Search

Machine learning models are biased toward data seen during the training steps. The models will tend to give good results in classes where there are many examples and poor results in those with few examples. This problem generally occurs when the classes to predict are imbalanced and this is frequent in educational data where for example, there are skills that are very difficult or very easy to master. There will be less data on students that correctly answered questions related to difficult skills and who incorrectly answered those related to skills easy to master. In this paper, we tackled this problem by proposing a hybrid architecture combining Deep Neural Network architectures— especially Long Short-Term Memory (LSTM) and Convolutional Neural Networks (CNN)—with expert knowledge for user modeling. The proposed solution uses attention mechanism to infuse expert knowledge into the Deep Neural Network. It has been tested in two contexts: knowledge tracing in an intelligent tutoring system (ITS) called Logic-Muse and prediction of socio-moral reasoning in a serious game called MorALERT. The proposed solution is compared to state-of-the-art machine learning solutions and experiments show that the resulting model can accurately predict the current student's knowledge state (in Logic-Muse) and thus enable an accurate personalization of the learning process. Other experiments show that the model can also be used to predict the level of socio-moral reasoning skills (in MorALERT). Our findings suggest the need for hybrid neural networks that integrate prior expert knowledge (especially when it is necessary to compensate for the strong dependency—of deep learning methods—on data size or the possible unbalanced datasets). Many domains can benefit from such an approach to building models that allow generalization even when there are small training data.

The online retail market shows a sharp increase in traffic during holiday sales. The ability to distinguish customers who will likely purchase is critical for provisioning traffic and for providing cost-effective promotions. This paper uniquely studies the browsing and purchasing behaviors of online shoppers during a yearly sale event in China, the world’s largest online marketplace. Based on 31 million action logs gathered from wide residential areas, we characterize the steps leading to purchases and determine their precursors. We investigate the effect of time (e.g., date, time of date), environment (e.g., platform, viewed category), and action (e.g., session time, clicks, sequence) on purchases. Action cues from shopping behaviors can be used for early detection. While most shoppers start with strong intentions to purchase, yet the moment of ordering comes rather impulsively within 30 seconds to several minutes of browsing. The predictive accuracy reaches as a high AUC of 0.924. The findings in this paper provide an understanding of traffic during mega sale events that can help online shops plan and provide a better user experience for upcoming shopping festivals. [ABSTRACT FROM AUTHOR]

User churn is an important issue in online services that threatens the health and profitability of services. Most of the previous works on churn prediction convert the problem into a binary classification task where the users are labeled as churned and non-churned. More recently, some works have tried to convert the user churn prediction problem into the prediction of user return time. In this approach which is more realistic in real world online services, at each time-step the model predicts the user return time instead of predicting a churn label. However, the previous works in this category suffer from lack of generality and require high computational complexity. In this paper, we introduce ChOracle, an oracle that predicts the user churn by modeling the user return times to service by utilizing a combination of Temporal Point Processes and Recurrent Neural Networks. Moreover, we incorporate latent variables into the proposed recurrent neural network to model the latent user loyalty to the system. We also develop an efficient approximate variational inference algorithm for learning parameters of the proposed RNN by using back propagation through time. Finally, we demonstrate the superior performance of ChOracle on a wide variety of real world datasets. [ABSTRACT FROM AUTHOR]

This study proposes a novel approach for measuring political polarization using a user-activity-based model. By exploiting data from comments, user activity in this study is defined based on features such as coverage, duration, and enthusiasm. To determine these features, we collect information on the activities of users from South Korean YouTube channels. Notably, the collected data of the model contains approximately 11 M comments from more than 600 K users based on 37 K videos of 77 YouTube channels. To handle the big data collection, we deploy a web-based platform called TubePlunger to collect video information (e.g., comments, replies, etc.) automatically from YouTube channels. The output of the model reveals that the users are strongly polarized because the number of neutral users is very small (approximately 8% of the total). We then applied this model to the other channels in the testing dataset to define polarization with a bias percentage and to visualize the user activity distribution. The experimental results show that there are 30 fully polarized YouTube channels (16 left-wing channels and 14 right-wing channels) with a measured bias ratio higher than 70%. Our method of analyzing social network data based on user activity provides the foundation for polarization analysis that can be applied to fields other than politics. [ABSTRACT FROM AUTHOR]

