Personalized scientific literature recommendation based on user's research interest

User interest and behavior modeling is a critical step in online digital advertising. On the one hand, user interests directly impact their response and actions to the displayed advertisement (Ad). On the other hand, user interests can further help determine the probability of an Ad viewer becoming a buying customer. To date, existing methods for Ad click prediction, or click-through rate prediction, mainly consider representing users as a static feature set and train machine learning classifiers to predict clicks. Such approaches do not consider temporal variance and changes in user behaviors, and solely rely on given features for learning. In this paper, we propose two deep learning-based frameworks, {hbox {LSTM}}_{mathrm{cp}} and {hbox {LSTM}}_{mathrm{ip}}, for user click prediction and user interest modeling. Our goal is to accurately predict (1) the probability of a user clicking on an Ad and (2) the probability of a user clicking a specific type of Ad campaign. To achieve the goal, we collect page information displayed to the users as a temporal sequence and use long short-term memory (LSTM) network to learn features that represents user interests as latent features. Experiments and comparisons on real-world data show that, compared to existing static set-based approaches, considering sequences and temporal variance of user requests results in improvements in user Ad response prediction and campaign specific user Ad click prediction.

Search engines usually return relevant sorted results based on the keywords. Because of the lack of considering the user's current search interest and intention, this kind of strategy may not meet users' personalized search requirements. In order to retrieve results associated with the user's current search interest, existing researches integrate the user's interest information into the search process and ranking, so as to make the search results fit to the user's current search target more better. Different from the unified model of interest which most current related researches used, we propose a model of combing user's interests for the purpose of ranking in the light of different query types. This paper first classified queries into several types based on the current keywords and the sources of user's interests, such as short-term interest in user clicks, and long-term interest in user's search history, search logs of the search system etc. Then based on the sources of interests, we established a model of combination of interests which based on the artificial neural network and was trained by related approaches and data sets. Depending on different query types, we combined these sources of user's interests that the most best reflecting the user's current search intention to enhance the user's search experience. Comparing with the existing model of single interest, the results of the experiments show that our model can obtain better ranking on the results.

Contributors

The rapid development of Community Question Answering (CQA) satisfies users' request for professional and personal knowledge. In CQA, one key issue is to recommend users with high expertise and willingness to answer the given questions, namely expert recommendation. However, most of existing methods for expert recommendation ignore some key information, such as time information and historical feedback information, degrading the performance. On the one hand, users' interest are changing over time. It is biased if we don't consider the dynamics. On the other hand, feedback information is critical to estimate users' expertise. To solve these problems, we propose a unified framework for expert recommendation to exploit user interest and expertise more precisely. Considering the inconsistency between them, we propose to learn their embeddings separately. We leverage Long Short-Term Memory (LSTM) to model user's short-term interest and combine it with long-term interest. The user expertise is learned by the designed user expertise network, which explicitly models feedback on users' historical behavior. The extensive experiments on a large-scale dataset from a realworld CQA site demonstrate the superior performance of our method than state-of-the-art solutions to the problem.

In the era of digitization, online digital advertising is one of the best techniques for modern marketing. This makes advertisers rely heavily on accurate user interest and behavior modelling to deliver precise advertisement impressions and increase click-through rates. The classic approach to model user interests has often required the use of predefined feature sets which are typically stagnant and not representative of temporal changes and trends in user behavior. While recent advances in deep learning offer potential solutions to these obstacles, many existing approaches fail to address the sequential nature of user interactions. In this paper, we propose an optimized Bi-Directional Long Short-Term Memory (Bi-LSTM) based user interest modeling approach together with an Updated version of Parrot Optimizer (UPO) to enhance performance. It treats the user activity as a temporal sequence which well fits the changing nature of user interest and preferences over time. The proposed approach is evaluated on two important tasks: predicting the probability that a user will click on an ad and predicting the probability that a user will click on a particular type of ad campaign. Simulation results demonstrate that the proposed method provides superior results than the static set-based approaches and achieves significant improvements on both user ad responses predictions and user ad clicks at the campaign level. The research also enhances the efficiency of user interest modeling with significant implications for online advertising, recommendation systems, and personalized marketing.

