William Rudman, Michal Golovanevsky, Dana Arad, Yonatan Belinkov, Carsten Eickhoff, Ritambhara Singh, Kyle Mahowald. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

As models grow more complex and societal demands for transparency increase with emerging regulations, explainability has become an increasingly important research area. However, despite its recognized relevance, progress in explainability research in information retrieval (IR) has been slower than in related fields. This full day workshop aims to advance research in explainable IR by providing a more in-depth platform to reflect on recent developments and facilitate discussions across both new and persistent challenges. Building upon the first edition of the workshop, which was a great success in bringing together multiple perspectives on explainability in IR, this second edition will focus on synthesizing a common agenda for the research community. Through a set of interactive activities, the workshop will bring together a diverse group of researchers to build a shared understanding of key tasks and challenges, and to help shape future directions for explainable IR research. The workshop will have as concrete outcomes a roadmap document and a special issue proposal for a journal issue on explainability in IR.

Learned sparse retrieval models such as SPLADE combine the effectiveness of neural architectures with the efficiency of inverted indices. As these models assign weights to terms from a fixed vocabulary, interpretability is often touted as a major benefit of these models. However, the emergence of wacky weights, i.e., expansion terms that appear semantically unrelated to the input, limits interpretability. While prior research has anecdotally observed this phenomenon, there is a lack of systematic understanding regarding their origins, prevalence, and contribution to retrieval effectiveness. In this paper, we reproduce SPLADE-v2 to systematically investigate wacky weights across the SPLADE family of models. We present a comprehensive dissection of wacky weights, providing a formal definition of wackiness based on the lexical utility of expansion terms. Furthermore, we introduce a novel measure to compare the prevalence of these tokens across models with varying vocabularies and sparsity levels. Beyond reproducing the original SPLADE-v2, we train it with various loss functions, datasets, and backbone transformers to isolate the factors contributing to wackiness. Our results show that larger vocabularies are associated with a higher prevalence of wacky tokens, while stricter sparsity regularizers are associated with lower prevalence. Finally, we find that wacky weights are used primarily for in-domain effectiveness rather than out-of-domain generalization.

BACKGROUND: Key challenges in leveraging unstructured clinician notes for predictive models include identifying and timing patient outcomes. To address these challenges we applied large language models (LLMs) to identify and temporally localize patient outcomes in clinician notes, and evaluated whether this contextual data enhances predictive modeling for conditions like sepsis. METHODOLOGY: We applied the Medical Concept Annotation Tool (MedCAT) and two LLMs, Meta-Llama-3.1-8B and BioMistral 7B, to clinician notes in the Medical Information Mart for Intensive Care version III to identify and time International Classification of Diseases-based patient outcomes. A physician manually validated the accuracy of the models' ability to identify and time outcomes including sepsis. Finally, downstream time series predictive modeling was used to assess the impact of unstructured clinician notes on sepsis prediction. RESULTS: Meta-Llama-3.1-8B demonstrated the best balance of coverage and precision, reliably identifying outcomes while minimizing false positives. Our manual physician validation evaluated model performance, which varied between outcomes. When comparing derived sepsis timestamps to clinical sepsis onset within a 5-h window, Meta-Llama-3.1-8B showed the smallest median time difference. Finally, incorporating temporally localized outcomes into downstream sepsis prediction improved model performance, area under the receiver operating characteristic (AUC) of 0.84 (95% CI: 0.83-0.86) from 0.77 (95% CI: 0.76-0.79). CONCLUSION: We demonstrate that LLM-driven identification and timing of patient outcomes from unstructured clinician notes is feasible. Contextual outcome identification unlocks the potential of unstructured clinician notes for predictive modeling. We provide TIMED-MIMIC, an automatically generated dataset encompassing 1697 temporally localized patient outcomes, as a publicly available resource.

Retrieval-Augmented Generation (RAG) systems have recently shown remarkable advancements by integrating retrieval mechanisms into language models, enhancing their ability to produce more accurate and contextually relevant responses. However, the influence of various components and configurations within RAG systems remains underexplored. A comprehensive understanding of these elements is essential for tailoring RAG systems to complex retrieval tasks and ensuring optimal performance across diverse applications. In this paper, we develop several advanced RAG system designs that incorporate query expansion, various novel retrieval strategies, and a novel Contrastive In-Context Learning RAG. Our study systematically investigates key factors, including language model size, prompt design, document chunk size, knowledge base size, retrieval stride, query expansion techniques, Contrastive In-Context Learning knowledge bases, multilingual knowledge bases, and Focus Mode retrieving relevant context at sentence-level. Through extensive experimentation, we provide a detailed analysis of how these factors influence response quality. Our findings offer actionable insights for developing RAG systems, striking a balance between contextual richness and retrieval-generation efficiency, thereby paving the way for more adaptable and high-performing RAG frameworks in diverse real-world scenarios. Our code and implementation details are publicly available.

Multimodal Large Language Models (MLLMs) perform well on tasks such as visual question answering, but it remains unclear whether their reasoning relies more on memorized world knowledge or on the visual information present in the input image.To investigate this, we introduce Visual CounterFact, a new dataset of visually-realistic counterfactuals that put world knowledge priors (e.g, red strawberry) into direct conflict with visual input (e.g, blue strawberry).Using Visual CounterFact, we show that model predictions initially reflect memorized priors, but shift toward visual evidence in mid-to-late layers.This dynamic reveals a competition between the two modalities, with visual input ultimately overriding priors during evaluation.To control this behavior, we propose Pixels Versus Priors (PvP) steering vectors, a mechanism for controlling model outputs toward either world knowledge or visual input through activation-level interventions.On average, PvP successfully shifts 99.3% of color and 80.8% of size predictions from priors to counterfactuals.Together, these findings offer new tools for interpreting and controlling factual behavior in multimodal models.* †

Mechanistic interpretation has greatly contributed to a more detailed understanding of generative language models, enabling significant progress in identifying structures that implement key behaviors through interactions between internal components.In contrast, interpretability in information retrieval (IR) remains relatively coarse-grained, and much is still unknown as to how IR models determine whether a document is relevant to a query.In this work, we address this gap by mechanistically analyzing how one commonly used model, a cross-encoder, estimates relevance.We find that the model extracts traditional relevance signals, such as term frequency and inverse document frequency, in early-to-middle layers.These concepts are then combined in later layers, similar to the well-known probabilistic ranking function, BM25.Overall, our analysis offers a more nuanced understanding of how IR models compute relevance.Isolating these components lays the groundwork for future interventions that could enhance transparency, mitigate safety risks, and improve scalability.

Multilingual large language models (LLMs) often exhibit factual inconsistencies across languages, usually with better performance in factual recall tasks in high-resource languages than in other languages. The causes of these failures, however, remain poorly understood. Using mechanistic analysis techniques, we uncover the underlying pipeline that LLMs employ, which involves using the English-centric factual recall mechanism to process multilingual queries and then translating English answers back into the target language. We identify two primary sources of error: insufficient engagement of the reliable English-centric mechanism for factual recall, and incorrect translation from English back into the target language for the final answer. To address these vulnerabilities, we introduce two vector interventions, both independent of languages and datasets, to redirect the model toward better internal paths for higher factual consistency. Our interventions combined increase the recall accuracy by over 35 percent for the lowest-performing language. Our findings demonstrate how mechanistic insights can be used to unlock latent multilingual capabilities in LLMs.