A session-based recommender system (SBRS) captures users' evolving behaviors and recommends the next item by profiling users in terms of items in a session. User intent and user preference are two factors affecting his (her) decisions. Specifically, the former narrows the selection scope to some item types, while the latter helps to compare items of the same type. Most SBRSs assume one arbitrary user intent dominates a session when making a recommendation. However, this oversimplifies the reality that a session may involve multiple types of items conforming to different intents. In current SBRSs, items conforming to different user intents have cross-interference in profiling users for whom only one user intent is considered. Explicitly identifying and differentiating items conforming to various user intents can address this issue and model rich contextual information of a session. To this end, we design a framework modeling user intent and preference explicitly, which empowers the two factors to play their distinctive roles. Accordingly, we propose a key-array memory network (KA-MemNN) with a hierarchical intent tree to model coarse-to-fine user intents. The two-layer weighting unit (TLWU) in KA-MemNN detects user intents and generates intent-specific user profiles. Furthermore, the hierarchical semantic component (HSC) integrates multiple sets of intent-specific user profiles along with different user intent distributions to model a multi-intent user profile. The experimental results on real-world datasets demonstrate the superiority of KA-MemNN over selected state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Although privacy settings are important not only for data privacy, but also to prevent hacking attacks like social engineering that depend on leaked private data, most users do not care about them. Research has tried to help users in setting their privacy settings by using some settings that have already been adapted by the user or individual factors like personality to predict the remaining settings. But in some cases, neither is available. However, the user might have already done privacy settings in another domain, for example, she already adapted the privacy settings on the smartphone, but not on her social network account. In this article, we investigate with the example of four domains (social network posts, location sharing, smartphone app permission settings and data of an intelligent retail store), whether and how precise privacy settings of a domain can be predicted across domains. We performed an exploratory study to examine which privacy settings of the aforementioned domains could be useful, and validated our findings in a validation study. Our results indicate that such an approach works with a prediction precision about 15%–20% better than random and a prediction without input coefficients. We identified clusters of domains that allow model transfer between their members, and discuss which kind of privacy settings (general or context-based) leads to a better prediction accuracy. Based on the results, we would like to conduct user studies to find out whether the prediction precision is perceived by users as a significant improvement over a "one-size-fits-all" solution, where every user is given the same privacy settings. [ABSTRACT FROM AUTHOR]

Gamification refers to the attempt to transform different kinds of systems to be able to better invoke positive experiences such as the flow state. However, the ability of such intervention to invoke flow state is commonly believed to depend on several moderating factors including the user's traits. Currently, there is a dearth of research on the effect of user traits on the results of gamification. Gamer types (personality traits related to gaming styles and preferences) are considered some of the most relevant factors affecting the individual's susceptibility to gamification. Therefore, in this study we investigate how gamer types from the BrainHex taxonomy (achiever, conqueror, daredevil, mastermind, seeker, socializer and survivor) moderate the effects of personalized/non-personalized gamification on users' flow experience (challenge-skill balance, merging of action and awareness, clear goals, feedback, concentration, control, loss of self-consciousness and autotelic experience), enjoyment, perception of gamification and motivation. We conducted a mixed factorial within-subject experiment involving 121 elementary school students comparing a personalized version against a non-personalized version of a gamified education system. There were no main effects between personalization and students' flow experience, perception of gamification and motivation, and enjoyment. Our results also indicate patterns of characteristics that can lead students to the high flow experience (e.g., those who prefer to play multiplayer have a high flow experience in both personalized and non-personalized versions). Based on our results, we provided recommendations to advance the design of gamifed educational systems. [ABSTRACT FROM AUTHOR]

The paper considers a problem of extracting user preferences based on their photo gallery. We propose a novel approach based on image captioning, i.e., automatic generation of textual descriptions of photos, and their classification. Known image captioning methods based on convolutional and recurrent (Long short-term memory) neural networks are analyzed. We train several models that combine the visual features of a photograph and the outputs of an Long short-term memory block by using Google's Conceptual Captions dataset. We examine application of natural language processing algorithms to transform obtained textual annotations into user preferences. Experimental studies are carried out using Microsoft COCO Captions, Flickr8k and a specially collected dataset reflecting the user’s interests. It is demonstrated that the best quality of preference prediction is achieved using keyword search methods and text summarization from Watson API, which are 8 % more accurate compared to traditional latent Dirichlet allocation. Moreover, descriptions generated by trained neural models are classified 1 – 7 % more accurately when compared to known image captioning models.