Focus on Authors

Sequential basket analysis is a challenging task that focuses on modeling user interests through their shopping basket records. This study focuses on a newly identified bias: the repetition bias , which typically arises due to repurchase behavior . Existing methods typically oversimplify the relationship between repetitions and predictions. They assume that frequent repetition of an item by a user indicates a strong preference of the user. However, this assumption is flawed as it fails to consider that repetitions are not driven solely by user interests, as they can also be influenced by external factors, resulting in a biased understanding of user interests. In this article, we propose the CA usal intervention for R epetition D e-biasing ( CARD ), a novel solution to comprehensively estimate various influencing factors and address the repetition bias, thereby ensuring a more accurate learning of user interests. Specifically, we design a multi-factor estimation debiasing framework with constructed causal graphs to formalize the data generation process within the recommendation. We then analyze the variables that influence the recommendation, with the goal of identifying confounding variables that affect repurchase behavior and thereby locating the source of repetition bias. Since repetition bias originates from the influence of confounding variables on repurchase behavior, we resort to causal intervention methods to prevent its impacts and thus eliminate repetition bias at its source for unbiased user interest modeling. We evaluate CARD by conducting extensive experiments over three real-world datasets. The results demonstrate our approach’s competitiveness over the representative state-of-the-art baselines in achieving unbiased user interest modeling.

Traditional teaching course recommendation methods mainly rely on user satisfaction or utility evaluation of courses. In web front-end teaching course recommendation, the personalized learning needs of users are ignored, resulting in poor quality of recommended courses. Therefore, to solve the problem of personalized course recommendation, this paper proposes a personalized course recommendation method based on implicit feedback. This method is based on user interaction behavior analysis, implicit feedback data is collected, including multidimensional data such as browsing history, click behavior, and dwell time, and stored in a relational database. Extract user learning features and group user browsing records to discover user groups with similar browsing behaviors. Build a knowledge graph for web-based teaching courses, combined with user interest models, integrate course knowledge points, associations, etc. into recommendation methods, generate personalized recommendation lists, and this approach effectively enhances recommendation accuracy and diversity. To verify the performance of the proposed recommendation method, a control group was set up to conduct comparative analysis based on indicators such as recommendation novelty, recommendation coverage, and normalized cumulative loss gain (NDCG). Experimental results demonstrate the novelty and coverage of recommendations are much higher than traditional methods, and it can cover a wider range of courses. Meanwhile, as the length of the recommendation list increases, the NDCG value maintains a stable growth, proving that this method can more accurately capture user interests and generate high-quality personalized recommendation lists. In future research, the sample size will be expanded to more comprehensively validate the effectiveness of recommendation methods in different user groups.

Digital map services and search services for geographical information are widely available on the Internet. Users can retrieve suitable geographical information using these digital maps through certain map operations and category selections. However, when a map region includes a large amount of information, it can be difficult for users to find specific geographical information. In addition, a user's interest in a region or a category may change. Therefore, we propose a method for recommending geographical information to users on the basis of their map operation or category selection history. We develop a model for determining a user's interest and use it to recommend suitable regions and categories.

Contributors

A major shortcoming of content-based approaches exists in the representation of the user model. Content-based approaches often employ term vectors to represent each user's interest. In doing so, they ignore the semantic relations between terms of the vector space model in which indexed terms are not orthogonal and often have semantic relatedness between one another. In this paper, we improve the representation of a user model during building user model in content-based approaches by performing these steps. First is the domain concept filtering in which concepts and items of interests are compared to the domain ontology to check the relevant items to our domain using ontology based semantic similarity. Second, is incorporating semantic content into the term vectors. We use word definitions and relations provided by WordNet to perform word sense disambiguation and employ domain-specific concepts as category labels for the semantically enhanced user models. The implicit information pertaining to the user behavior was extracted from click stream data or web usage sessions captured within the web server logs. Also, our proposed approach aims to update user model, we should analysis user's history query keywords. For a certain keyword, we extract the words which have the semantic relationships with the keyword and add them into the user interest model as nodes according to semantic relationships in the WordNet. Keywords-User model; Domain ontology; Semantic Similarity; Wordnet.

