Recommending venues or points-of-interest (POIs) is a hot topic in recent years, especially for tourism applications and mobile users. We propose and evaluate several suggestion methods, taking an effectiveness, feasibility, efficiency, and privacy perspective. The task is addressed by two content-based methods (a Weighted kNN classifier and a Rated Rocchio personalized query), Collaborative Filtering methods, as well as several (rank-based or rating-based) methods of merging results of different systems. Effectiveness is evaluated on two standard benchmark datasets, provided and used by TREC’s Contextual Suggestion Tracks in 2015 and 2016. First, we enrich these datasets with more information on venues, collected from web services like Foursquare and Yelp; we make this extra data available for future experimentation. Then, we find that the content-based methods provide state-of-the-art effectiveness, the collaborative filtering variants mostly suffer from data sparsity problems in the current datasets, and the merging methods further improve results by mainly promoting the first relevant suggestion. Concerning mobile feasibility, efficiency, and user privacy, the content-based methods, especially Rated Rocchio, are the best. Collaborative filtering has the worst efficiency and privacy leaks. Our findings can be very useful for developing effective and efficient operational systems, respecting user privacy. Last, our experiments indicate that better benchmark datasets would be welcome, and the use of additional evaluation measures—more sensitive in recall—is recommended.

We present Pythia, a privacy-enhanced non-invasive contextual suggestion system for tourists, with important architectural innovations. The system offers high quality personalized recommendations, non-invasive operation and protection of user privacy. A key feature of Pythia is the exploitation of the vast amounts of personal data generated by smartphones to automatically build user profiles, and make contextual suggestions to tourists. More precisely, the system utilizes (sensitive) personal data, such as location traces, browsing history and web searches (query logs), to build a POI-based user profile. This profile is then used by a contextual suggestion engine for making POI recommendations to the user based on her current location. Strong privacy guarantees are achieved by placing both mechanisms at the user-side. As a proof of concept, we present a Pythia prototype which implements the aforementioned mechanisms as mobile applications for Android, as well as, web applications.

We give a fresh look into score normalization for merging result-lists, isolating the problem from other components. We focus on three of the simplest, practical, and widely-used linear methods which do not require any training data, i.e. MinMax, Sum, and Z-Score. We provide theoretical arguments on why and when the methods work, and evaluate them experimentally. We find that MinMax is the most robust under many circumstances, and that Sum is - in contrast to previous literature - the worst. Based on the insights gained, we propose another three simple methods which work as good or better than the baselines.

The Bag-Of-Visual-Words (BOVW) paradigm is fast becoming a popular image representation for Content-Based Image Retrieval (CBIR), mainly because of its better retrieval effectiveness over global feature representations on collections with images being near-duplicate to queries. In this experimental study we demonstrate that this advantage of BOVW is diminished when visual diversity is enhanced by using a secondary modality, such as text, to pre-filter images.

Traditional tools for information retrieval (IR) evaluation, such as TREC's trec_eval, have outdated command-line interfaces with many unused features, or 'switches', accumulated over the years. They are usually seen as cumbersome applications by new IR researchers, steepening the learning curve. We introduce a platform-independent application for IR evaluation with a graphical easy-to-use interface: the TREC_Files Evaluator. The application supports most of the standard measures used for evaluation in TREC, CLEF, and elsewhere, such as MAP, P10, P20, and bpref, as well as the Averaged Normalized Modified Retrieval Rank (ANMRR) proposed by MPEG for image retrieval evaluation. Additional features include a batch mode and statistical significance testing of the results against a pre-selected baseline.

Compact composite descriptors (CCDs) are global image features, capturing more than one types of information at the same time in a very compact representation. Their qual-ity has so far been evaluated in retrieval from several homo-geneous databases containing images of only the type that each CCD is intended for, and has been found better than other descriptors in the literature such as the MPEG-7 de-scriptors. In this study, we consider heterogeneous databases and investigate query-time fusion techniques for CCDs. The results show that fusion is beneficial, even with simple score normalization and combination methods due to the com-patibility of the score distributions produced by the CCDs considered.

Score normalization is indispensable in distributed retrieval and fu-sion or meta-search where merging of result-lists is required. Dis-tributional approaches to score normalization with reference to rel-evance, such as binary mixture models like the normal-exponential, suffer from lack of universality and troublesome parameter estima-tion especially under sparse relevance. We develop a new approach which tackles both problems by using aggregate score distributions without reference to relevance, and is suitable for uncooperative engines. The method is based on the assumption that scores pro-duced by engines consist of a signal and a noise component which can both be approximated by submitting well-defined sets of arti-ficial queries to each engine. We evaluate in a standard distributed retrieval testbed and show that the signal-to-noise approach yields better results than other distributional methods. As a significant by-product, we investigate query-length distributions.

Ranked retrieval has a particular disadvantage in comparison with traditional Boolean retrieval: there is no clear cut-off point where to stop consulting results. This is a serious problem in some setups. We investigate and further develop methods to select the rank cut-off value which optimizes a given effectiveness measure. Assuming no other input than a system's output for a query--document scores and their distribution--the task is essentially a score-distributional threshold optimization problem. The recent trend in modeling score distributions is to use a normal-exponential mixture: normal for relevant, and exponential for non-relevant document scores. We discuss the two main theoretical problems with the current model, support incompatibility and non-convexity, and develop new models that address them. The main contributions of the paper are two truncated normal-exponential models, varying in the way the out-truncated score ranges are handled. We conduct a range of experiments using the TREC 2007 and 2008 Legal Track data, and show that the truncated models lead to significantly better results.

We analyse query length, and fit power-law and Poisson distributions to four different query sets. We provide a practical model for query length, based on the truncation of a Poisson distribution for short queries and a power-law distribution for longer queries, that better fits real query length distributions than earlier proposals.

Much of the information stored on the web contains geographical context, but current search engines treat such context in the same way as all other content. In this paper we describe the design, implementation and evaluation of a spatially aware search engine which is capable of handling queries in the form of the triplet of ⟨theme⟩⟨spatial relationship⟩⟨location⟩. The process of identifying geographic references in documents and assigning appropriate footprints to documents, to be stored together with document terms in an appropriate indexing structure allowing real‐time search, is described. Methods allowing users to query and explore results which have been relevance‐ranked in terms of both thematic and spatial relevance have been implanted and a usability study indicates that users are happy with the range of spatial relationships available and intuitively understand how to use such a search engine. Normalised precision for 38 queries, containing four types of spatial relationships, is significantly higher (p<0.001) for searches exploiting spatial information than pure text search.

The thresholding of document scores has proved critical for the effectiveness of classification tasks. We review the most important approaches to thresholding, and introduce thescore-distributional (S-D) threshold optimizationmethod. The method is based on score distributions and is capable of optimizing any effectiveness measure defined in terms of the traditional contingency table.