Large language models (LLMs) exhibit sophisticated behavior, notably solving arithmetic with only a few in-context examples (ICEs).Yet the computations that connect those examples to the answer remain opaque.We probe four open-weight LLMs, Pythia-12B, Llama-3.1-8B,MPT-7B, and OPT-6.7B, on basic arithmetic to illustrate how they process ICEs.Our study integrates activation patching, information-flow analysis, automatic circuit discovery, and the logit-lens perspective into a unified pipeline.Within this framework we isolate partial-sum representations in three-operand tasks, investigate their influence on final logits, and derive linear function vectors that characterize tasks and align with ICEinduced activations.Controlled ablations show that strict pattern consistency in the formatting of ICEs guides the models more strongly than the symbols chosen or even the factual correctness of the examples.By unifying four complementary interpretability tools, this work delivers one of the most comprehensive interpretability studies of LLM arithmetic to date, and the first on three-operand tasks.Our code is publicly available 1 .

We employ new tools from mechanistic interpretability to ask whether the internal structure of large language models (LLMs) shows correspondence to the linguistic structures which underlie the languages on which they are trained. In particular, we ask (1) when two languages employ the same morphosyntactic processes, do LLMs handle them using shared internal circuitry? and (2) when two languages require different morphosyntactic processes, do LLMs handle them using different internal circuitry? In a focused case study on English and Chinese multilingual and monolingual models, we analyze the internal circuitry involved in two tasks. We find evidence that models employ the same circuit to handle the same syntactic process independently of the language in which it occurs, and that this is the case even for monolingual models trained completely independently. Moreover, we show that multilingual models employ language-specific components (attention heads and feed-forward networks) when needed to handle linguistic processes (e.g., morphological marking) that only exist in some languages. Together, our results are revealing about how LLMs trade off between exploiting common structures and preserving linguistic differences when tasked with modeling multiple languages simultaneously, opening the door for future work in this direction.

Biomedical knowledge graphs (KGs) encode rich, structured information critical for drug discovery tasks, but extracting meaningful insights from large-scale KGs remains challenging due to their complex structure. Existing biomedical subgraph retrieval methods are tailored for graph neural networks (GNNs), limiting compatibility with other paradigms, including large language models (LLMs). We introduce K-Paths, a model-agnostic retrieval framework that extracts structured, diverse, and biologically meaningful multi-hop paths from dense biomedical KGs. These paths enable prediction of unobserved drug-drug and drug-disease interactions, including those involving entities not seen during training, thus supporting inductive reasoning. K-Paths is training-free and employs a diversity-aware adaptation of Yen's algorithm to extract the K shortest loopless paths between entities in a query, prioritizing biologically relevant and relationally diverse connections. These paths serve as concise, interpretable reasoning chains that can be directly integrated with LLMs or GNNs to improve generalization, accuracy, and enable explainable inference. Experiments on benchmark datasets show that K-Paths improves zero-shot reasoning across state-of-the-art LLMs. For instance, Tx-Gemma 27B improves by 19.8 and 4.0 F1 points on interaction severity prediction and drug repurposing tasks, respectively. Llama 70B achieves gains of 8.5 and 6.2 points on the same tasks. K-Paths also boosts the training efficiency of EmerGNN, a state-of-the-art GNN, by reducing the KG size by 90% while maintaining predictive performance. Beyond efficiency, K-Paths bridges the gap between KGs and LLMs, enabling scalable and explainable LLM-augmented scientific discovery. We release our code and the retrieved paths as a benchmark for inductive reasoning.

Michal Golovanevsky, William Rudman, Vedant Palit, Carsten Eickhoff, Ritambhara Singh. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

The meteoric rise of Artificial Intelligence (AI), with its rapidly expanding market capitalization, presents both transformative opportunities and critical challenges. Chief among these is the urgent need for a new, unified paradigm for trustworthy evaluation, as current benchmarks increasingly reveal critical vulnerabilities. Issues like data contamination and selective reporting by model developers fuel hype, while inadequate data quality control can lead to biased evaluations that, even if unintentionally, may favor specific approaches. As a flood of participants enters the AI space, this "Wild West" of assessment makes distinguishing genuine progress from exaggerated claims exceptionally difficult. Such ambiguity blurs scientific signals and erodes public confidence, much as unchecked claims would destabilize financial markets reliant on credible oversight from agencies like Moody's.In high-stakes human examinations (e.g., SAT, GRE), substantial effort is devoted to ensuring fairness and credibility; why settle for less in evaluating AI, especially given its profound societal impact? This position paper argues that a laissez-faire approach is untenable. For true and sustainable AI advancement, we call for a paradigm shift to a unified, live, and quality-controlled benchmarking framework—robust by construction rather than reliant on courtesy or goodwill. Accordingly, we dissect the systemic flaws undermining today’s evaluation ecosystem and distill the essential requirements for next-generation assessments. To concretize this position, we introduce the idea of PeerBench, a community-governed, proctored evaluation blueprint that seeks to improve security and credibility through sealed execution, item banking with rolling renewal, and delayed transparency. PeerBench is presented as a complementary, certificate-grade layer alongside open benchmarks, not a replacement. We discuss trade-offs and limits and call for further research on mechanism design, governance, and reliability guarantees. Our goal is to lay the groundwork for evaluations that restore integrity and deliver genuinely trustworthy measures of AI progress.

As models grow more complex and societal demands for transparency increase with emerging regulations, explainability has become an even more important research area. However, despite its recognized relevance, explainability research in IR has seen slower progress than in related fields. This full day workshop aims to advance research in explainable information retrieval by providing a more in-depth platform to reflect on recent developments and facilitate discussions to address new and persistent challenges. Our goal is to bring together a diverse group of researchers to build a shared understanding of key tasks and challenges that will lay the foundation for the future of explainable IR research.

Mechanistic interpretability research, which aims to uncover the internal processes of machine learning models, has gained significant attention. One state-of-the-art technique, activation patching, has been applied to analyzing neural ranker behavior in relation to information retrieval (IR) axioms. To date, however, this remains a rapidly evolving topic in IR, with no established methodology for measuring results or constructing datasets to ensure pronounced, robust, and consistent patching effects. In this study, based on experimental results, we provide recommendations on measuring patching effects and designing diagnostic datasets for investigating term frequency. We identify the rareness and informativeness of injected terms as a key factor influencing the magnitude of patching effects. Additionally, we find that low score differences between baseline and perturbed documents introduce significant noise, which can be mitigated by filtering or applying penalty scores to the metric. More generally, we provide practical recommendations for the reliable application of activation patching in IR, advancing future interpretability research of neural ranking models. Our code is available at https://github.com/polgrisha/best-practices-ir-patching.

