Europeana is the leading digital library of Europe’s cultural heritage, providing access to over 60 million items in more than 40 languages. Its search infrastructure relies on Solr and BM25 over the items’ metadata, thus depending heavily on keyword matching and resource-intensive treatments such as translation and multilingual metadata enrichment. This paper explores the application of Neural Information Retrieval (NIR) approaches in Europeana, focusing on multilinguality. We created a dataset for comparative evaluation, and show that while NIR demonstrates strong potential for multilingual search, challenges remain regarding its performance, particularly for entity-centric queries. This work also highlights the need for more reliable evaluation data.

The purpose of the Strategic Workshop on Information Retrieval in Lorne (SWIRL)¹ is to explore the long-range issues of the information retrieval (IR) field, to recognise challenges that are on – or even over – the horizon, to build consensus on key challenges, and to disseminate the resulting information to the research community. The intent is that this description of open problems will help to inspire researchers and graduate students to address the questions and will provide funding agencies with data to focus and coordinate support for IR research.

Rank-Biased Overlap (RBO) is a similarity measure for indefinite rankings: it is top-weighted, and can be computed when only a prefix of the rankings is known or when they have only some items in common. It is widely used for instance to analyze differences between search engines by comparing the rankings of documents they retrieve for the same queries. In these situations, though, it is very frequent to find tied documents that have the same score. Unfortunately, the treatment of ties in RBO remains superficial and incomplete, in the sense that it is not clear how to calculate it from the ranking prefixes only. In addition, the existing way of dealing with ties is very different from the one traditionally followed in the field of Statistics, most notably found in rank correlation coefficients such as Kendall's and Spearman's. In this paper we propose a generalized formulation for RBO to handle ties, thanks to which we complete the original definitions by showing how to perform prefix evaluation. We also use it to fully develop two variants that align with the ones found in the Statistics literature: one when there is a reference ranking to compare to, and one when there is not. Overall, these three variants provide researchers with flexibility when comparing rankings with RBO, by clearly determining what ties mean, and how they should be treated. Finally, using both synthetic and TREC data, we demonstrate the use of these new tie-aware RBO measures. We show that the scores may differ substantially from the original tie-unaware RBO measure, where ties had to be broken at random or by arbitrary criteria such as by document ID. Overall, these results evidence the need for a proper account of ties in rank similarity measures such as RBO.

Rank-Biased Overlap (RBO) is a popular measure of the similarity between two rankings. A key characteristic of RBO is that it can be computed even when the rankings are not fully seen and only a prefix is known, but this introduces uncertainty in the computation. In such cases, one would normally compute the point estimate RBO_EXT, as well as bounds representing the best and worst cases; their difference is thus a residual quantifying the amount of uncertainty. Another source of uncertainty is the presence of tied items, because their actual relative order is unknown. Current approaches to this issue similarly provide a point estimate by considering the average RBO score over all the permutations of the ties, such as RBO^a. However, there is currently no approach to quantify and bound the uncertainty due to ties, just as there is for the uncertainty due to unseen items. In this paper we fill this gap and provide algorithmic solutions to the problem of finding the arrangements of tied items that yield the lowest and highest possible RBO scores, naturally leading to total bounds and residuals. We also show that the current RBO^a estimate only equals the average RBO over permutations when the rankings have the same length, so we also generalize it to rankings of different lengths. In summary, this work provides a full account for the uncertainty in RBO, allowing practitioners to make more sensible decisions on the grounds of rank similarity. The main realization is that residuals can actually be much larger once we account for both sources of uncertainty. To illustrate this, we present empirical results using both synthetic and TREC data, demonstrating that a realistic picture for the residual of RBO can only be provided by considering both sources of uncertainty.

Mainstream bias, where some users receive poor recommendations because their preferences are uncommon or simply because they are less active, is an important aspect to consider regarding fairness in recommender systems. Existing methods to mitigate mainstream bias do not explicitly model the importance of these non-mainstream users or, when they do, it is in a way that is not necessarily compatible with the data and recommendation model at hand. In contrast, we use the recommendation utility as a more generic and implicit proxy to quantify mainstreamness, and propose a simple user-weighting approach to incorporate it into the training process while taking the cost of potential recommendation errors into account. We provide extensive experimental results showing that quantifying mainstreamness via utility is better able at identifying non-mainstream users, and that they are indeed better served when training the model in a cost-sensitive way. This is achieved with negligible or no loss in overall recommendation accuracy, meaning that the models learn a better balance across users. In addition, we show that research of this kind, which evaluates recommendation quality at the individual user level, may not be reliable if not using enough interactions when assessing model performance.

