Trust-based automated recommendation making

Recommender systems are one of the recent inventions to deal with ever growing information overload in relation to the selection of goods and services in a global economy. Collaborative Filtering (CF) is one of the most popular techniques in recommender systems. The CF recommends items to a target user based on the preferences of a set of similar users known as the neighbours, generated from a database made up of the preferences of past users. With sufficient background information of item ratings, its performance is promising enough but research shows that it performs very poorly in a cold start situation where there is not enough previous rating data. As an alternative to ratings, trust between the users could be used to choose the neighbour for recommendation making. Better recommendations can be achieved using an inferred trust network which mimics the real world "friend of a friend" recommendations. To extend the boundaries of the neighbour, an effective trust inference technique is required. This thesis proposes a trust interference technique called Directed Series Parallel Graph (DSPG) which performs better than other popular trust inference algorithms such as TidalTrust and MoleTrust. Another problem is that reliable explicit trust data is not always available. In real life, people trust "word of mouth" recommendations made by people with similar interests. This is often assumed in the recommender system. By conducting a survey, we can confirm that interest similarity has a positive relationship with trust and this can be used to generate a trust network for recommendation. In this research, we also propose a new method called SimTrust for developing trust networks based on user's interest similarity in the absence of explicit trust data. To identify the interest similarity, we use user's personalised tagging information. However, we are interested in what resources the user chooses to tag, rather than the text of the tag applied. The commonalities of the resources being tagged by the users can be used to form the neighbours used in the automated recommender system. Our experimental results show that our proposed tag-similarity based method outperforms the traditional collaborative filtering approach which usually uses rating data.

Personalized recommender systems integrating tags and item taxonomy

The social tags in Web 2.0 are becoming another important information source to profile users' interests and preferences to make personalized recommendations. To solve the problem of low information sharing caused by the free-style vocabulary of tags and the long tails of the distribution of tags and items, this paper proposes an approach to integrate the social tags given by users and the item taxonomy with standard vocabulary and hierarchical structure provided by experts to make personalized recommendations. The experimental results show that the proposed approach can effectively improve the information sharing and recommendation accuracy.

A combined method for mitigating sparsity problem in tag recommendation

Tag recommendation is a specific recommendation task for recommending metadata (tag) for a web resource (item) during user annotation process. In this context, sparsity problem refers to situation where tags need to be produced for items with few annotations or for user who tags few items. Most of the state of the art approaches in tag recommendation are rarely evaluated or perform poorly under this situation. This paper presents a combined method for mitigating sparsity problem in tag recommendation by mainly expanding and ranking candidate tags based on similar items’ tags and existing tag ontology. We evaluated the approach on two public social bookmarking datasets. The experiment results show better accuracy for recommendation in sparsity situation over several state of the art methods.

Item reputation-aware recommender systems

Recommender systems provide personalized advice for customers online based on their own preferences, while reputation systems generate a community advice on the quality of items on the Web. Both systems use users’ ratings to generate their output. In this paper, we propose to combine reputation models with recommender systems to enhance the accuracy of recommendations. The main contributions include two methods for merging two ranked item lists which are generated based on recommendation scores and reputation scores, respectively, and a personalized reputation method to generate item reputations based on users’ interests. The proposed merging methods can be applicable to any recommendation methods and reputation methods, i.e., they are independent from generating recommendation scores and reputation scores. The experiments we conducted showed that the proposed methods could enhance the accuracy of existing recommender systems.

History-guided conversational recommendation

Product recommendation is an important aspect of many e-commerce systems. It provides an effective way to help users navigate complex product spaces. In this paper, we focus on critiquing-based recommenders. We present a new critiquing-based approach, History-Guided Recommendation (HGR), which is capable of using the recommendation pairs (item and critique) or critiques only so far in the current recommendation session to predict the most likely product recommendations and therefore short-cut the sometimes protracted recommendation sessions in standard critiquing approaches. The HGR approach shows a significant improvement in the interaction between the user and the recommender. It also enables successfully accepted recommendations to be made much earlier in the session

An ontological approach to personalized medical knowledge recommendation

Knowledge recommendation has become a promising method in supporting the clinicians decisions and improving the quality of medical services in the constantly changing clinical environment. However, current medical knowledge management systems cannot understand users requirements accurately and realize personalized recommendation. Therefore this paper proposes an ontological approach based on semiotic principles to personalized medical knowledge recommendations. In particular, healthcare domain knowledge is conceptualized and an ontology-based user profile is built. Furthermore, the personalized recommendation mechanism is illustrated.

A graph-based model for context-aware recommendation using implicit feedback data

Recommender systems have been successfully dealing with the problem of information overload. However, most recommendation methods suit to the scenarios where explicit feedback, e.g. ratings, are available, but might not be suitable for the most common scenarios with only implicit feedback. In addition, most existing methods only focus on user and item dimensions and neglect any additional contextual information, such as time and location. In this paper, we propose a graph-based generic recommendation framework, which constructs a Multi-Layer Context Graph (MLCG) from implicit feedback data, and then performs ranking algorithms in MLCG for context-aware recommendation. Specifically, MLCG incorporates a variety of contextual information into a recommendation process and models the interactions between users and items. Moreover, based on MLCG, two novel ranking methods are developed: Context-aware Personalized Random Walk (CPRW) captures user preferences and current situations, and Semantic Path-based Random Walk (SPRW) incorporates semantics of paths in MLCG into random walk model for recommendation. The experiments on two real-world datasets demonstrate the effectiveness of our approach.

