Search

Sparse neural retrievers, such as DeepImpact, uniCOIL and SPLADE, have been introduced recently as an efficient and effective way to perform retrieval with inverted indexes. They aim to learn term importance and, in some cases, document expansions, to provide a more effective document ranking compared to traditional bag-of-words retrieval models such as BM25. However, these sparse neural retrievers have been shown to increase the computational costs and latency of query processing compared to their classical counterparts. To mitigate this, we apply a well-known family of techniques for boosting the efficiency of query processing over inverted indexes: static pruning. We experiment with three static pruning strategies, namely document-centric, term-centric and agnostic pruning, and we assess, over diverse datasets, that these techniques still work with sparse neural retrievers. In particular, static pruning achieves $2\times$ speedup with negligible effectiveness loss ($\leq 2\%$ drop) and, depending on the use case, even $4\times$ speedup with minimal impact on the effectiveness ($\leq 8\%$ drop). Moreover, we show that neural rerankers are robust to candidates from statically pruned indexes., Short accepted at SIGIR 2023. Has extra content in the Appendix that is not present in the ACM version

The MS MARCO-passage dataset has been the main large-scale dataset open to the IR community and it has fostered successfully the development of novel neural retrieval models over the years. But, it turns out that two different corpora of MS MARCO are used in the literature, the official one and a second one where passages were augmented with titles, mostly due to the introduction of the Tevatron code base. However, the addition of titles actually leaks relevance information, while breaking the original guidelines of the MS MARCO-passage dataset. In this work, we investigate the differences between the two corpora and demonstrate empirically that they make a significant difference when evaluating a new method. In other words, we show that if a paper does not properly report which version is used, reproducing fairly its results is basically impossible. Furthermore, given the current status of reviewing, where monitoring state-of-the-art results is of great importance, having two different versions of a dataset is a large problem. This is why this paper aims to report the importance of this issue so that researchers can be made aware of this problem and appropriately report their results., Short paper accepted at SIGIR 2023

This paper describes our participation in the 2022 TREC NeuCLIR challenge. We submitted runs to two out of the three languages (Farsi and Russian), with a focus on first-stage rankers and comparing mono-lingual strategies to Adhoc ones. For monolingual runs, we start from pretraining models on the target language using MLM+FLOPS and then finetuning using the MSMARCO translated to the language either with ColBERT or SPLADE as the retrieval model. While for the Adhoc task, we test both query translation (to the target language) and back-translation of the documents (to English). Initial result analysis shows that the monolingual strategy is strong, but that for the moment Adhoc achieved the best results, with back-translating documents being better than translating queries., Notebook detailing our participation and analysis on the TREC NeuCLIR 2022

Evaluation in Information Retrieval relies on post-hoc empirical procedures, which are time-consuming and expensive operations. To alleviate this, Query Performance Prediction (QPP) models have been developed to estimate the performance of a system without the need for human-made relevance judgements. Such models, usually relying on lexical features from queries and corpora, have been applied to traditional sparse IR methods - with various degrees of success. With the advent of neural IR and large Pre-trained Language Models, the retrieval paradigm has significantly shifted towards more semantic signals. In this work, we study and analyze to what extent current QPP models can predict the performance of such systems. Our experiments consider seven traditional bag-of-words and seven BERT-based IR approaches, as well as nineteen state-of-the-art QPPs evaluated on two collections, Deep Learning '19 and Robust '04. Our findings show that QPPs perform statistically significantly worse on neural IR systems. In settings where semantic signals are prominent (e.g., passage retrieval), their performance on neural models drops by as much as 10% compared to bag-of-words approaches. On top of that, in lexical-oriented scenarios, QPPs fail to predict performance for neural IR systems on those queries where they differ from traditional approaches the most.