Cross-lingual summarization (CLS) has attracted increasing interest in recent years due to the availability of large-scale web-mined datasets and the advancements of multilingual language models. However, given the rareness of naturally occurring CLS resources, the majority of datasets are forced to rely on translation which can contain overly literal artifacts. This restricts our ability to observe naturally occurring CLS pairs that capture organic diction, including instances of code-switching. This alteration between languages in mid-message is a common phenomenon in multilingual settings yet has been largely overlooked in cross-lingual contexts due to data scarcity. To address this gap, we introduce CroCoSum, a dataset of cross-lingual code-switched summarization of technology news. It consists of over 24,000 English source articles and 18,000 human-written Chinese news summaries, with more than 92% of the summaries containing code-switched phrases. For reference, we evaluate the performance of existing approaches including pipeline, end-to-end, and zero-shot methods. We show that leveraging existing CLS resources as a pretraining step does not improve performance on CroCoSum, indicating the limited generalizability of current datasets. Finally, we discuss the challenges of evaluating cross-lingual summarizers on code-switched generation through qualitative error analyses.

Jessica Zosa Forde, Ruochen Zhang, Lintang Sutawika, Alham Fikri Aji, Samuel Cahyawijaya, Genta Indra Winata, Minghao Wu, Carsten Eickhoff, Stella Biderman, Ellie Pavlick. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024.

Recent work in mechanistic interpretability has shown that behaviors in language models can be successfully reverse-engineered through circuit analysis. A common criticism, however, is that each circuit is task-specific, and thus such analysis cannot contribute to understanding the models at a higher level. In this work, we present evidence that insights (both low-level findings about specific heads and higher-level findings about general algorithms) can indeed generalize across tasks. Specifically, we study the circuit discovered in (Wang, 2022) for the Indirect Object Identification (IOI) task and 1.) show that it reproduces on a larger GPT2 model, and 2.) that it is mostly reused to solve a seemingly different task: Colored Objects (Ippolito & Callison-Burch, 2023). We provide evidence that the process underlying both tasks is functionally very similar, and contains about a 78% overlap in in-circuit attention heads. We further present a proof-of-concept intervention experiment, in which we adjust four attention heads in middle layers in order to ‘repair’ the Colored Objects circuit and make it behave like the IOI circuit. In doing so, we boost accuracy from 49.6% to 93.7% on the Colored Objects task and explain most sources of error. The intervention affects downstream attention heads in specific ways predicted by their interactions in the IOI circuit, indicating that this subcircuit behavior is invariant to the different task inputs. Overall, our results provide evidence that it may yet be possible to explain large language models' behavior in terms of a relatively small number of interpretable task-general algorithmic building blocks and computational components.

Given the success of Large Language Models (LLMs), there has been considerable interest in studying the properties of model activations. The literature overwhelmingly agrees that LLM representations are dominated by a few ``outlier dimensions'' with exceedingly high variance and magnitude. Several studies in Natural Language Processing (NLP) have sought to mitigate the impact of such outlier dimensions and force LLMs to be isotropic (i.e., have uniform variance across all dimensions in embedding space). Isotropy is thought to be a desirable property for LLMs that improves model performance and more closely aligns textual representations with human intuition. However, many claims regarding isotropy in NLP have been based on the average cosine similarity of embeddings, which has recently been shown to be a flawed measure of isotropy. In this paper, we propose I-STAR: IsoScore$^{\star}$-based STable Anisotropic Regularization, a novel regularization method that can be used to increase or decrease levels of isotropy in embedding space during training. I-STAR uses IsoScore$^{\star}$, the first accurate measure of isotropy that is both differentiable and stable on mini-batch computations. In contrast to several previous works, we find that \textit{decreasing} isotropy in contextualized embeddings improves performance on the majority of tasks and models considered in this paper.

Jack Merullo, Carsten Eickhoff, Ellie Pavlick. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024.

Although it is known that transformer language models (LMs) pass features from early layers to later layers, it is not well understood how this information is represented and routed by the model. We analyze a mechanism used in two LMs to selectively inhibit items in a context in one task, and find that it underlies a commonly used abstraction across many context-retrieval behaviors. Specifically, we find that models write into low-rank subspaces of the residual stream to represent features which are then read out by later layers, forming low-rank *communication channels* (Elhage et al., 2021) between layers. A particular 3D subspace in model activations in GPT-2 can be traversed to positionally index items in lists, and we show that this mechanism can explain an otherwise arbitrary-seeming sensitivity of the model to the order of items in the prompt. That is, the model has trouble copying the correct information from context when many items ``crowd" this limited space. By decomposing attention heads with the Singular Value Decomposition (SVD), we find that previously described interactions between heads separated by one or more layers can be predicted via analysis of their weight matrices alone. We show that it is possible to manipulate the internal model representations as well as edit model weights based on the mechanism we discover in order to significantly improve performance on our synthetic Laundry List task, which requires recall from a list, often improving task accuracy by over 20\%. Our analysis reveals a surprisingly intricate interpretable structure learned from language model pretraining, and helps us understand why sophisticated LMs sometimes fail in simple domains, facilitating future analysis of more complex behaviors.

As information retrieval (IR) systems, such as search engines and conversational agents, become ubiquitous in various domains, the need for transparent and explainable systems grows to ensure accountability, fairness, and unbiased results. Despite recent advances in explainable AI and IR techniques, there is no consensus on the definition of explainability. Existing approaches often treat it as a singular notion, disregarding the multidimensional definition postulated in the literature. In this paper, we use psychometrics and crowdsourcing to identify human-centered factors of explainability in Web search systems and introduce SSE (Search System Explainability), an evaluation metric for explainable IR (XIR) search systems. In a crowdsourced user study, we demonstrate SSE's ability to distinguish between explainable and non-explainable systems, showing that systems with higher scores indeed indicate greater interpretability. We hope that aside from these concrete contributions to XIR, this line of work will serve as a blueprint for similar explainability evaluation efforts in other domains of machine learning and natural language processing.

Neural models have demonstrated remarkable performance across diverse ranking tasks. However, the processes and internal mechanisms along which they determine relevance are still largely unknown. Existing approaches for analyzing neural ranker behavior with respect to IR properties rely either on assessing overall model behavior or employing probing methods that may offer an incomplete understanding of causal mechanisms. To provide a more granular understanding of internal model decision-making processes, we propose the use of causal interventions to reverse engineer neural rankers, and demonstrate how mechanistic interpretability methods can be used to isolate components satisfying term-frequency axioms within a ranking model. We identify a group of attention heads that detect duplicate tokens in earlier layers of the model, then communicate with downstream heads to compute overall document relevance. More generally, we propose that this style of mechanistic analysis opens up avenues for reverse engineering the processes neural retrieval models use to compute relevance. This work aims to initiate granular interpretability efforts that will not only benefit retrieval model development and training, but ultimately ensure safer deployment of these models.

