualization system, but also advance the AI methods by incubating them with human expertise. In this dissertation, visual searching performance is characterized from multiple scales, namely, 1) the population scale to understand the visual searching performance in regardless of individual differences; 2) the individual scale to model the performance by quantifying individual differences; and 3) the attention scale to quantify the human visual searching-based decision-making process. Thanks to the advancements in wearable sensing techniques, this dissertation attempts to quantify visual searching process from multiple scales by proposing (1) a data-fusion method to model the quantitative relationship between physiological signals and human's perceived task complexities (population scale, Chapter 2); (2) a recommender system to suggest the best visualization design to the right person at the right time via sensor analytics (individual scale, Chapter 3); and (3) a visual language processing modeling framework to model humans' quality inspection decisions (attention scale, Chapter 4). Finally, Chapter 5 summarizes the contributions and proposes future research directions. Merits of the proposed methodologies are demonstrated in a visualization evaluation user study, and a cognitive hacking user study. The proposed methodologies can be readily extended to other applications and research studies to support multi-scale quantitative analytics.

Recommender System is a tool helping users find content and overcome information overload. It predicts interests of users and makes recommendation according to the interest model of users. The original content-based recommender system is the continuation and development of collaborative filtering, which doesn’t need the user’s evaluation for items. Instead, the similarity is calculated based on the information of items that are chose by users, and then make the recommendation accordingly. With the improvement of machine learning, current content-based recommender system can build profile for users and products respectively. Building or updating the profile according to the analysis of items that are bought or visited by users. The system can compare the user and the profile of items and then recommend the most similar products. So this recommender method that compare user and product directly cannot be brought into collaborative filtering model. The foundation of content-based algorithm is acquisition and quantitative analysis of the content. As the research of acquisition and filtering of text information are mature, many current content-based recommender systems make recommendation according to the analysis of text information. This paper introduces content-based recommender system for the movie website of VionLabs. There are a lot of features extracted from the movie, they are diversity and unique, which is also the difference from other recommender systems. We use these features to construct movie model and calculate similarity. We introduce a new approach for setting weight of features, which improves the representative of movies. Finally we evaluate the approach to illustrate the improvement. / Recommender System är ett verktyg som hjälper användarna att hitta innehåll och övervinna informationsöverflöd. Det förutspår användarnas intressen och gör rekommendation enligt räntemodellen användare. Den ursprungliga innehållsbaserade recommender är en fortsättning och utveckling av samarbete filtrering, som inte behöver användarens utvärdering artiklar. Istället är likheten beräknas baserat på informationen objekt som har varit valde av användare, och sedan göra rekommendationen därefter. Med förbättringen av maskininlärning, kan nuvarande innehållsbaserad recommender systemet bygga profil för användare och produkt respektive. Bygga eller uppdatera profilen enligt analysen av objekt som köps eller besöks av användare. Systemet kan jämföra användaren och profilen av artiklar och rekommendera den mest liknande produkt. Så här recommender metod som jämför användaren och produkten direkt kan inte föras in collaborative filtreringsmodell. Grunden för innehållsbaserad algoritm är förvärv och kvantitativ analys av innehållet. Eftersom forskning förvärv och filtrering av textinformation är mogen, många aktuella innehållsbaserade recommender system gör rekommendation enligt analysen av textinformation. Denna uppsats införa innehållsbaserad recommender system för film webbplats VionLabs. Det finns en mängd funktioner som extraherats från en film, är de mångfald och unik, vilket är också skillnaden med andra recommender system. Vi använder dessa funktioner för att konstruera film vektor och beräkna likheter. Vi introducerar en ny metod för att fastställa vikten av funktioner, vilket förbättrar företrädare för filmer. Slutligen utvärderar vi tillvägagångssättet för att illustrera förbättringen.

Chatbots are computer programs with the capability to lead a conversation with a human user. When a chatbot is unable to match a user’s utterance to any predefined answer, it will use a fallback intent; a generic response that does not contribute to the conversation in any meaningful way. This report aims to investigate if a content-based recommender system could provide support to a chatbot agent in case of these fallback experiences. Content-based recommender systems use content to filter, prioritize and deliver relevant information to users. Their purpose is to search through a large amount of content and predict recommendations based on user requirements. The recommender system developed in this project consists of four components: a web spider, a Bag-of-words model, a graph database, and the GraphQL API. The anticipation was to capture web page articles and rank them with a numeric scoring to figure out which articles that make for the best recommendation concerning given subjects. The chatbot agent could then use these recommended articles to provide the user with value and help instead of a generic response. After the evaluation, it was found that the recommender system in principle fulfilled all requirements, but that the scoring algorithm used could achieve significant improvements in its recommendations if a more advanced algorithm would be implemented. The scoring algorithm used in this project is based on word count, which lacks taking the context of the dialogue between the user and the agent into consideration, among other things.