The advent of social networks and micro-blogging sites online has led to an abundance of user-generated content. Hence, the enormous amount of content is viewed as inappropriate and unimportant information by many users on social media. Therefore, there is a need to use personalization to select information related to users' interests or searchers on social media platforms. Therefore, in recent years, user interest mining has been a prominent research area. However, almost all of the emerging research suffers from significant gaps and drawbacks. Firstly, it suffers from focusing on the explicit content of the users to determine the interests of the users while neglecting the multiple facts as the personality of the users; demographic data may be a valuable source of influence on the interests of the users. Secondly, existing work represents users with their interesting topics without considering the semantic similarity between the topics based on clusters to extract the users' implicit interests. This paper is aims to propose a novel user interest mining approach and model based on demographic data, big five personality traits and similarity between the topics based on clusters. To demonstrate the leverage of combining user personality traits and demographic data into interest investigation, various experiments were conducted on the collected data. The experimental results showed that looking at personality and demographic data gives more accurate results in mining systems, increases utility, and can help address cold start problems for new users. Moreover, the results also showed that interesting topics were the dominant factor. On the other hand, the results showed that the current users' implicit interests can be predicted through the cluster based on similar topics. Moreover, the hybrid model based on graphs facilitates the study of the patterns of interaction between users and topics. This model can be beneficial for researchers, people on social media, and for certain research in related fields. [ABSTRACT FROM AUTHOR]

When conducting a search task, users may find it difficult to articulate their need, even more so when the task is complex. To help them complete their search, search engine usually provide query suggestions. A good query suggestion system requires to model user behavior during the search session. In this article, we study multiple Transformer architectures applied to the query suggestion task and compare them with recurrent neural network (RNN)-based models. We experiment Transformer models with different tokenizers, with different Encoders (large pretrained models or fully trained ones), and with two kinds of architectures (flat or hierarchic). We study the performance and the behaviors of these various models, and observe that Transformer-based models outperform RNN-based ones. We show that while the hierarchical architectures exhibit very good performances for query suggestion, the flat models are more suitable for complex and long search tasks. Finally, we investigate the flat models behavior and demonstrate that they indeed learn to recover the hierarchy of a search se [ABSTRACT FROM AUTHOR]

User engagement is crucial to the long-term success of a mobile app. Several metrics, such as dwell time, have been used for measuring user engagement. However, how to effectively predict user engagement in the context of mobile apps is still an open research question. For example, do the mobile usage contexts (e.g., time of day) in which users access mobile apps impact their dwell time? Answers to such questions could help mobile operating system and publishers to optimize advertising and service placement. In this article, we first conduct an empirical study for assessing how user characteristics, temporal features, and the short/long-term contexts contribute to gains in predicting users’ app dwell time on the population level. The comprehensive analysis is conducted on large app usage logs collected through a mobile advertising company. The dataset covers more than 12K anonymous users and 1.3 million log events. Based on the analysis, we further investigate a novel mobile app engagement prediction problem—can we predict simultaneously what app the user will use next and how long he/she will stay on that app? We propose several strategies for this joint prediction problem and demonstrate that our model can improve the performance significantly when compared with the state-of-the-art baselines. Our work can help mobile system developers in designing a better and more engagement-aware mobile app user experience. [ABSTRACT FROM AUTHOR]

Gamification has been discussed as a standout approach to improve user experience, with different studies showing that users can have different preferences over game elements according to their user types. However, relatively less is known how different kinds of users may react to different types of gamification. Therefore, in this study ( N = 331 ) we investigate how user orientation (Achiever, Disruptor, Free Spirit, Philanthropist, Player, and Socializer) is associated with the preference for and perceived sense of accomplishment from different gamification designs. Beyond singular associations between the user orientation and the gamification designs, the findings indicate no comprehensive and consistent patterns of associations. From the six user orientations, five presented significant associations: Socializer orientation was positively associated with Social, Fictional, and Personal designs, while negatively associated with Performance design; Player orientation was positively associated with Social (Accomplishment), Personal, and Ecological designs, while negatively associated with the Social design (Preference); Disruptor orientation was positively associated with Social design; Achiever orientation was positively associated with Performance and Social designs; and Free Spirit orientation was negatively associated with Social design. Based on the results, we provide recommendations on how to personalize gamified systems and set further research trajectories on personalized gamification. [ABSTRACT FROM AUTHOR]

Event‐based social networks (EBSNs) have become increasingly popular, which provide online social event management platforms for event organizers to publish and share social events (e.g., outdoor activities). In EBSNs, a major challenge for a social event organizer is how to plan a social event to attract the maximum number of attendance. To organize an event, three essential elements are required, namely, what (i.e., event content), where (i.e., event location), and when (i.e., event time). In this paper, we focus on the social event planning problem, which selects a location and time to hold a social event for the organizer with the given event content, to maximize the total number of participants. The solution of the social event planning problem could support decision‐making for social event organizers. For simplicity, we denote a location and time pair as an item in this paper. To solve the social event planning problem, we present a hybrid pairwise Markov random field (H‐PMRF) model which takes latent preferences of users, latent attributes of items, similarities between users and similarities between items into consideration. In particular, we construct an undirected graph where each node represents a user's decision on a specific item and each edge represents the relationship between the nodes, define the node potentials and edge potentials which model the dependency relationships between nodes, and give a joint probability distribution over the graph. Further, we adopt the Loopy Belief Propagation algorithm to compute the posterior probability distribution of each node in H‐PMRF and select the location and time to hold the event which could attract the maximum number of participants. We collect real‐world data set from DoubanEvent website and conduct extensive experiments on it. Experimental results show that the proposed model outperforms several baselines. [ABSTRACT FROM AUTHOR]