Research on knowledge graph embedding (KGE) has emerged as an active field in which most existing KGE approaches mainly focus on static structural data and ignore the influence of temporal variation involved in time-aware triples. In order to deal with this issue, several temporal knowledge graph embedding (TKGE) approaches have been proposed to integrate temporal and structural information. However, these methods only employ a uniformly random sampling to construct negative facts. As a consequence, the corrupted samples are often too simplistic for training an effective model. In this paper, we propose a new temporal knowledge graph embedding framework by introducing adversarial learning to further refine the performance of traditional TKGE models. In our framework, a generator is utilized to construct high-quality plausible quadruples and a discriminator learns to obtain the embeddings of entities and relations based on both positive and negative samples. Meanwhile, we also apply a Gumbel-Softmax relaxation and the Wasserstein distance to prevent vanishing gradient problems on discrete data; an inherent flaw in traditional generative adversarial networks . Through comprehensive experimentation on temporal datasets, the results indicate that our proposed framework can attain significant improvements based on benchmark models and also demonstrate the effectiveness and applicability of our framework.

Document-level text simplification is a specific type of simplification which involves simplifying documents consisting of several sentences by rewriting them into fewer or more sentences.In this paper, we propose a new two-stage framework SIMSUM for automated document-level text simplification.Our model is designed with explicit summarization and simplification models and guides the generation using the main keywords of a source text.In order to evaluate our new model, we use two existing benchmark datasets for simplification, namely D-Wikipedia and Wiki-Doc.We compare our model's performance with state of the art and show that SIMSUM achieves top results on the D-Wikipedia dataset SARI (+1.20),D-SARI (+1.64), and FKGL (-0.35) scores, improving over the best baseline models.In order to evaluate the quality of the generated text, we analyze the outputs from different models qualitatively and demonstrate the merit of our new model.Our code and datasets are available 1 .

Societal biases that are contained in retrieved documents have received increased interest. Such biases, which are often prevalent in the training data and learned by the model, can cause societal harms, by misrepresenting certain groups, and by enforcing stereotypes. Mitigating such biases demands algorithms that balance the trade-off between maximized utility for the user with fairness objectives, which incentivize unbiased rankings. Prior work on bias mitigation often assumes that ranking scores, which correspond to the utility that a document holds for a user, can be accurately determined. In reality, there is always a degree of uncertainty in the estimate of expected document utility. This uncertainty can be approximated by viewing ranking models through a Bayesian perspective, where the standard deterministic score becomes a distribution.

Deepak Kumar, Oleg Lesota, George Zerveas, Daniel Cohen, Carsten Eickhoff, Markus Schedl, Navid Rekabsaz. Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics. 2023.

Representations from large language models (LLMs) are known to be dominated by a small subset of dimensions with exceedingly high variance.Previous works have argued that although ablating these outlier dimensions in LLM representations hurts downstream performance, outlier dimensions are detrimental to the representational quality of embeddings.In this study, we investigate how fine-tuning impacts outlier dimensions and show that 1) outlier dimensions that occur in pre-training persist in fine-tuned models and 2) a single outlier dimension can complete downstream tasks with a minimal error rate.Our results suggest that outlier dimensions can encode crucial taskspecific knowledge and that the value of a representation in a single outlier dimension drives downstream model decisions.1

Sparse annotation poses persistent challenges to training dense retrieval models, for example by distorting the training signal when unlabeled relevant documents are used spuriously as negatives in contrastive learning.To alleviate this problem, we introduce evidence-based label smoothing, a novel, computationally efficient method that prevents penalizing the model for assigning high relevance to false negatives.To compute the target relevance distribution over candidate documents within the ranking context of a given query, those candidates most similar to the ground truth are assigned a nonzero relevance probability based on the degree of their similarity to the ground-truth document(s).To estimate relevance we leverage an improved similarity metric based on reciprocal nearest neighbors, which can also be used independently to rerank candidates in postprocessing.Through extensive experiments on two large-scale ad hoc text retrieval datasets, we demonstrate that reciprocal nearest neighbors can improve the ranking effectiveness of dense retrieval models, both when used for label smoothing, as well as for reranking.This indicates that by considering relationships between documents and queries beyond simple geometric distance we can effectively enhance the ranking context. 1 1 Our code and other resources are available at: https://github.com/gzerveas/CODER 2 E.g., on average 1.

Contrastive learning has been the dominant approach to training dense retrieval models.In this work, we investigate the impact of ranking context -an often overlooked aspect of learning dense retrieval models.In particular, we examine the effect of its constituent parts: jointly scoring a large number of negatives per query, using retrieved (query-specific) instead of random negatives, and a fully list-wise loss.To incorporate these factors into training, we introduce Contextual Document Embedding Reranking (CODER), a highly efficient retrieval framework.When reranking, it incurs only a negligible computational overhead on top of a firststage method at run time (∼ 5 ms delay per query), allowing it to be easily combined with any state-of-the-art dual encoder method.Models trained through CODER can also be used as stand-alone retrievers.Evaluating CODER in a large set of experiments on the MS MARCO and TripClick collections, we show that the contextual reranking of precomputed document embeddings leads to a significant improvement in retrieval performance.This improvement becomes even more pronounced when more relevance information per query is available, shown in the TripClick collection, where we establish new state-of-the-art results by a large margin.

Epilepsy is one of the most common neurological disorders, typically observed via seizure episodes. Epileptic seizures are commonly monitored through electroencephalogram (EEG) recordings due to their routine and low expense collection. The stochastic nature of EEG makes seizure identification via manual inspections performed by highly-trained experts a tedious endeavor, motivating the use of automated identification. The literature on automated identification focuses mostly on supervised learning methods requiring expert labels of EEG segments that contain seizures, which are difficult to obtain. Motivated by these observations, we pose seizure identification as an unsupervised anomaly detection problem. To this end, we employ the first unsupervised transformer-based model for seizure identification on raw EEG. We train an autoencoder involving a transformer encoder via an unsupervised loss function, incorporating a novel masking strategy uniquely designed for multivariate time-series data such as EEG. Training employs EEG recordings that do not contain any seizures, while seizures are identified with respect to reconstruction errors at inference time. We evaluate our method on three publicly available benchmark EEG datasets for distinguishing seizure vs. non-seizure windows. Our method leads to significantly better seizure identification performance than supervised learning counterparts, by up to 16% recall, 9% accuracy, and 9% Area under the Receiver Operating Characteristics Curve (AUC), establishing particular benefits on highly imbalanced data. Through accurate seizure identification, our method could facilitate widely accessible and early detection of epilepsy development, without needing expensive label collection or manual feature extraction.

Given a query, neural retrieval models predict point estimates of relevance for each document; however, a significant drawback of relying solely on point estimates is that they contain no indication of the model's confidence in its predictions. Despite this lack of information, downstream methods such as reranking, cutoff prediction, and none-of-the-above classification are still able to learn effective functions to accomplish their respective tasks. Unfortunately, these downstream methods can suffer poor performance when the initial ranking model loses confidence in its score predictions. This becomes increasingly important in high-stakes settings, such as medical searches that can influence health decision making.