Parameter-Efficient transfer learning with Adapters have been studied in Natural Language Processing (NLP) as an alternative to full fine-tuning. Adapters are memory-efficient and scale well with downstream tasks by training small bottle-neck layers added between transformer layers while keeping the large pretrained language model (PLMs) frozen. In spite of showing promising results in NLP, these methods are under-explored in Information Retrieval. While previous studies have only experimented with dense retriever or in a cross lingual retrieval scenario, in this paper we aim to complete the picture on the use of adapters in IR. First, we study adapters for SPLADE, a sparse retriever, for which adapters not only retain the efficiency and effectiveness otherwise achieved by finetuning, but are memory-efficient and orders of magnitude lighter to train. We observe that Adapters-SPLADE not only optimizes just 2\% of training parameters, but outperforms fully fine-tuned counterpart and existing parameter-efficient dense IR models on IR benchmark datasets. Secondly, we address domain adaptation of neural retrieval thanks to adapters on cross-domain BEIR datasets and TripClick. Finally, we also consider knowledge sharing between rerankers and first stage rankers. Overall, our study complete the examination of adapters for neural IR, Comment: accepted at ECIR'23

This paper presents the AToMiC (Authoring Tools for Multimedia Content) dataset, designed to advance research in image/text cross-modal retrieval. While vision-language pretrained transformers have led to significant improvements in retrieval effectiveness, existing research has relied on image-caption datasets that feature only simplistic image-text relationships and underspecified user models of retrieval tasks. To address the gap between these oversimplified settings and real-world applications for multimedia content creation, we introduce a new approach for building retrieval test collections. We leverage hierarchical structures and diverse domains of texts, styles, and types of images, as well as large-scale image-document associations embedded in Wikipedia. We formulate two tasks based on a realistic user model and validate our dataset through retrieval experiments using baseline models. AToMiC offers a testbed for scalable, diverse, and reproducible multimedia retrieval research. Finally, the dataset provides the basis for a dedicated track at the 2023 Text Retrieval Conference (TREC), and is publicly available at https://github.com/TREC-AToMiC/AToMiC.

Middle training methods aim to bridge the gap between the Masked Language Model (MLM) pre-training and the final finetuning for retrieval. Recent models such as CoCondenser, RetroMAE, and LexMAE argue that the MLM task is not sufficient enough to pre-train a transformer network for retrieval and hence propose various tasks to do so. Intrigued by those novel methods, we noticed that all these models used different finetuning protocols, making it hard to assess the benefits of middle training. We propose in this paper a benchmark of CoCondenser, RetroMAE, and LexMAE, under the same finetuning conditions. We compare both dense and sparse approaches under various finetuning protocols and middle training on different collections (MS MARCO, Wikipedia or Tripclick). We use additional middle training baselines, such as a standard MLM finetuning on the retrieval collection, optionally augmented by a CLS predicting the passage term frequency. For the sparse approach, our study reveals that there is almost no statistical difference between those methods: the more effective the finetuning procedure is, the less difference there is between those models. For the dense approach, RetroMAE using MS MARCO as middle-training collection shows excellent results in almost all the settings. Finally, we show that middle training on the retrieval collection, thus adapting the language model to it, is a critical factor. Overall, a better experimental setup should be adopted to evaluate middle training methods. Code available at https://github.com/naver/splade/tree/benchmarch-SIGIR23

This paper describes our participation to the 2022 TREC Deep Learning challenge. We submitted runs to all four tasks, with a focus on the full retrieval passage task. The strategy is almost the same as 2021, with first stage retrieval being based around SPLADE, with some added ensembling with ColBERTv2 and DocT5. We also use the same strategy of last year for the second stage, with an ensemble of re-rankers trained using hard negatives selected by SPLADE. Initial result analysis show that the strategy is still strong, but is still unclear to us what next steps should we take.

Finetuning Pretrained Language Models (PLM) for IR has been de facto the standard practice since their breakthrough effectiveness few years ago. But, is this approach well understood? In this paper, we study the impact of the pretraining collection on the final IR effectiveness. In particular, we challenge the current hypothesis that PLM shall be trained on a large enough generic collection and we show that pretraining from scratch on the collection of interest is surprisingly competitive with the current approach. We benchmark first-stage ranking rankers and cross-encoders for reranking on the task of general passage retrieval on MSMARCO, Mr-Tydi for Arabic, Japanese and Russian, and TripClick for specific domain. Contrary to popular belief, we show that, for finetuning first-stage rankers, models pretrained solely on their collection have equivalent or better effectiveness compared to more general models. However, there is a slight effectiveness drop for rerankers pretrained only on the target collection. Overall, our study sheds a new light on the role of the pretraining collection and should make our community ponder on building specialized models by pretraining from scratch. Last but not least, doing so could enable better control of efficiency, data bias and replicability, which are key research questions for the IR community., Comment: Accepted at ECIR 2023