This report documents the program and the outcomes of Dagstuhl Seminar 23031 "Frontiers of Information Access Experimentation for Research and Education", which brought together 38 participants from 12 countries. The seminar addressed technology-enhanced information access (information retrieval, recommender systems, natural language processing) and specifically focused on developing more responsible experimental practices leading to more valid results, both for research as well as for scientific education.The seminar featured a series of long and short talks delivered by participants, who helped in setting a common ground and in letting emerge topics of interest to be explored as the main output of the seminar. This led to the definition of five groups which investigated challenges, opportunities, and next steps in the following areas: reality check, i.e. conducting real-world studies, human-machine-collaborative relevance judgment frameworks, overcoming methodological challenges in information retrieval and recommender systems through awareness and education, results-blind reviewing, and guidance for authors.Date: 15--20 January 2023.Website: https://www.dagstuhl.de/23031.

Direct optimization of IR metrics has often been adopted as an approach to devise and develop ranking-based recommender systems. Most methods following this approach (e.g. TFMAP, CLiMF, Top-N-Rank) aim at optimizing the same metric being used for evaluation, under the assumption that this will lead to the best performance. A number of studies of this practice bring this assumption, however, into question. In this paper, we dig deeper into this issue in order to learn more about the effects of the choice of the metric to optimize on the performance of a ranking-based recommender system. We present an extensive experimental study conducted on different datasets in both pairwise and listwise learning-to-rank (LTR) scenarios, to compare the relative merit of four popular IR metrics, namely RR, AP, nDCG and RBP, when used for optimization and assessment of recommender systems in various combinations. For the first three, we follow the practice of loss function formulation available in literature. For the fourth one, we propose novel loss functions inspired by RBP for both the pairwise and listwise scenario. Our results confirm that the best performance is indeed not necessarily achieved when optimizing the same metric being used for evaluation. In fact, we find that RBP-inspired losses perform at least as well as other metrics in a consistent way, and offer clear benefits in several cases. Interesting to see is that RBP-inspired losses, while improving the recommendation performance for all uses, may lead to an individual performance gain that is correlated with the activity level of a user in interacting with items. The more active the users, the more they benefit. Overall, our results challenge the assumption behind the current research practice of optimizing and evaluating the same metric, and point to RBP-based optimization instead as a promising alternative when learning to rank in the recommendation context.

Statistical significance tests are the main tool that IR practitioners use to determine the reliability of their experimental evaluation results. The question of which test behaves best with IR evaluation data has been around for decades, and has seen all kinds of results and recommendations. Definitive answer to this question has recently been attempted via stochastic simulation of IR evaluation data, allowing researchers to compute actual Type I error rates because they can control the null hypothesis. One such research line simulates metric scores for a fixed set of systems on random topics, and concluded that the t-test behaves the best. Another such line simulates retrieval runs by random systems on a fixed set of topics, and concluded that the Wilcoxon test behaves the best. Interestingly, two recent surveys of the IR literature have shown that the community has a clear preference precisely for these two tests, so further investigation is critical to understand why the above simulation studies reach opposite conclusions. It has been recently postulated that a reason for the disagreement is the distributions of metric scores used by one of these simulation methods. In this paper we investigate this issue and extend the argument to another key aspect of the simulation, namely the dependence between systems. Following a principled approach, we analyze the robustness of statistical tests to different factors, thus identifying under what conditions they behave well or not with respect to the Type I error rate. Our results suggest that differences between the Wilcoxon and t-test may be explained by the skewness of score differences.

In a collaborative-filtering recommendation scenario, biases in the data will likely propagate in the learned recommendations. In this paper we focus on the so-called mainstream bias: the tendency of a recommender system to provide better recommendations to users who have a mainstream taste, as opposed to non-mainstream users. We propose NAECF, a conceptually simple but effective idea to address this bias. The idea consists of adding an autoencoder (AE) layer when learning user and item representations with text-based Convolutional Neural Networks. The AEs, one for the users and one for the items, serve as adversaries to the process of minimizing the rating prediction error when learning how to recommend. They enforce that the specific unique properties of all users and items are sufficiently well incorporated and preserved in the learned representations. These representations, extracted as the bottlenecks of the corresponding AEs, are expected to be less biased towards mainstream users, and to provide more balanced recommendation utility across all users. Our experimental results confirm these expectations, significantly improving the recommendations for nonmainstream users while maintaining the recommendation quality for mainstream users. Our results emphasize the importance of deploying extensive content-based features, such as online reviews, in order to better represent users and items to maximize the debiasing effect.

With the advent of deep learning, global tempo estimation accuracy has reached a new peak, which presents a great opportunity to evaluate our evaluation practices. In this article, we discuss presumed and actual applications, the pros and cons of commonly used metrics, and the suitability of popular datasets. To guide future research, we present results of a survey among domain experts that investigates today’s applications, their requirements, and the usefulness of currently employed metrics. To aid future evaluations, we present a public repository containing evaluation code as well as estimates by many different systems and different ground truths for popular datasets.