Societal biases can influence Information Retrieval system results, and conversely, search results can potentially reinforce existing societal biases. Recent research has therefore focused on developing methods for quantifying and mitigating bias in search results and applied them to contemporary retrieval systems that leverage transformer-based language models. In the present work, we expand this direction of research by considering bias mitigation within a framework for contextual document embedding reranking. In this framework, the transformer-based query encoder is optimized for relevance ranking through a list-wise objective, by jointly scoring for the same query a large set of candidate document embeddings in the context of one another, instead of in isolation. At the same time, we impose a regularization loss which penalizes highly scoring documents that deviate from neutrality with respect to a protected attribute (e.g., gender). Our approach for bias mitigation is end-to-end differentiable and efficient. Compared to the existing alternatives for deep neural retrieval architectures, which are based on adversarial training, we demonstrate that it can attain much stronger bias mitigation/fairness. At the same time, for the same amount of bias mitigation, it offers significantly better relevance performance (utility). Crucially, our method allows for a more finely controllable and predictable intensity of bias mitigation, which is essential for practical deployment in production systems.

OBJECTIVE: Alzheimer's disease (AD) is the most common neurodegenerative disorder with one of the most complex pathogeneses, making effective and clinically actionable decision support difficult. The objective of this study was to develop a novel multimodal deep learning framework to aid medical professionals in AD diagnosis. MATERIALS AND METHODS: We present a Multimodal Alzheimer's Disease Diagnosis framework (MADDi) to accurately detect the presence of AD and mild cognitive impairment (MCI) from imaging, genetic, and clinical data. MADDi is novel in that we use cross-modal attention, which captures interactions between modalities-a method not previously explored in this domain. We perform multi-class classification, a challenging task considering the strong similarities between MCI and AD. We compare with previous state-of-the-art models, evaluate the importance of attention, and examine the contribution of each modality to the model's performance. RESULTS: MADDi classifies MCI, AD, and controls with 96.88% accuracy on a held-out test set. When examining the contribution of different attention schemes, we found that the combination of cross-modal attention with self-attention performed the best, and no attention layers in the model performed the worst, with a 7.9% difference in F1-scores. DISCUSSION: Our experiments underlined the importance of structured clinical data to help machine learning models contextualize and interpret the remaining modalities. Extensive ablation studies showed that any multimodal mixture of input features without access to structured clinical information suffered marked performance losses. CONCLUSION: This study demonstrates the merit of combining multiple input modalities via cross-modal attention to deliver highly accurate AD diagnostic decision support.

Neural sequence-to-sequence models are the state-of-the-art approach used in abstractive summarization of textual documents, useful for producing condensed versions of source text narratives without being restricted to using only words from the original text. Despite the advances in abstractive summarization, custom generation of summaries (e.g., towards a user’s preference) remains unexplored. In this article, we present CATS, an abstractive neural summarization model that summarizes content in a sequence-to-sequence fashion while also introducing a new mechanism to control the underlying latent topic distribution of the produced summaries. We empirically illustrate the efficacy of our model in producing customized summaries and present findings that facilitate the design of such systems. We use the well-known CNN/DailyMail dataset to evaluate our model. Furthermore, we present a transfer-learning method and demonstrate the effectiveness of our approach in a low resource setting, i.e., abstractive summarization of meetings minutes, where combining the main available meetings’ transcripts datasets, AMI and International Computer Science Institute(ICSI) , results in merely a few hundred training documents.

We present a novel framework for multivariate time series representation learning based on the transformer encoder architecture. The framework includes an unsupervised pre-training scheme, which can offer substantial performance benefits over fully supervised learning on downstream tasks, both with but even without leveraging additional unlabeled data, i.e., by reusing the existing data samples. Evaluating our framework on several public multivariate time series datasets from various domains and with diverse characteristics, we demonstrate that it performs significantly better than the best currently available methods for regression and classification, even for datasets which consist of only a few hundred training samples. Given the pronounced interest in unsupervised learning for nearly all domains in the sciences and in industry, these findings represent an important landmark, presenting the first unsupervised method shown to push the limits of state-of-the-art performance for multivariate time series regression and classification.

In the pursuit of natural language understanding, there has been a long standing interest in tracking state changes throughout narratives.Impressive progress has been made in modeling the state of transaction-centric dialogues and procedural texts.However, this problem has been less intensively studied in the realm of general discourse where ground truth descriptions of states may be loosely defined and state changes are less densely distributed over utterances.This paper proposes to turn to simplified, fully observable systems that show some of these properties: Sports events.We curated 2,263 soccer matches including timestamped natural language commentary accompanied by discrete events such as a team scoring goals, switching players or being penalized with cards.We propose a new task formulation where, given paragraphs of commentary of a game at different timestamps, the system is asked to recognize the occurrence of in-game events.This domain allows for rich descriptions of state while avoiding the complexities of many other real-world settings.As an initial point of performance measurement, we include two baseline methods from the perspectives of sentence classification with temporal dependence and current state-of-the-art generative model, respectively, and demonstrate that even sophisticated existing methods struggle on the state tracking task when the definition of state broadens or non-event chatter becomes prevalent.

In any ranking system, the retrieval model outputs a single score for a document based on its belief on how relevant it is to a given search query. While retrieval models have continued to improve with the introduction of increasingly complex architectures, few works have investigated a retrieval model's belief in the score beyond the scope of a single value. We argue that capturing the model's uncertainty with respect to its own scoring of a document is a critical aspect of retrieval that allows for greater use of current models across new document distributions, collections, or even improving effectiveness for down-stream tasks. In this paper, we address this problem via an efficient Bayesian framework for retrieval models which captures the model's belief in the relevance score through a stochastic process while adding only negligible computational overhead. We evaluate this belief via a ranking based calibration metric showing that our approximate Bayesian framework significantly improves a retrieval model's ranking effectiveness through a risk aware reranking as well as its confidence calibration. Lastly, we demonstrate that this additional uncertainty information is actionable and reliable on down-stream tasks represented via cutoff prediction.

Click logs are valuable resources for a variety of information retrieval (IR) tasks. This includes query understanding/analysis, as well as learning effective IR models particularly when the models require large amounts of training data. We release a large-scale domain-specific dataset of click logs, obtained from user interactions of the Trip Database health web search engine. Our click log dataset comprises approximately 5.2 million user interactions collected between 2013 and 2020. We use this dataset to create a standard IR evaluation benchmark - TripClick - with around 700,000 unique free-text queries and 1.3 million pairs of query-document relevance signals, whose relevance is estimated by two click-through models. As such, the collection is one of the few datasets offering the necessary data richness and scale to train neural IR models with a large amount of parameters, and notably the first in the health domain. Using TripClick, we conduct experiments to evaluate a variety of IR models, showing the benefits of exploiting this data to train neural architectures. In particular, the evaluation results show that the best performing neural IR model significantly improves the performance by a large margin relative to classical IR models, especially for more frequent queries.