The emergence of mobile terminals operating at millimeter-wave frequencies necessitates the ability to evaluate the effect the environment, and in particular, users have on their radiation properties. Some studies evaluated the shadowing effects of a hand or an entire body for simple antenna configurations. This letter proposes a method for reliably predicting the performance of different array geometries in the presence of the users when they operate the mobile with one or with two hands. In practice, the way a mobile is operated is varying strongly between users, and hence, it is of great interest to draw a methodology to both numerically and experimentally evaluate any handset design in a large number of use cases in a repeatable manner. The use of numerical models and realistic phantoms allow high repeatability when evaluating the terminal radiation under real use conditions. Both the simulated body and the human phantom are used to study the field scattering from the handset arrays subject to the user interaction, yielding consistent results between them. Results suggest that shadowing by the user’s torso usually decreases gain between 20–30 dB close to the region of the user. The user posture largely affects the spherical coverage, particularly for those antennas close to the corners in a two-hand mode. [ABSTRACT FROM AUTHOR]

Recommender systems have been applied in several areas, including e-Health systems, which refers to information and health services enhanced through technology. However, most studies aim at imposing rules to improve lifestyle, rather than recommending nutrition and physical activities. In this context, this study aims to develop a system for recommending physical activities for hypertensive patients to create opportunities for the patients so they can search for and create a healthy lifestyle. To achieve this goal, we elaborated on a hypertensive user profile model, called HyperModel2PAR, and a physical activity recommender system for hypertensive patients, called HyperRecSysPA. The model resulting from this study is composed of 32 elements divided into three groups, which were used in the modeling of user profiles within the system for generating HyperRecSysPA recommendations. The developed system was validated by physicians who answered a specific questionnaire. As a result, ~75% of the recommendations generated were approved. Therefore, this study has prospective contributions to the literature, since both models obtained conclusive results in the assessments performed.

Cultural heritage (CH) institutions attract wide and heterogeneous audiences, which should be efficiently supported and have access to meaningful CH content. This introduces numerous challenges when delivering such experiences, given that people have different cognitive characteristics which influence the way we process information, experience, behave, and acquire knowledge. Our recent studies provide evidence that human cognition should be considered as a personalization factor within CH contexts, and thus we developed a framework that delivers cognition-centered personalized CH activities. The efficiency and the efficacy of the framework have been successfully assessed through two user studies, but non-technical professionals (e.g., CH designers) may face difficulties when attempting to use it and create personalized CH activities. In this paper, we present DeCACHe, which supports CH designers in creating cognition-centered personalized CH activities throughout different phases of the design lifecycle. We also report a user study with seventeen professional CH designers, who used our tool to design CH activities for people with different cognitive characteristics.

One of the travelers’ main challenges is that they have to spend a great effort to find and choose the most desired travel offer(s) among a vast list of non-categorized and non-personalized items. Recommendation systems provide an effective way to solve the problem of information overload. In this work, we design and implement “The Hybrid Offer Ranker” (THOR), a hybrid, personalized recommender system for the transportation domain. THOR assigns every traveler a unique contextual preference model built using solely their personal data, which makes the model sensitive to the user’s choices. This model is used to rank travel offers presented to each user according to their personal preferences. We reduce the recommendation problem to one of binary classification that predicts the probability with which the traveler will buy each available travel offer. Travel offers are ranked according to the computed probabilities, hence to the user’s personal preference model. Moreover, to tackle the cold start problem for new users, we apply clustering algorithms to identify groups of travelers with similar profiles and build a preference model for each group. To test the system’s performance, we generate a dataset according to some carefully designed rules. The results of the experiments show that the THOR tool is capable of learning the contextual preferences of each traveler and ranks offers starting from those that have the higher probability of being selected.