A North American man has approximately a one in six chance of being diagnosed with prostate cancer in his lifetime. In most cases, there is no clearly optimal treatment, so he may be invited to participate in a treatment decision between several medically reasonable options, each with potential short- and long-term side effects. Information needs are high at diagnosis and can continue to be elevated for years or decades. Many men and their families seek information online, where, due partly to the array of websites available and high variation in information preferences, it can be difficult to find personally relevant and useful websites. This research sought to address this issue by developing methods to categorize prostate cancer websites and exploring quantitative and qualitative relationships between websites, information-seekers, and individuals’ assessments of websites. The research involved a series of three studies. In the first study, 29 men with prostate cancer participated in a needs assessment involving questionnaires, an interview, and interaction with a prototype website. In the second study, a detailed classification system was developed and applied to a set of forty websites selected to be representative of the variety of prostate cancer websites available. The third (online) study collected clinical, cognitive, and psychosocial details from 65 participants along with their ratings of websites from study two. A number of hypotheses were tested. One finding was that, compared to men with greater trust, men with lower trust in their physician tended to judge commercial websites as less relevant and useful, and found websites with descriptions of personal experiences more relevant and useful. Analyses also addressed a number of exploratory questions, including whether website and individual attributes might predict preferences for websites. Using discriminant analysis on 80% of the data, two functions were identified that predicted ratings significantly better than chance. These relationships were then validated with 20% of the data held back for testing. The results are discussed in terms of their implications for information tailoring and recommender systems for prostate cancer patients searching for information online. Limitations of the current research and recommendations for future research are also presented.

A North American man has approximately a one in six chance of being diagnosed with prostate cancer in his lifetime. In most cases, there is no clearly optimal treatment, so he may be invited to participate in a treatment decision between several medically reasonable options, each with potential short- and long-term side effects. Information needs are high at diagnosis and can continue to be elevated for years or decades. Many men and their families seek information online, where, due partly to the array of websites available and high variation in information preferences, it can be difficult to find personally relevant and useful websites. This research sought to address this issue by developing methods to categorize prostate cancer websites and exploring quantitative and qualitative relationships between websites, information-seekers, and individuals’ assessments of websites. The research involved a series of three studies. In the first study, 29 men with prostate cancer participated in a needs assessment involving questionnaires, an interview, and interaction with a prototype website. In the second study, a detailed classification system was developed and applied to a set of forty websites selected to be representative of the variety of prostate cancer websites available. The third (online) study collected clinical, cognitive, and psychosocial details from 65 participants along with their ratings of websites from study two. A number of hypotheses were tested. One finding was that, compared to men with greater trust, men with lower trust in their physician tended to judge commercial websites as less relevant and useful, and found websites with descriptions of personal experiences more relevant and useful. Analyses also addressed a number of exploratory questions, including whether website and individual attributes might predict preferences for websites. Using discriminant analysis on 80% of the data, two functions were identified that predicted ratings significantly better than chance. These relationships were then validated with 20% of the data held back for testing. The results are discussed in terms of their implications for information tailoring and recommender systems for prostate cancer patients searching for information online. Limitations of the current research and recommendations for future research are also presented.

Despite recommender systems being useful, for some applications it is hard to accumulate all the required information needed for the recommendation. In today‟s ubiquitous environment, mobile devices with different characteristics are widely available. Our work focuses on the recommendation service built on mobile environment to support tourists‟ traveling need. When tourists visit a new attraction, their recommender systems can exchange data with the attraction system to help obtain rating information of people with similar tastes. Such asynchronous rating exchange mechanisms allow a tourist to receive ratings from other people even though they may not collocate at the same time. We proposed four data exchange methods between a user and an attraction system. Our recommendation mechanism incorporates other users‟ opinions to provide recommendations once the user has collected enough ratings. Every method is compared under four conditions which attraction systems carry different amount of existing data. Then we compare these methods under different amount of existing rating data and shed the light on their advantages and disadvantages. Finally, we compare our proposed asynchronous methods with other synchronous data exchange methods proposed previously.