An interaction between pharmacological agents can trigger unexpected adverse events. Capturing richer and more comprehensive information about drug-drug interactions (DDIs) is one of the key tasks in public health and drug development. Recently, several knowledge graph (KG) embedding approaches have received increasing attention in the DDI domain due to their capability of projecting drugs and interactions into a low-dimensional feature space for predicting links and classifying triplets. However, existing methods only apply a uniformly random mode to construct negative samples. As a consequence, these samples are often too simplistic to train an effective model. In this paper, we propose a new KG embedding framework by introducing adversarial autoencoders (AAEs) based on Wasserstein distances and Gumbel-Softmax relaxation for DDI tasks. In our framework, the autoencoder is employed to generate high-quality negative samples and the hidden vector of the autoencoder is regarded as a plausible drug candidate. Afterwards, the discriminator learns the embeddings of drugs and interactions based on both positive and negative triplets. Meanwhile, in order to solve vanishing gradient problems on the discrete representation-an inherent flaw in traditional generative models-we utilize the Gumbel-Softmax relaxation and the Wasserstein distance to train the embedding model steadily. We empirically evaluate our method on two tasks: link prediction and DDI classification. The experimental results show that our framework can attain significant improvements and noticeably outperform competitive baselines. Supplementary information: Supplementary data and code are available at https://github.com/dyf0631/AAE_FOR_KG.

In politics, neologisms are frequently invented for partisan objectives.For example, "undocumented workers" and "illegal aliens" refer to the same group of people (i.e., they have the same denotation), but they carry clearly different connotations.Examples like these have traditionally posed a challenge to referencebased semantic theories and led to increasing acceptance of alternative theories (e.g., Two-Factor Semantics) among philosophers and cognitive scientists.In NLP, however, popular pretrained models encode both denotation and connotation as one entangled representation.In this study, we propose an adversarial neural network that decomposes a pretrained representation as independent denotation and connotation representations.For intrinsic interpretability, we show that words with the same denotation but different connotations (e.g., "immigrants" vs. "aliens", "estate tax" vs. "death tax") move closer to each other in denotation space while moving further apart in connotation space.For extrinsic application, we train an information retrieval system with our disentangled representations and show that the denotation vectors improve the viewpoint diversity of document rankings.

Search engines often provide only limited explanation on why results are ranked in a particular order. This lack of transparency prevents users from understanding results and can potentially give rise to biased or unfair systems. Opaque search engines may also hurt user trust in the presented ranking. This paper presents an investigation of system quality when different degrees of explanation are provided on search engine result pages. Our user study demonstrates that the inclusion of even simplistic explanations leads to better transparency, increased user trust and better search efficiency.

We present HSDM, a full-day workshop on Health Search and Data Mining co-located with WSDM 2020's Health Day. This event builds on recent biomedical workshops in the NLP and ML communities but puts a clear emphasis on search and data mining (and their intersection) that is lacking in other venues. The program will include two keynote addresses by key opinion leaders in the clinical, search, and data mining domains. The technical program consists of 6 original research presentations. Finally, we will close with a panel discussion with keynote speakers, PC members, and the audience.

Low-frequency terms are a recurring challenge for information retrieval models, especially neural IR frameworks struggle with adequately capturing infrequently observed words. While these terms are often removed from neural models - mainly as a concession to efficiency demands - they traditionally play an important role in the performance of IR models. In this paper, we analyze the effects of low-frequency terms on the performance and robustness of neural IR models. We conduct controlled experiments on three recent neural IR models, trained on a large-scale passage retrieval collection. We evaluate the neural IR models with various vocabulary sizes for their respective word embeddings, considering different levels of constraints on the available GPU memory. We observe that despite the significant benefits of using larger vocabularies, the performance gap between the vocabularies can be, to a great extent, mitigated by extensive tuning of a related parameter: the number of documents to re-rank. We further investigate the use of subword-token embedding models, and in particular FastText, for neural IR models. Our experiments show that using FastText brings slight improvements to the overall performance of the neural IR models in comparison to models trained on the full vocabulary, while the improvement becomes much more pronounced for queries containing low-frequency terms.

Motivation: Technological advancements in high-throughput DNA sequencing have led to an exponential growth of sequencing data being produced and stored as a byproduct of biomedical research. Despite its public availability, a majority of this data remains hard to query for the research community due to a lack of efficient data representation and indexing solutions. One of the available techniques to represent read data is a condensed form as an assembly graph. Such a representation contains all sequence information but does not store contextual information and metadata. Results: We present two new approaches for a compressed representation of a graph coloring: a lossless compression scheme based on a novel application of wavelet tries as well as a highly accurate lossy compression based on a set of Bloom filters. Both strategies retain a coloring even when adding to the underlying graph topology. We present construction and merge procedures for both methods and evaluate their performance on a wide range of different datasets. By dropping the requirement of a fully lossless compression and using the topological information of the underlying graph, we can reduce memory requirements by up to three orders of magnitude. Representing individual colors as independently stored modules, our approaches can be efficiently parallelized and provide strategies for dynamic use. These properties allow for an easy upscaling to the problem sizes common to the biomedical domain. Availability and implementation: We provide prototype implementations in C++, summaries of our experiments as well as links to all datasets publicly at https://github.com/ratschlab/graph_annotation. Supplementary information: Supplementary data are available at Bioinformatics online.

Crowdsourcing has become a popular paradigm in data curation, annotation and evaluation for many artificial intelligence and information retrieval applications. Considerable efforts have gone into devising effective quality control mechanisms that identify or discourage cheat submissions in an attempt to improve the quality of noisy crowd judgments. Besides purposeful cheating, there is another source of noise that is often alluded to but insufficiently studied: Cognitive biases.

Many social network applications depend on robust representations of spatio-temporal data. In this work, we present an embedding model based on feed-forward neural networks which transforms social media check-ins into dense feature vectors encoding geographic, temporal, and functional aspects for modeling places, neighborhoods, and users. We employ the embedding model in a variety of applications including location recommendation , urban functional zone study , and crime prediction . For location recommendation , we propose a Spatio-Temporal Embedding Similarity algorithm (STES) based on the embedding model. In a range of experiments on real life data collected from Foursquare, we demonstrate our model’s effectiveness at characterizing places and people and its applicability in aforementioned problem domains. Finally, we select eight major cities around the globe and verify the robustness and generality of our model by porting pre-trained models from one city to another, thereby alleviating the need for costly local training.

Topic models such as Latent Dirichlet Allocation (LDA) have been widely used in information retrieval for tasks ranging from smoothing and feedback methods to tools for exploratory search and discovery. However, classical methods for inferring topic models do not scale up to the massive size of today's publicly available Web-scale data sets. The state-of-the-art approaches rely on custom strategies, implementations and hardware to facilitate their asynchronous, communication-intensive workloads. We present APS-LDA, which integrates state-of-the-art topic modeling with cluster computing frameworks such as Spark using a novel asynchronous parameter server. Advantages of this integration include convenient usage of existing data processing pipelines and eliminating the need for disk writes as data can be kept in memory from start to finish. Our goal is not to outperform highly customized implementations, but to propose a general high-performance topic modeling framework that can easily be used in today's data processing pipelines. We compare APS-LDA to the existing Spark LDA implementations and show that our system can, on a 480-core cluster, process up to 135× more data and 10× more topics without sacricing model quality.