Latency and efficiency issues are often overlooked when evaluating IR models based on Pretrained Language Models (PLMs) in reason of multiple hardware and software testing scenarios. Nevertheless, efficiency is an important part of such systems and should not be overlooked. In this paper, we focus on improving the efficiency of the SPLADE model since it has achieved state-of-the-art zero-shot performance and competitive results on TREC collections. SPLADE efficiency can be controlled via a regularization factor, but solely controlling this regularization has been shown to not be efficient enough. In order to reduce the latency gap between SPLADE and traditional retrieval systems, we propose several techniques including L1 regularization for queries, a separation of document/query encoders, a FLOPS-regularized middle-training, and the use of faster query encoders. Our benchmark demonstrates that we can drastically improve the efficiency of these models while increasing the performance metrics on in-domain data. To our knowledge, {we propose the first neural models that, under the same computing constraints, \textit{achieve similar latency (less than 4ms difference) as traditional BM25}, while having \textit{similar performance (less than 10\% MRR@10 reduction)} as the state-of-the-art single-stage neural rankers on in-domain data}., Accepted at SIGIR 2022 and ReNeuIR. Huggingface models available at https://huggingface.co/naver and code at https://github.com/naver/splade

Neural retrievers based on dense representations combined with Approximate Nearest Neighbors search have recently received a lot of attention, owing their success to distillation and/or better sampling of examples for training -- while still relying on the same backbone architecture. In the meantime, sparse representation learning fueled by traditional inverted indexing techniques has seen a growing interest, inheriting from desirable IR priors such as explicit lexical matching. While some architectural variants have been proposed, a lesser effort has been put in the training of such models. In this work, we build on SPLADE -- a sparse expansion-based retriever -- and show to which extent it is able to benefit from the same training improvements as dense models, by studying the effect of distillation, hard-negative mining as well as the Pre-trained Language Model initialization. We furthermore study the link between effectiveness and efficiency, on in-domain and zero-shot settings, leading to state-of-the-art results in both scenarios for sufficiently expressive models., Comment: Accepted at SIGIR22 as a short paper (this work is the extension of SPLADE v2)

We are interested in the problem of cross-modal retrieval for web image search, where the goal is to retrieve images relevant to a text query. While most of the current approaches for cross-modal retrieval revolve around learning how to represent text and images in a shared latent space, we take a different direction: we propose to generalize the cross-modal relevance feedback mechanism, a simple yet effective unsupervised method, that relies on standard information retrieval heuristics and the choice of a few hyper-parameters. We show that we can cast it as a supervised representation learning problem on graphs, using graph convolutions operating jointly over text and image features, namely cross-modal graph convolutions. The proposed architecture directly learns how to combine image and text features for the ranking task, while taking into account the context given by all the other elements in the set of images to be (re-)ranked. We validate our approach on two datasets: a public dataset from a MediaEval challenge, and a small sample of proprietary image search query logs, referred as WebQ. Our experiments demonstrate that our model improves over standard baselines.

Pre-trained Language Models have recently emerged in Information Retrieval as providing the backbone of a new generation of neural systems that outperform traditional methods on a variety of tasks. However, it is still unclear to what extent such approaches generalize in zero-shot conditions. The recent BEIR benchmark provides partial answers to this question by comparing models on datasets and tasks that differ from the training conditions. We aim to address the same question by comparing models under more explicit distribution shifts. To this end, we build three query-based distribution shifts within MS MARCO (query-semantic, query-intent, query-length), which are used to evaluate the three main families of neural retrievers based on BERT: sparse, dense, and late-interaction -- as well as a monoBERT re-ranker. We further analyse the performance drops between the train and test query distributions. In particular, we experiment with two generalization indicators: the first one based on train/test query vocabulary overlap, and the second based on representations of a trained bi-encoder. Intuitively, those indicators verify that the further away the test set is from the train one, the worse the drop in performance. We also show that models respond differently to the shifts -- dense approaches being the most impacted. Overall, our study demonstrates that it is possible to design more controllable distribution shifts as a tool to better understand generalization of IR models. Finally, we release the MS MARCO query subsets, which provide an additional resource to benchmark zero-shot transfer in Information Retrieval., Accepted at ECIR 2023