When designing rule-based dialogue systems, the need for the creation of an elaborate design by the designer is a challenge. One way to reduce the cost of creating content is to generate utterances from data collected in an objective and reproducible manner. This study focuses on rule-based dialogue systems using survey data and, more specifically, on opinion dialogue in which the system models the user. In the field of opinion dialogue, there has been little study on the topic of transition methods for modeling users while maintaining their motivation to engage in dialogue. To model them, we adopted information content. Our contribution includes the design of a rule-based dialogue system that does not require an elaborate design. We also reported an appropriate topic transition method based on information content. This is confirmed by the influence of the user’s personality characteristics. The content of the questions gives the user a sense of the system’s intention to understand them. We also reported the possibility that the system’s rational intention contributes to the user’s motivation to engage in dialogue with the system.

Personality accounts for how individuals differ in their enduring emotional, interpersonal, experiential, attitudinal and motivational styles. Personality, especially in the form of the Five Factor Model, has shown usefulness in personalized systems, such as recommender systems. In this work, we focus on a personality model that is targeted at motivations for multimedia consumption. The model is composed of two dimensions: the (i) eudaimonic orientation of users (EO) and (ii) hedonic orientation of users (HO). While the former accounts for how much a user is interested in content that deals with meaningful topics, the latter accounts for how much a user is interested in the entertaining quality of the content. Our research goal is to devise a model that predicts the EH and HO of users from interaction data with movies, such as ratings. We collected a dataset of 350 users, 703 movies and 3499 ratings. We performed a comparison of various predictive algorithms, as both regression and classification problems. Finally, we demonstrate that our proposed approach is able to predict the EO and HO of users from traces of interactions with movies substantially better than the baseline approaches. The outcomes of this work have implications for exploitation in recommender systems.

Understanding user’s behavior and their interactions with artificial-intelligent-based systems is as important as analyzing the performance of the algorithms used in these systems. For instance, in the Recommender Systems domain, the accuracy of the recommendation algorithm was the ultimate goal for most systems designers. However, researchers and practitioners have realized that providing accurate recommendations is insufficient to enhance users’ acceptance. A recommender system needs to focus on other factors that enhance its interactions with the users. Recent researches suggest augmenting these systems with persuasive capabilities. Persuasive features lead to increasing users’ acceptance of the recommendations, which, in turn, enhances users’ experience with these systems. Nonetheless, the literature still lacks a comprehensive view of the actual effect of persuasive principles on recommender users. To fill this gap, this study diagnoses how users of different characteristics get influenced by various persuasive principles that a recommender system uses. The study considers four users’ aspects: age, gender, culture (continent), and personality traits. The paper also investigates the impact of the context (or application domain) on the influence of the persuasive principles. Two application domains (namely eCommerce and Movie recommendations) are considered. A within-subject user study was conducted. The analysis of (279) responses revealed that persuasive principles have the potential to enhance users’ experience with recommender systems. The study also shows that, among the considered factors, culture, personality traits, and the domain of recommendations have a higher impact on the influence of persuasive principles than other factors. Based on the analysis of the results, the study provides insights and guidelines for recommender systems designers. These guidelines can be used as a reference for designing recommender systems with users’ experience in mind. We suggest that considering the results presented in this paper could help to improve recommender-users interaction.

Across a wide variety of domains, artificial agents that can adapt and personalize to users have potential to improve and transform how social services are provided. Because of the need for personalized interaction data to drive this process, long-term (or longitudinal) interactions between users and agents, which unfold over a series of distinct interaction sessions, have attracted substantial research interest. In recognition of the expanded scope and structure of a long-term interaction, researchers are also adjusting the personalization models and algorithms used, orienting toward “continual learning” methods, which do not assume a stationary modeling target and explicitly account for the temporal context of training data. In parallel, researchers have also studied the effect of “multitask personalization,” an approach in which an agent interacts with users over multiple different tasks contexts throughout the course of a long-term interaction and learns personalized models of a user that are transferrable across these tasks. In this paper, we unite these two paradigms under the framework of “Lifelong Personalization,” analyzing the effect of multitask personalization applied to dynamic, non-stationary targets. We extend the multi-task personalization approach to the more complex and realistic scenario of modeling dynamic learners over time, focusing in particular on interactive scenarios in which the modeling agent plays an active role in teaching the student whose knowledge the agent is simultaneously attempting to model. Inspired by the way in which agents use active learning to select new training data based on domain context, we augment a Gaussian Process-based multitask personalization model with a mechanism to actively and continually manage its own training data, allowing a modeling agent to remove or reduce the weight of observed data from its training set, based on interactive context cues. We evaluate this method in a series of simulation experiments comparing different approaches to continual and multitask learning on simulated student data. We expect this method to substantially improve learning in Gaussian Process models in dynamic domains, establishing Gaussian Processes as another flexible modeling tool for Long-term Human-Robot Interaction (HRI) Studies.