Crowdsourcing has long established itself as a viable alternative to corpus annotation by domain experts for tasks such as document relevance assessment. The crowdsourcing process traditionally relies on high degrees of label redundancy in order to mitigate the detrimental effects of individually noisy worker submissions. Such redundancy comes at the cost of increased label volume, and, subsequently, monetary requirements. In practice, especially as the size of datasets increases, this is undesirable. In this paper, we focus on an alternate method that exploits document information instead, to infer relevance labels for unjudged documents. We present an active learning scheme for document selection that aims at maximising the overall relevance label prediction accuracy, for a given budget of available relevance judgements by exploiting system-wide estimates of label variance and mutual information.

The proliferation of Internet-enabled devices and services has led to a shifting balance between digital and analogue aspects of our everyday lives. In the face of this development there is a growing demand for the study of privacy hazards, the potential for unique user deanonymization and information leakage between the various social media profiles many of us maintain. To enable the structured study of such adversarial effects, this paper presents a dedicated dataset of cross-platform social network personas (i.e., the same person has accounts on multiple platforms). The corpus comprises 850 users who generate predominantly English content. Each user object contains the online footprint of the same person in three distinct social networks: Twitter, Instagram and Foursquare. In total, it encompasses over 2.5M tweets, 340k check-ins and 42k Instagram posts. We describe the collection methodology, characteristics of the dataset, and how to obtain it. Finally, we discuss a common use case, cross-platform user identification.

Many fundamental problems in natural language processing rely on determining what entities appear in a given text. Commonly referenced as entity linking, this step is a fundamental component of many NLP tasks such as text understanding, automatic summarization, semantic search or machine translation. Name ambiguity, word polysemy, context dependencies and a heavy-tailed distribution of entities contribute to the complexity of this problem. We here propose a probabilistic approach that makes use of an effective graphical model to perform collective entity disambiguation. Input mentions (i.e., linkable token spans) are disambiguated jointly across an entire document by combining a document-level prior of entity co-occurrences with local information captured from mentions and their surrounding context. The model is based on simple sufficient statistics extracted from data, thus relying on few parameters to be learned.

The use of crowdsourcing for document relevance assessment has been found to be a viable alternative to corpus annotation by highly trained experts. The question of quality control is a recurring challenge that is often addressed by aggregating multiple individual assessments of the same topic-document pair from independent workers. In the past, such aggregation schemes have been weighted or filtered by estimates of worker reliability based on a multitude of behavioral features. In this paper, we propose an alternative approach by relying on document information. Inspired by the clustering hypothesis of information retrieval, we assume textually similar documents to show similar degrees of relevance towards a given topic. Following up on this intuition, we propagate crowd-generated relevance judgments to similar documents, effectively smoothing the distribution of relevance labels across the similarity space.

Information about a user's domain knowledge and interest can be important signals for many information retrieval tasks such as query suggestion or result ranking. State-of-the-art user models rely on coarse-grained representations of the user's previous knowledge about a topic or domain. In this paper, we study query refinement using eye-tracking in order to gain precise and detailed insight into which terms the user was exposed to in a search session and which ones they showed a particular interest in. We measure fixations on the term level, allowing for a detailed model of user attention. To allow for a wide-spread exploitation of our findings, we generalize from the restrictive eye-gaze tracking to using more accessible signals: mouse cursor traces. Based on the public API of a popular search engine, we demonstrate how query suggestion candidates can be ranked according to traces of user attention and interest, resulting in significantly better performance than achieved by an attention-oblivious industry solution. Our experiments suggest that modelling term-level user attention can be achieved with great reliability and holds significant potential for supporting a range of traditional IR tasks.

Many generative language and relevance models assume conditional independence between the likelihood of observing individual terms. This assumption is obviously naive, but also hard to replace or relax. There are only very few term pairs that actually show significant conditional dependencies while the vast majority of co-located terms has no implications on the document's topical nature or relevance towards a given topic. It is exactly this situation that we capture in a formal framework: A limited number of meaningful dependencies in a system of largely independent observations. Making use of the formal copula framework, we describe the strength of causal dependency in terms of a number of established term co-occurrence metrics. Our experiments based on the well known ClueWeb'12 corpus and TREC 2013 topics indicate significant performance gains in terms of retrieval performance when we formally account for the dependency structure underlying pieces of natural language text.

Modern relevance models consider a wide range of criteria in order to identify those documents that are expected to satisfy the user's information need. With growing dimensionality of the underlying relevance spaces the need for sophisticated score combination and estimation schemes arises. In this paper, we investigate the use of copulas, a model family from the domain of robust statistics, for the formal estimation of the probability of relevance in high-dimensional spaces. Our experiments are based on the MSLR-WEB10K and WEB30K datasets, two annotated, publicly available samples of hundreds of thousands of real Web search impressions, and suggest that copulas can significantly outperform linear combination models for high-dimensional problems. Our models achieved a performance on par with that of state-of-the-art machine learning approaches.

The Internet is the largest source of information in the world. Search engines help people navigate the huge space of available data in order to acquire new skills and knowledge. In this paper, we present an in-depth analysis of sessions in which people explicitly search for new knowledge on the Web based on the log files of a popular search engine. We investigate within-session and cross-session developments of expertise, focusing on how the language and search behavior of a user on a topic evolves over time. In this way, we identify those sessions and page visits that appear to significantly boost the learning process. Our experiments demonstrate a strong connection between clicks and several metrics related to expertise. Based on models of the user and their specific context, we present a method capable of automatically predicting, with good accuracy, which clicks will lead to enhanced learning. Our findings provide insight into how search engines might better help users learn as they search.

In many domains of information retrieval, system estimates of document relevance are based on multidimensional quality criteria that have to be accommodated in a unidimensional result ranking. Current solutions to this challenge are often inconsistent with the formal probabilistic framework in which constituent scores were estimated, or use sophisticated learning methods that make it difficult for humans to understand the origin of the final ranking. To address these issues, we introduce the use of copulas, a powerful statistical framework for modeling complex multi-dimensional dependencies, to information retrieval tasks. We provide a formal background to copulas and demonstrate their effectiveness on standard IR tasks such as combining multidimensional relevance estimates and fusion of results from multiple search engines. We introduce copula-based versions of standard relevance estimators and fusion methods and show that these lead to significant performance improvements on several tasks, as evaluated on large-scale standard corpora, compared to their non-copula counterparts. We also investigate criteria for understanding the likely effect of using copula models in a given retrieval scenario.

Most research in Web search personalization models users as static or slowly evolving entities with a given set of preferences defined by their past behavior. However, recent publications as well as empirical evidence suggest that for a significant number of search sessions, users diverge from their regular search profiles in order to satisfy atypical, limited-duration information needs. In this work, we conduct a large-scale inspection of real-life search sessions to further understand this scenario. Subsequently, we design an automatic means of detecting and supporting such atypical sessions. We demonstrate significant improvements over state-of-the-art Web search personalization techniques by accounting for the typicality of search sessions. The proposed method is evaluated based on Web-scale search session data spanning several months of user activity.