In the study of scholars, the research interests of scholars have strong personalized features, which can well reflect the interest preferences of scholars in specific research fields to serve the tasks of user profiling and academic recommendation of scholars. Research interest mining aims to mine topics that can be used to describe scholars' research interests directly from the text. The existing methods lack the comprehensive use of semantic information of academic papers and academic network information. In this paper, we construct a research interest mining method for scholars based on the text of their papers and citation networks. On the one hand, the LSI model and Doc2Vec model are used for text representation of scholars' papers, which can effectively extract the topics of papers and mine the semantic information in the text; on the other hand, the MinHash algorithm is used to reduce the complexity of citation network similarity calculation. Finally, the integration method is used for fusion of interest labels to obtain the final scholars' research interest labels. The experimental results demonstrate that the integrated approach can get higher accuracy rates.

Microblog, like Twitter and Sina Weibo, produce large content everyday by millions of user, which reflect user interest. Grasping user interest is important for content recommendation and ad targeting. In this paper, We take a novel method to predict user interest by using automatic topic learning and recommendation method. We crawl lots of user tweet data from Weibo, and use Latent Dirichlet Allocation (LDA) topic model to exact topics. We assign each tweet to one topic, then get a User-Interest matrix by accumulating each user's preference for topics. Finally we use singular value decomposition (SVD) method to predict user preference for the topics that user may interest. We evaluate our method on test data and get state of the art performance.