We propose a Composite Code Sparse Autoencoder (CCSA) approach for Approximate Nearest Neighbor (ANN) search of document representations based on Siamese-BERT models. In Information Retrieval (IR), the ranking pipeline is generally decomposed in two stages: the first stage focus on retrieving a candidate set from the whole collection. The second stage re-ranks the candidate set by relying on more complex models. Recently, Siamese-BERT models have been used as first stage ranker to replace or complement the traditional bag-of-word models. However, indexing and searching a large document collection require efficient similarity search on dense vectors and this is why ANN techniques come into play. Since composite codes are naturally sparse, we first show how CCSA can learn efficient parallel inverted index thanks to an uniformity regularizer. Second, CCSA can be used as a binary quantization method and we propose to combine it with the recent graph based ANN techniques. Our experiments on MSMARCO dataset reveal that CCSA outperforms IVF with product quantization. Furthermore, CCSA binary quantization is beneficial for the index size, and memory usage for the graph-based HNSW method, while maintaining a good level of recall and MRR. Third, we compare with recent supervised quantization methods for image retrieval and find that CCSA is able to outperform them., Comment: 24 pages, longer version of short paper accepted at SIGIR 2021

Learning with noisy labels is an active research area for image classification. However, the effect of noisy labels on image retrieval has been less studied. In this work, we propose a noise-resistant method for image retrieval named Teacher-based Selection of Interactions, T-SINT, which identifies noisy interactions, ie. elements in the distance matrix, and selects correct positive and negative interactions to be considered in the retrieval loss by using a teacher-based training setup which contributes to the stability. As a result, it consistently outperforms state-of-the-art methods on high noise rates across benchmark datasets with synthetic noise and more realistic noise., Comment: Accepted at WACV 2022. 13 pages, 5 figures

Multilingual NMT has become an attractive solution for MT deployment in production. But to match bilingual quality, it comes at the cost of larger and slower models. In this work, we consider several ways to make multilingual NMT faster at inference without degrading its quality. We experiment with several "light decoder" architectures in two 20-language multi-parallel settings: small-scale on TED Talks and large-scale on ParaCrawl. Our experiments demonstrate that combining a shallow decoder with vocabulary filtering leads to more than twice faster inference with no loss in translation quality. We validate our findings with BLEU and chrF (on 380 language pairs), robustness evaluation and human evaluation., EMNLP 2021 long paper

In neural Information Retrieval, ongoing research is directed towards improving the first retriever in ranking pipelines. Learning dense embeddings to conduct retrieval using efficient approximate nearest neighbors methods has proven to work well. Meanwhile, there has been a growing interest in learning sparse representations for documents and queries, that could inherit from the desirable properties of bag-of-words models such as the exact matching of terms and the efficiency of inverted indexes. In this work, we present a new first-stage ranker based on explicit sparsity regularization and a log-saturation effect on term weights, leading to highly sparse representations and competitive results with respect to state-of-the-art dense and sparse methods. Our approach is simple, trained end-to-end in a single stage. We also explore the trade-off between effectiveness and efficiency, by controlling the contribution of the sparsity regularization., Comment: 5 pages, SIGIR'21 short paper

National audience; Les modèles de RI basés sur les Transformers sont aujourd’hui état de l’art en Recherche d’Information ad-hoc, mais leur comportement reste encore incompris. Des travaux récents ont montré que BERT ne satisfait pas les axiomes classiques de la RI. Nous proposons d’étudier le processus d’appariement par l’analyse de l’importance des termes et des mécanismes d’appariement exact et sémantique. Même si les axiomes classiques ne sont pas formellement vérifiés, notre analyse révèle que des modèles comme ColBERT: (i) sont capables d’inclure une notion d’importance du terme ; (ii) s’appuient sur des correspondances exactes pour les termes importants.