Crowdsourcing successfully strives to become a widely used means of collecting large-scale scientific corpora. Many research fields, including Information Retrieval, rely on this novel way of data acquisition. However, it seems to be undermined by a significant share of workers that are primarily interested in producing quick generic answers rather than correct ones in order to optimise their time-efficiency and, in turn, earn more money. Recently, we have seen numerous sophisticated schemes of identifying such workers. Those, however, often require additional resources or introduce artificial limitations to the task. In this work, we take a different approach by investigating means of a priori making crowdsourced tasks more resistant against cheaters.

The continued development and maturation of advanced HTML features such as Cascading style sheets (CSS), Javascript, and AJAX, as well as their widespread adoption by browsers, has enabled web pages to flourish with sophistication and interactivity. Unfortunately, this presents challenges to the web search community, as a web page's representation in the browser (i.e., what users see) can diverge dramatically from its raw HTML content (i.e., what search engines index and retrieve). For example, interactive pages may contain content in regions that are not visible before a user action, such as focusing a tab, but which are nonetheless still contained within the raw HTML. We study this divergence by comparing raw HTML to its fully rendered form across a number of metrics spanning presentation, geometry, and content, using a large, representative sample of popular web pages. We find that a large divergence currently exists, and we show via a historical analysis that this divergence has grown more pronounced over the last decade. The general finding of our study is that continuing to index the web via simple HTML parsing will diminish the effectiveness of retrieval on the modern web, and that the IR community should work toward more sophisticated web page processing in indexing technology.

BooksOnline'12, the fifth workshop in the series, aims to offer a forum for bringing together expertise from academia, industry and libraries to facilitate the exchange of research results and technology in the field of digital libraries with specific focus on online books and complementary social media. The focus of this year's workshop is "engaging reading experiences", starting from the act of deciding what to read, through the exploration and interpretation of a book's content, to sharing the overall experience. Within this overall umbrella theme, the accepted papers naturally showed three salient themes: (1) Search and Discovery, (2) Personalization and Recommendation, and Reading Experiences beyond Text. The contributions demonstrate a range of technologies, including a collaborative tabletop visual approach to support the searching and discovery of books, co-citation methods to enhance document retrieval; exploring open issues in audio-book production to support non-text based reading and improving e-book accessibility; new approaches to recommendation that take into account writing style as well as looking specifically to young readers and their needs in order to develop recommendation tools that consider both content and reading level and match these against the readers' specific interests and reading ability. Following in the theme of the reader playing a central role in the future of our digital era, we are honored to welcome Maribeth Back from FX Palo Alto and Natasa Milic-Frayling from Microsoft Research as our keynote speakers.

Relevance is the central concept of information retrieval. Although its important role is unanimously accepted among researchers, numerous different definitions of the term have emerged over the years. Considerable effort has been put into creating consistent and universally applicable descriptions of relevance in the form of relevance frameworks. Across these various formal systems of relevance, a wide range of relevance criteria has been identified. The probably most frequently used single criterion, that in some applications even becomes a synonym for relevance, is topicality. It expresses a document's topical overlap with the user's information need. For textual resources, it is often estimated based on term co-occurrences between query and document. There is, however, a significant number of further noteworthy relevance criteria. Prominent specimen are: (Currency) determines how recent and up to date the document is. Outdated information may have become invalid over time.

Crowdsourcing is a market of steadily-growing importance upon which both academia and industry increasingly rely. However, this market appears to be inherently infested with a significant share of malicious workers who try to maximise their profits through cheating or sloppiness. This serves to undermine the very merits crowdsourcing has come to represent. Based on previous experience as well as psychological insights, we propose the use of a game in order to attract and retain a larger share of reliable workers to frequently-requested crowdsourcing tasks such as relevance assessments and clustering. In a large-scale comparative study conducted using recent TREC data, we investigate the performance of traditional HIT designs and a game-based alternative that is able to achieve high quality at significantly lower pay rates, facing fewer malicious submissions.

Twitter and Foursquare are two well-connected platforms for sharing information where growing numbers of users post location-related messages. In contrast to the longitude-latitude geotags commonly used online, e.g., on photos and tweets, new place-tags containing category information show more human-readable high-level information rather than a pair of coordinates. This grants an opportunity for better understanding users' physical locations which can be used as context to facilitate other applications, e.g., location context-aware advertisement. In this paper, we verify the assumption that users' current trails contain cues of their future routes. The results from the preliminary experiments show promising performance of a basic Markov Chain-based model.

The BooksOnline Workshop series aims to foster the discussion and exchange of research ideas towards addressing challenges and exploring opportunities around large collections of digital books and complementary media. The fourth workshop in the series, BooksOnline'11 pays special attention to the role of social media and the phenomena of crowdsourcing in the context of online books, which is expected to be key in defining new user experiences in digital libraries and on the Web. The workshop boasts a high quality program, including keynote addresses by Ville Miettinnen, CEO of Microtask and Adam Farquhar, Head of Digital Library Technology at The British Library. From the accepted papers two main themes became salient: 1) Information retrieval and information extraction methods focused on enhancing digital libraries, and 2) Studies and analyses of reading experience and behaviour. This paper provides an overview of the workshop and the accepted contributions.

Twitter is a widely-used social networking service which enables its users to post text-based messages, so-called tweets. POI tags on tweets can show more human-readable high-level information about a place rather than just a pair of coordinates. In this paper, we attempt to predict the POI tag of a tweet based on its textual content and time of posting. Potential applications include accurate positioning when GPS devices fail and disambiguating places located near each other. We consider this task as a ranking problem, i.e., we try to rank a set of candidate POIs according to a tweet by using language and time models. To tackle the sparsity of tweets tagged with POIs, we use web pages retrieved by search engines as an additional source of evidence. From our experiments, we find that users indeed leak some information about their accurate locations in their tweets.

Today children interact more and more frequently with information services. Especially in on-line scenarios there is a great amount of content that is not suitable for their age group. Due to the growing importance and ubiquity of the Internet in today's world, denying children any unsupervised Web access is often not possible. This work presents an automatic way of distinguishing web pages for children from those for adults in order to improve child-appropriate web search engine performance. A range of 80 different features based on findings from cognitive sciences and children's psychology are discussed and evaluated. We conducted a large scale user study on the suitability of web sites and give detailed information about the insights gained. Finally a comparison to traditional web classification methods as well as human annotator performance reveals that our automatic classifier can reach a performance close to that of human agreement.

Children spend significant amounts of time on the Internet. Recent studies showed, that during these periods they are often not under adult supervision. This work presents an automatic approach to identifying suitable web pages for children based on topical and non-topical web page aspects. We discuss the characteristics of children's web sites with respect to recent findings in children's psychology and cognitive sciences. We finally evaluate our approach in a large-scale user study, finding, that it compares favourably to state of the art methods while approximating human performance.