New technologies are constantly being introduced in radiation oncology primarily because they are new and not because they are clearly better than the technologies they are replacing. Often there is a “belief” that the new technology “ought” to be better but many believe that they should be widely adopted in the clinic only after evidence has shown that they are at least as safe and efficacious as existing technologies, which are often less expensive. This is the concern debated in this month's Point/Counterpoint. Arguing for the Proposition is Christopher Njeh, Ph.D. Dr. Njeh obtained his Ph.D. degree in Medical Physics from Sheffield Hallam University, UK, and, after graduation, he worked at the Addenbrooke's Hospital in Cambridge and Queen Elizabeth's Hospital, Birmingham, UK. He then came to the United States as a Visiting Postdoctoral Fellow at the University of California, San Francisco, where he was subsequently appointed as an Assistant Professor of Radiology. He later completed a Medical Physics residency at Johns Hopkins University, Baltimore, and is currently Chief Medical Physicist at Texas Oncology in Tyler, TX, and holds an adjunct faculty position at the University of Texas at Tyler. Dr. Njeh is certified in Therapeutic Radiologic Physics by the ABR. His major research interests include image-guided radiation therapy and accelerated partial breast irradiation. He is author or coauthor of over 50 papers and 10 book chapters and is coeditor of two books. Arguing against the Proposition is Dr. Christian McDonald Langton. Dr. Langton obtained his M.Sc. degree in Medical Physics from the University of Aberdeen and his doctoral degrees from the University of Hull. After working in industry for two years, he returned to academia in the UK and ultimately attained the rank of Professor of Medical Physics at the University of Hull. In 2008, Dr. Langton was appointed Professor of Medical Physics at Queensland University of Technology in Brisbane, Australia, and Director of the Queensland Cancer Physics Collaborative. Dr. Langton's main research interest has been in quantitative bone imaging and characterization, and his work on the science, technology, and clinical utility of ultrasound assessment of cancellous bone and osteoporosis has resulted in over 1800 publication citations. He holds several related patents and there are seven commercial devices currently available adopting his broadband ultrasonic attenuation technique, with over 12 000 systems utilized worldwide. Medical technology encompasses all drugs, devices, and medical and surgical procedures used in medical care as well as the organizational supportive systems within which such care is provided.1 Radiation oncology has recently witnessed an explosion in innovation including but not limited to: proton therapy, CyberKnife, tomotherapy, IGRT, and IMRT. Efficacy, safety, and cost effectiveness, however, remains the focus in the provision of optimal care to patients. While some of these innovations offer unprecedented breakthroughs for some patients, they have the potential to also result in unintended harm if not used appropriately. It is, therefore, essential that adoption of these new technologies be evidence based. Evidence-based medicine can be perceived as “the conscientious, explicit and judicious use of current best evidence in making decisions about the care of individual patients.”2 Its practice presupposes the integration of individual clinical expertise with the best available external clinical evidence from systematic collection and synthesis of data, including patients’ values and expectations.3 The gold standard for the attainment of level 1 evidence is usually through randomized controlled trials (RCTs) and meta-analysis of such trials. There are many reported instances in the scientific literature where RCTs refuted evidence from theoretical, observational, physiologic studies or common sense. In the 1890s, Dr. William Halsted, for instance, developed radical mastectomy for breast cancer. His procedure was performed unchallenged for over 80 years. It was, however, not until an RCT was conducted in the late 1980s that it dawned upon the scientific community that radical mastectomy had no advantage over simpler forms of treatment for early-stage breast cancer.4 Another more recent example is vascular brachytherapy that was used to treat in-stent restenosis until an RCT showed that this therapy yielded comparatively inferior outcomes to polymer-based slow-release paclitaxel-eluting stents.5 Opponents of RCTs may be surprised to learn that not all new therapies amount to an improvement compared to the standard of therapy. For example, an analysis of outcome data from 58 RCTs, including a total of 12734 patients, conducted between 1968 and 2002 by the Radiation Therapy Oncology Group, found that, overall, experimental and standard arms were equally successful.6 They also found that treatment-related mortality and morbidity were, on average, higher in the innovative arm. In contemplating these facts, we are reminded of the economist who once said “…man's wants are numerous but his means are limited.” His view is applicable to national health care. Were one to place the cost of new technology into proper context it would be safe to assert that U.S. health care costs have risen faster than the gross domestic product (GDP), often by a substantial margin. While in 1960 approximately 5% of the GDP of the United States was spent on medical care,1 by 2004 it accounted for over 15% and is expected to be as much as 20% of GDP by 2015.7 According to a landmark study by The Kaiser Family Foundation,8 new technology has been identified as one of the causes of this exponential rise in health care cost. I submit, therefore, that new technology needs to fulfill the triple condition of efficacy, safety, and cost-effectiveness so that our limited resources can be put to the most judicious use. Evidence-based medicine is commonly defined as “the conscientious, explicit, and judicious use of current best evidence in making decisions about the care of individual patients.”9 These are admirable words, but what do they really mean? Are they a realistic aspiration for adoption of new technologies in radiation oncology? A key factor is the validity of the evidence, so variable in reality that a number of category levels are widely utilized. For example, the U.S. Preventative Services Task Force10 lists three levels of “quality of evidence,” the highest level being “evidence obtained from at least one properly designed randomized controlled trial.” For the laudable randomized clinical trial, there has been a dramatic expansion in the number of publications associated with “radiation oncology” alone. Considering “Web of Science” publications per year using the topic search category of (“RCT” or “randomized clinical trial” or “randomized control trial”) and (“radiation oncology” or “radiotherapy”) yields: 1989 = 2, 1994 = 32, 1999 = 55, 2004 = 107, and 2009 = 228. How do we arrive at a consensus based upon such a wealth of information? Will a comparison be made against an untarnished “gold standard” or with current practice? The latter will inevitably necessitate large cohort numbers in each study arm, often impracticable from a recruitment perspective. Is there potential for a high attrition rate? Will it be difficult to assign and maintain inclusion and exclusion criteria? Will it be difficult to avoid bias? Very few cases are truly “equivalent.” Noting the understandable need to maintain a primary focus on patient welfare, as circumstances potentially change, will it be difficult to maintain a rigid protocol? Will the protocol be readily and reliably transferable multicenter and multinational? How long will it be before the technique could be routinely adopted? Technology developments appear at a fast time rate and may evolve during the course of an RCT such that they are used differently at the end of a trial than at the beginning and might even become outdated before the trial is over. How important are factors such as quality of life and secondary cancer risk? Other criticisms of adopting “evidence-based medicine” include stagnation, bland uniformity, and lowering of standards through deskilling practitioners. Instead of using clinical judgment, they will be encouraged to follow protocols that treat all patients as essentially interchangeable.9 There is also a threat to the adoption of new techniques in radiation oncology through a growing movement of “lack-of-evidence based medicine” that has been used to restrict access to a number of therapies,11 particularly by the UK's National Institute of Clinical Excellence. Perhaps the hottest “new technology in radiation oncology” debate relates to proton verses photon IMRT, with a question raised as to whether large randomized phase III comparative trials should be performed?; that would inherently encompass a significant number of scientific and ethical issues11,12 — I will end with that thought! The sheer volume of information available in the literature is more reason for a unified and systematic approach to synthesize them. Recruitment bias can be avoided by proper randomization.13 EBM makes decision making more thoughtful and more transparent, providing a stronger scientific backbone to medical practice. Not all studies are carried out with the same degree of rigor (quality, quantity, and consistency) hence a need to grade the quality of the research such as required by SORT,14 GRADE,15 or the Center for Evidence Based Medicine. The FDA is not thorough enough in its technology approval process. Recent studies have shown that the FDA premarket approval process is often based on weak studies.16 Ethical dilemmas in RCTs are eliminated by the acknowledgement of the equipoise principle which assumes that the two arms in a study have an equal chance of performing well.17 This principle has been validated by the fact that only 25%–50% of new technology is better than traditional technology.18 Another issue with pursuing RCTs has to do with the vested interests of three players if the RCT proves that the procedure is ineffective: the physician (new technology is accompanied by higher reimbursement), the hospital (need to pay for the equipment), and the manufacturer (need to make a profit).19 In conclusion, obtaining the relevant high quality evidence is a challenging, demanding, time-consuming, and costly pursuit. Nevertheless, it is a rigorous process, which we must demand of new technology so as to remain accountable to our patients. Having carefully considered my opponent's Opening Statement, I am confident that the arguments provided within my own Opening Statement remain valid and wholly intact. There are two primary components of my opposition to the proposition. First, it is impossible in reality to acquire irrefutable evidence as to whether a “new technology” will indeed improve individual patient care. Second, it is impossible to create a single unifying consensus based upon reported data. Allied to this, there has been a dramatic increase in the number of evidence-based medicine derived “clinical guidelines,” so much so that there has been a call for “guidelines for clinical guidelines” within a British Medical Journal Editorial.20 A fundamental question that we must address is whether so-called “evidence-based medicine” serves its purpose of helping clinicians make better decisions for the individual patient; related not only to the primary factors of diagnosis and treatment, but also encompassing prognosis, benefit, risk, and cost. Continuing this somewhat broader perspective, I wish to consider another component of the proposition's title, specifically, what do we really mean by “new technology in radiation oncology”? Are many so-called “new technologies” simply part of a fundamentally evolutionary advancement process? Noting the age-old tenet of “maximally treating the cancer by maximally sparing normal tissue,” are we in danger of being distracted away from aspects of radiation oncology that are of greater importance from an individual patient's perspective? For example, are we in need of true “new technologies” that better target regions of a particular tumor or organ that require, or maybe do not require, “treatment;” and to determine how these relate to both static and temporal anatomy? In summary, while appreciating the ideological paradigm that “the adoption of new technology in radiation oncology should rely on evidence-based medicine,” in reality, this cannot be achieved and we should concentrate on the primary role of helping clinicians make better decisions for the individual patient.

In online business, it is important to construct sale web pages offering attractive services for popular products in order to improve page access as well as purchase rates. Moreover, online shop owners need to hold various types of (campaign) sales frequently throughout the year to keep customers coming back. Also, online shopping systems have to adapt to its circumstances such as customers' needs and the surrounding economic situations. To cope with these requirements, a self-organized IoT (Internet of Things) aware systems is proposed for online shopping. In our approach, awareness is achieved by monitoring/analyzing the data of user's behavior, in particular the eye gaze. The particular focus here is the analysis of the users' eye gaze with the objective to derive insights about products the user is interested in or not and adapt the system accordingly by presenting other products with similar attributes, hiding products with less interesting attributes and building a user's interest model by integrating all the positive preferences