Alan Eckhardt

Similarity of users' (content-based) preference models for Collaborative filtering in few ratings scenario

Collaborative filtering is an efficient way to find best objects to recommend. This technique is particularly useful when there is a lot of users that rated a lot of objects. In this paper, we propose a method that improve the Collaborative filtering in situations, where the number of ratings or users is small. The proposed approach is experimentally evaluated on real datasets with very convincing results.

A system recommending top-k objects for multiple users preferences

We discuss models of user preferences in Web environment. We construct a model for user preference querying over a number of data sources and ordering of answers by a combination of particular attribute rankings. We generalize Fagins algorithm in two directions - we develop some new heuristics for top-k search in the model without random access and propose a method of ordering lists of objects by user fuzzy function. To enable different user preferences our system does not require objects to be sorted -instead we use a B+-tree on each of the attribute domains. This leads to a more realistic model of Web services. We implement our methods and heuristics for search of top-k answers into Tokaf middleware framework prototype. We describe experiments with Tokaf and compare different performance measures with some other methods.

PHASES: A User Profile Learning Approach for Web Search

Web search heuristics based on Fagins threshold algorithm assume we have the user profile in the form of particular attribute ordering and a fuzzy aggregation function representing the user combining function. Having these, there are sufficient algorithms for searching top-k answers. Finding particular attribute ordering and aggregation for a user still remains a problem. In this short paper our main contribution is a proof of concept of a new iterative process of acquisition of user preferences and attribute ordering .

Learning User Preferences for 2CP-Regression for a Recommender System

In this paper we deal with a task to learn a general user model from user ratings of a small set of objects. This general model is used to recommend top-k objects to the user. We consider several (also some new) alternatives of learning local preferences and several alternatives of aggregation (with or without 2CP-regression). The main contributions are evaluation of experiments on our prototype tool PrefWork with respect to several satisfaction measures and the proposal of method Peak for normalisation of numerical attributes. Our main objective is to keep the number of sample data which the user has to rate reasonable small.

Learning Different User Profile Annotated Rules for Fuzzy Preference Top-k Querying

Uncertainty querying of large data can be solved by providing top-k answers according to a user fuzzy ranking/scoring function. Usually different users have different fuzzy scoring function --- a user preference model. Main goal of this paper is to assign a user a preference model automatically. To achieve this we decompose users fuzzy ranking function to ordering of particular attributes and to a combination function. To solve the problem of automatic assignment of user model we design two algorithms, one for learning user preference on particular attribute and second for learning the combination function. Methods were integrated into a Fagin-like top-k querying system with some new heuristics and tested.

UPComp - A PHP Component for Recommendation Based on User Behaviour

In this paper, we investigate the possibilities of interpreting user behaviour in order to learn his/her preferences. UP Comp, a PHP component enabling use of user preferences for recommendation, is described. UP Comp is a standalone component that can be integrated into any PHP web with only basic knowledge of PHP, HTML and SQL. The methods of user behaviour interpretation are evaluated on a real web shop with tourist trips using UP Comp.

Combining Various Methods of Automated User Decision and Preferences Modelling

In this paper we present a proposal of a system that combines various methods of user modelling. This system may find its application in e-commerce, recommender systems, etc. The main focus of this paper is on automatic methods that require only a small amount of data from user. The different ways of integration of user models are studied. A proof-of-concept implementation is compared to standard methods in an initial experiment with artificial user data...

On Fuzzy vs. Metric Similarity Search in Complex Databases

The task of similarity search is widely used in various areas of computing, including multimedia databases, data mining, bioinformatics, social networks, etc. For a long time, the database-oriented applications of similarity search employed the definition of similarity restricted to metric distances. Due to the metric postulates (reflexivity, non-negativity, symmetry and triangle inequality), a metric similarity allows to build a metric index above the database which can be subsequently used for efficient (fast) similarity search. On the other hand, the metric postulates limit the domain experts (providers of the similarity measure) in similarity modeling. In this paper we propose an alternative non-metric method of indexing for efficient similarity search. The requirement on metric is replaced by the requirement on fuzzy similarity satisfying the transitivity property with a tuneable fuzzy conjunctor. We also show a duality between the fuzzy approach and the metric one.

Integrating user and group preferences for top-k search from distributed web resources

We discuss models of user and group preferences in social networks and the Semantic web. We construct a model for user and group preference querying over RDF data as well as for ordering of answers by aggregation of particular attribute ranking. We have implemented our methods and heuristics into the Tokaf middleware framework prototype. We describe also experiments with Tokaf.In this paper we investigate the problem of automatically identifying the genre of TV programmes. The approach here proposed is based on two foundations: Gaussian mixture models (GMMs) and artificial neural networks (ANNs). Firstly, we use Gaussian mixtures to model the probability distributions of low-level audiovisual features. Secondly, we use the parameters of each mixture model as new feature vectors. Finally, we train a multilayer perceptron (MLP), using GMM parameters as input data, to identify seven television programme genres. We evaluated the effectiveness of the proposed approach testing our system on a large set of data, summing up to more than 100 hours of broadcasted programmes.

Uncertainty Issues and Algorithms in Automating Process Connecting Web and User

We focus on replacing human processing web resources by automated processing. On an experimental system we identify uncertainty issues making this process difficult for automated processing and try to minimize human intervention. In particular we focus on uncertainty issues in a Web content mining system and a user preference mining system. We conclude with possible future development heading to an extension of OWL with uncertainty features.