Rarely selected distractors in high stakes medical multiple-choice examinations and their recognition by item authors: A simulation and survey

Background Many medical exams use 5 options for multiple choice questions (MCQs), although the literature suggests that 3 options are optimal. Previous studies on this topic have often been based on non-medical examinations, so we sought to analyse rarely selected, 'non-functional' distractors (NF-D) in high stakes medical examinations, and their detection by item authors as well as psychometric changes resulting from a reduction in the number of options. Methods Based on Swiss Federal MCQ examinations from 2005-2007, the frequency of NF-D (selected by <1% or <5% of the candidates) was calculated. Distractors that were chosen the least or second least were identified and candidates who chose them were allocated to the remaining options using two extreme assumptions about their hypothetical behaviour: In case rarely selected distractors were eliminated, candidates could randomly choose another option - or purposively choose the correct answer, from which they had originally been distracted. In a second step, 37 experts were asked to mark the least plausible options. The consequences of a reduction from 4 to 3 or 2 distractors - based on item statistics or on the experts' ratings - with respect to difficulty, discrimination and reliability were modelled. Results About 70% of the 5-option-items had at least 1 NF-D selected by <1% of the candidates (97% for NF-Ds selected by <5%). Only a reduction to 2 distractors and assuming that candidates would switch to the correct answer in the absence of a 'non-functional' distractor led to relevant differences in reliability and difficulty (and to a lesser degree discrimination). The experts' ratings resulted in slightly greater changes compared to the statistical approach. Conclusions Based on item statistics and/or an expert panel's recommendation, the choice of a varying number of 3-4 (or partly 2) plausible distractors could be performed without marked deteriorations in psychometric characteristics.

Modeling the evolution of item rating networks using time-domain preferential attachment

The understanding of the structure and dynamics of the intricate network of connections among people that consumes products through Internet appears as an extremely useful asset in order to study emergent properties related to social behavior. This knowledge could be useful, for example, to improve the performance of personal recommendation algorithms. In this contribution, we analyzed five-year records of movie-rating transactions provided by Netflix, a movie rental platform where users rate movies from an online catalog. This dataset can be studied as a bipartite user-item network whose structure evolves in time. Even though several topological properties from subsets of this bipartite network have been reported with a model that combines random and preferential attachment mechanisms [Beguerisse Díaz et al., 2010], there are still many aspects worth to be explored, as they are connected to relevant phenomena underlying the evolution of the network. In this work, we test the hypothesis that bursty human behavior is essential in order to describe how a bipartite user-item network evolves in time. To that end, we propose a novel model that combines, for user nodes, a network growth prescription based on a preferential attachment mechanism acting not only in the topological domain (i.e. based on node degrees) but also in time domain. In the case of items, the model mixes degree preferential attachment and random selection. With these ingredients, the model is not only able to reproduce the asymptotic degree distribution, but also shows an excellent agreement with the Netflix data in several time-dependent topological properties.

Letter of recommendation for Amos Main from Joseph Moodey to Nathan Prince, 1725 July 2

Handwritten letter sent by Joseph Moody, schoolmaster in York, to Harvard Tutor Nathan Prince recommending student Amos Main for acceptance to the College. In the letter, Moody requests Prince give Main an examination for admission, with the caveat that though Main has been studying Latin and Greek he has a difficult home life and is "somewhat Raw; yet I hope you'l wink at it." The letter, dated July 2, 1725, is written on a folded folio-sized leaf; there are handwritten notes about Massachusetts towns on the verso.

'On the Effects of Idiotypic Interactions for Recommendation Communities in Artificial Immune Systems'

It has previously been shown that a recommender based on immune system idiotypic principles can outperform one based on correlation alone. This paper reports the results of work in progress, where we undertake some investigations into the nature of this beneficial effect. The initial findings are that the immune system recommender tends to produce different neighbourhoods, and that the superior performance of this recommender is due partly to the different neighbourhoods, and partly to the way that the idiotypic effect is used to weight each neighbour's recommendations.

'Movie Recommendation Systems Using An Artificial Immune System'

We apply the Artificial Immune System (AIS)technology to the collaborative Filtering (CF)technology when we build the movie recommendation system. Two different affinity measure algorithms of AIS, Kendall tau and Weighted Kappa, are used to calculate the correlation coefficients for this movie recommendation system. From the testing we think that Weighted Kappa is more suitable than Kendall tau for movie problems.

'Movie Recommendation Systems Using An Artificial Immune System'

We apply the Artificial Immune System (AIS)technology to the collaborative Filtering (CF)technology when we build the movie recommendation system. Two different affinity measure algorithms of AIS, Kendall tau and Weighted Kappa, are used to calculate the correlation coefficients for this movie recommendation system. From the testing we think that Weighted Kappa is more suitable than Kendall tau for movie problems.

'On the Effects of Idiotypic Interactions for Recommendation Communities in Artificial Immune Systems'

It has previously been shown that a recommender based on immune system idiotypic principles can outperform one based on correlation alone. This paper reports the results of work in progress, where we undertake some investigations into the nature of this beneficial effect. The initial findings are that the immune system recommender tends to produce different neighbourhoods, and that the superior performance of this recommender is due partly to the different neighbourhoods, and partly to the way that the idiotypic effect is used to weight each neighbour's recommendations.