Recently proposed methods allow the generation of simulated scores representing the values of an effectiveness metric, but they do not investigate the generation of the actual lists of retrieved documents. In this paper we address this limitation: we present an approach that exploits an evolutionary algorithm and, given a metric score, creates a simulated relevance profile (i.e., a ranked list of relevance values) that produces that score. We show how the simulated relevance profiles are realistic under various analyses.

This paper introduces the AcousticBrainz Genre Dataset, a large-scale collection of hierarchical multi-label genre annotations from different metadata sources. It allows researchers to explore how the same music pieces are annotated differently by different communities following their own genre taxonomies, and how this could be addressed by genre recognition systems. Genre labels for the dataset are sourced from both expert annotations and crowds, permitting comparisons between strict hierarchies and folksonomies. Music features are available via the AcousticBrainz database. To guide research, we suggest a concrete research task and provide a baseline as well as an evaluation method. This task may serve as an example of the development and validation of automatic annotation algorithms on complementary datasets with different taxonomies and coverage. With this dataset, we hope to contribute to developments in content-based music genre recognition as well as cross-disciplinary studies on genre metadata analysis

Statistical significance testing is widely accepted as a means to assess how well a difference in effectiveness reflects an actual difference between systems, as opposed to random noise because of the selection of topics. According to recent surveys on SIGIR, CIKM, ECIR and TOIS papers, the t-test is the most popular choice among IR researchers. However, previous work has suggested computer intensive tests like the bootstrap or the permutation test, based mainly on theoretical arguments. On empirical grounds, others have suggested non-parametric alternatives such as the Wilcoxon test. Indeed, the question of which tests we should use has accompanied IR and related fields for decades now. Previous theoretical studies on this matter were limited in that we know that test assumptions are not met in IR experiments, and empirical studies were limited in that we do not have the necessary control over the null hypotheses to compute actual Type I and Type II error rates under realistic conditions. Therefore, not only is it unclear which test to use, but also how much trust we should put in them. In contrast to past studies, in this paper we employ a recent simulation methodology from TREC data to go around these limitations. Our study comprises over 500 million p-values computed for a range of tests, systems, effectiveness measures, topic set sizes and effect sizes, and for both the 2-tail and 1-tail cases. Having such a large supply of IR evaluation data with full knowledge of the null hypotheses, we are finally in a position to evaluate how well statistical significance tests really behave with IR data, and make sound recommendations for practitioners.

This paper introduces the AcousticBrainz Genre Dataset, a large-scale collection of hierarchical multi-label genre annotations from different metadata sources. It allows researchers to explore how the same music pieces are annotated differently by different communities following their own genre taxonomies, and how this could be addressed by genre recognition systems. Genre labels for the dataset are sourced from both expert annotations and crowds, permitting comparisons between strict hierarchies and folksonomies. Music features are available via the Acoustic- Brainz database. To guide research, we suggest a concrete research task and provide a baseline as well as an evaluation method. This task may serve as an example of the development and validation of automatic annotation algorithms on complementary datasets with different taxonomies and coverage. With this dataset, we hope to contribute to developments in content-based music genre recognition as well as cross-disciplinary studies on genre metadata analysis.

Metaphors are commonly used in interface design within Human-Computer Interaction (HCI). Interface metaphors provide users with a way to interact with the computer that resembles a known activity, giving instantaneous knowledge or intuition about how the interaction works. A widely used one in Digital Musical Instruments (DMIs) is the conductor-orchestra metaphor, where the orchestra is considered as an instrument controlled by the movements of the conductor. We propose a DMI based on the conductor metaphor that allows to control tempo and dynamics and adapts its mapping specifically for each user by observing spontaneous conducting movements (i.e., movements performed on top of fixed music without any instructions). We refer to this as mapping by observation given that, even though the systemis trained specifically for each
user, this training is not done explicitly and consciously by the user. More specifically, the system adapts its mapping based on the tendency of the user to anticipate or fall behind the beat and observing the Motion Capture descriptors that best correlate to loudness during spontaneous conducting. We evaluate the proposed system in an experiment with twenty four (24) participants where we compare it with a baseline that does not perform this user-specific adaptation. The comparison is done in a context where the user does not receive instructions and, instead, is allowed to discover by playing. We evaluate objective and subjective measures from tasks where participants have to make
the orchestra play at different loudness levels or in synchrony with a metronome. Results of the experiment prove that the usability of the system that automatically learns its mapping from spontaneous movements is better both in terms of providing a more intuitive control over loudness and a more precise control over beat timing. Interestingly, the results also show a strong correlation betweenmeasures taken fromthe data used for training and the improvement introduced by the adapting system. This indicates that it is possible to estimate in advance how useful the observation of spontaneous movements is to build user-specific adaptations. This opens interesting directions for creating more
intuitive and expressive DMIs, particularly in public installations.