Accompanying with the Internet growth explosion, more and more information disseminates on the Web. The large amount of information, however, causes the information overload problem that disturbs users who desire to search and find useful information online. Information retrieval and information filtering arise to compensate for the searching and comprehending ability of the users. Recommender systems as one of the information filtering techniques emerge when users cannot describe their requirements precisely as keywords. Collaborative filtering (CF) compares novel information with common interests shared by a group of people to make the recommendations. One of its methods, the Model-based CF, generates predicted recommendation based on the model learned from the past opinions of the users. However, two issues on model-based CF should be addressed. First, data quality of the rating matrix input can affect the prediction performance. Second, most current models treat the data class as the nominal scale instead of ordinal nature in ratings. The objective of this research is thus to propose a model-based CF algorithm that considers data reliability and data scale in the model. Three experiments are conducted accordingly, and the results show our proposed method outperforms other counterparts especially under data of mild sparsity degree and of large scale. These results justify the feasibility of our proposed method in real applications.

Interpersonal relationship and recommendation are the important relation and popular mechanism. Living in the information-overloading age, the original information searching mechanisms, which require the specification of keywords, are ineffective and impractical. Moreover, a variety of recommendation techniques have been proposed and many of them have been implemented in real systems, especially in online stores. Among different recommendation techniques proposed in the literature, the content-based and collaborative filtering approaches have been broadly adopted by membership stores that maintain users¡¦ long term interest. For short-term interest, by far the content-based approach is the most popular one for recommendation. However, most of the proposed recommendation approaches do not take the social information as an important factor. In this study, we proposed several social network-based recommendation approaches that take into account the similarities of items with respect to their social closeness for meeting users¡¦ short term interests. Our experiment evaluation results show that social network-based approaches perform better than the content-based counterpart, if the user¡¦s short term interest profile contains articles of similar content. Nonetheless, content-based approach becomes better when articles in the profile are diversified in their content. Besides, contrast to content-based approach, social network-based approach is less likely to recommend articles which readers do not value. Finally, the desired articles recommended by content-based approach are very different from those by social network-based approach. This suggests the development of some hybrid recommendation method that utilizes both content and social network when making recommendations.

Recommender systems are information retrieval tools helping users in their information seeking tasks and guiding them in a large space of possible options. Many hybrid recommender systems are proposed so far to overcome shortcomings born of pure content-based (PCB) and pure collaborative filtering (PCF) systems. Most studies on recommender systems aim to improve the accuracy and efficiency of predictions. In this thesis, we propose an online hybrid recommender strategy (CBCFdfc) based on content boosted collaborative filtering algorithm which aims to improve the prediction accuracy and efficiency. CBCFdfc combines content-based and collaborative characteristics to solve problems like sparsity, new item and over-specialization. CBCFdfc uses fuzzy clustering to keep a certain level of prediction accuracy while decreasing online prediction time. We compare CBCFdfc with PCB and PCF according to prediction accuracy metrics, and with CBCFonl (online CBCF without clustering) according to online recommendation time. Test results showed that CBCFdfc performs better than other approaches in most cases. We, also, evaluate the effect of user-specified parameters to the prediction accuracy and efficiency. According to test results, we determine optimal values for these parameters. In addition to experiments made on simulated data, we also perform a user study and evaluate opinions of users about recommended movies. The results that are obtained in user evaluation are satisfactory. As a result, the proposed system can be regarded as an accurate and efficient hybrid online movie recommender.

Living in an information-overloading age, the original information searching mechanisms are ineffective and impractical. As the e-commerce is more and more popular, using information technology to discover the latent demand of customers becomes an important issue. Hence, a variety of recommendation techniques have been proposed and many of them have been implemented in real systems, mostly in online stores. Among the techniques, the content-based and collaborative filtering approaches are the ones broadly adopted and proved to be successful. Recently, social network-based recommendation approach has been proposed that takes into account the similarities of items with respect to their social closeness. The social network-based approach performs better than content-based approach in some scenarios and it can also avoid recommending articles that have high content similarity to a user¡¦s favorite articles but low quality. Therefore, we propose three hybrid approaches, Switching, Proportional, and Fusion that combine content-based and social network-based approaches in order to achieve a better performance. Our experimental result shows that even though the proposed approaches have pros and cons under different scenarios, in general they achieve better performance than individual approaches. Besides, we generate some synthetic articles that have close content similarities to articles in our collection to evaluate the fidelity of each approach. The experimental results show that approaches incorporating social network information have lower chance to recommend these faked articles.