Deep neural networks have frequently been used to directly learn representations useful for a given task from raw input data. In terms of overall performance metrics, machine learning solutions employing deep representations frequently have been reported to greatly outperform those using hand-crafted feature representations. At the same time, they may pick up on aspects that are predominant in the data, yet not actually meaningful or interpretable. In this paper, we therefore propose a systematic way to test the trustworthiness of deep music representations, considering musical semantics. The underlying assumption is that in case a deep representation is to be trusted, distance consistency between known related points should be maintained both in the input audio space and corresponding latent deep space. We generate known related points through semantically meaningful transformations, both considering imperceptible and graver transformations. Then, we examine within- and between-space distance consistencies, both considering audio space and latent embedded space, the latter either being a result of a conventional feature extractor or a deep encoder. We illustrate how our method, as a complement to task-specific performance, provides interpretable insight into what a network may have captured from training data signals.

Inspired by the success of deploying deep learning in the fields of Computer Vision and Natural Language Processing, this learning paradigm has also found its way into the field of Music Information Retrieval. In order to benefit from deep learning in an effective, but also efficient manner, deep transfer learning has become a common approach. In this approach, it is possible to reuse the output of a pre-trained neural network as the basis for a new learning task. The underlying hypothesis is that if the initial and new learning tasks show commonalities and are applied to the same type of input data (e.g., music audio), the generated deep representation of the data is also informative for the new task. Since, however, most of the networks used to generate deep representations are trained using a single initial learning source, their representation is unlikely to be informative for all possible future tasks. In this paper, we present the results of our investigation of what are the most important factors to generate deep representations for the data and learning tasks in the music domain. We conducted this investigation via an extensive empirical study that involves multiple learning sources, as well as multiple deep learning architectures with varying levels of information sharing between sources, in order to learn music representations. We then validate these representations considering multiple target datasets for evaluation. The results of our experiments yield several insights into how to approach the design of methods for learning widely deployable deep data representations in the music domain.

In test collection based evaluation of IR systems, score standardization has been proposed to compare systems across collections and minimize the effect of outlier runs on specific topics. The underlying idea is to account for the difficulty of topics, so that systems are scored relative to it. Webber et al. first proposed standardization through a non-linear transformation with the standard normal distribution, and recently Sakai proposed a simple linear transformation. In this paper, we show that both approaches are actually special cases of a simple standardization which assumes specific distributions for the per-topic scores. From this viewpoint, we argue that a transformation based on the empirical distribution is the most appropriate choice for this kind of standardization. Through a series of experiments on TREC data, we show the benefits of our proposal in terms of score stability and statistical test behavior.

Part of Information Retrieval evaluation research is limited by the fact that we do not know the distributions of system effectiveness over the populations of topics and, by extension, their true mean scores. The workaround usually consists in resampling topics from an existing collection and approximating the statistics of interest with the observations made between random subsamples, as if one represented the population and the other a random sample. However, this methodology is clearly limited by the availability of data, the impossibility to control the properties of these data, and the fact that we do not really measure what we intend to. To overcome these limitations, we propose a method based on vine copulas for stochastic simulation of evaluation results where the true system distributions are known upfront. In the basic use case, it takes the scores from an existing collection to build a semi-parametric model representing the set of systems and the population of topics, which can then be used to make realistic simulations of the scores by the same systems but on random new topics. Our ability to simulate this kind of data not only removes the current limitations, but also offers new opportunities for research. As an example, we show the benefits of this approach in two sample applications replicating typical experiments found in the literature. We provide a full R package to simulate new data following the proposed method, which can also be used to fully reproduce the results in this paper.

Named Entity Recognition is a basic task in Information Extraction that aims at identifying entities of interest within full text documents. The patterns used to recognize entities can be rule based, as in the popular JAPE system. However, hand-crafting effective patterns is often difficult, and yet there is little research devoted to methods capable of learning human-readable patterns, possibly with arbitrary sets of features. In this paper, we present a semi-Automatic method to generate both regular expressions and a subset of the JAPE language. It does not need a corpus annotated beforehand. Instead, it employs active learning and combines clustering with an algorithm that finds alignments between symbols present in the entities discovered during the learning process. The method currently supports a fixed set of character features and an arbitrary set of token features, but it can incorporate other kinds of features as well. Through several experiments with an English corpus, we show the ability of the method to generate effective patterns at a low annotation cost, and how it can successfully help in the annotation of brand new corpora.