Graph neural networks (GNNs) have recently gained significant momentum in the recommendation community, demonstrating state-of-the-art performance in top-k recommendation and next-item recommendation. Despite promising results on GNN-based recommendation and search, most of the current GNN research remains essentially concentrated on more traditional tasks such as classification or regression. The GReS workshop on Graph Neural Networks for Recommendation and Search is then a first endeavor to bridge the gap between the RecSys and GNN communities, and promote recommendation and search problems amongst GNN practitioners.

The ColBERT model has recently been proposed as an effective BERT based ranker. By adopting a late interaction mechanism, a major advantage of ColBERT is that document representations can be precomputed in advance. However, the big downside of the model is the index size, which scales linearly with the number of tokens in the collection. In this paper, we study various designs for ColBERT models in order to attack this problem. While compression techniques have been explored to reduce the index size, in this paper we study token pruning techniques for ColBERT. We compare simple heuristics, as well as a single layer of attention mechanism to select the tokens to keep at indexing time. Our experiments show that ColBERT indexes can be pruned up to 30\% on the MS MARCO passage collection without a significant drop in performance. Finally, we experiment on MS MARCO documents, which reveal several challenges for such mechanism., Comment: 5 pages

In neural Information Retrieval (IR), ongoing research is directed towards improving the first retriever in ranking pipelines. Learning dense embeddings to conduct retrieval using efficient approximate nearest neighbors methods has proven to work well. Meanwhile, there has been a growing interest in learning \emph{sparse} representations for documents and queries, that could inherit from the desirable properties of bag-of-words models such as the exact matching of terms and the efficiency of inverted indexes. Introduced recently, the SPLADE model provides highly sparse representations and competitive results with respect to state-of-the-art dense and sparse approaches. In this paper, we build on SPLADE and propose several significant improvements in terms of effectiveness and/or efficiency. More specifically, we modify the pooling mechanism, benchmark a model solely based on document expansion, and introduce models trained with distillation. We also report results on the BEIR benchmark. Overall, SPLADE is considerably improved with more than $9$\% gains on NDCG@10 on TREC DL 2019, leading to state-of-the-art results on the BEIR benchmark., Comment: 5 pages. arXiv admin note: substantial text overlap with arXiv:2107.05720

We are interested in the problem of cross-modal retrieval for web image search, where the goal is to retrieve images relevant to a text query. While most of the current approaches for cross-modal retrieval revolve around learning how to represent text and images in a shared latent space, we take a different direction: we propose to generalize the cross-modal relevance feedback mechanism, a simple yet effective unsupervised method, that relies on standard information retrieval heuristics and the choice of a few hyper-parameters. We show that we can cast it as a supervised representation learning problem on graphs, using graph convolutions operating jointly over text and image features, namely cross-modal graph convolutions. The proposed architecture directly learns how to combine image and text features for the ranking task, while taking into account the context given by all the other elements in the set of images to be (re-)ranked. We validate our approach on two datasets: a public dataset from a MediaEval challenge, and a small sample of proprietary image search query logs, referred as WebQ. Our experiments demonstrate that our model improves over standard baselines.

We present in this paper experiments on Table Recognition in hand-written registry books. We first explain how the problem of row and column detection is modeled, and then compare two Machine Learning approaches (Conditional Random Field and Graph Convolutional Network) for detecting these table elements. Evaluation was conducted on death records provided by the Archive of the Diocese of Passau. Both methods show similar results, a 89 F1 score, a quality which allows for Information Extraction. Software and dataset are open source/data., Comment: DAS 2018

Exploiting large pretrained models for various NMT tasks have gained a lot of visibility recently. In this work we study how BERT pretrained models could be exploited for supervised Neural Machine Translation. We compare various ways to integrate pretrained BERT model with NMT model and study the impact of the monolingual data used for BERT training on the final translation quality. We use WMT-14 English-German, IWSLT15 English-German and IWSLT14 English-Russian datasets for these experiments. In addition to standard task test set evaluation, we perform evaluation on out-of-domain test sets and noise injected test sets, in order to assess how BERT pretrained representations affect model robustness., Comment: 10 pages

In this paper, we present an approach to generate pleasant running tours for unknown environments. We start from a first version, able to generate basic pleasant tours, but without explicitly catering for elevation. Based on observations on how runners appreciate elevation gain and steepness and on interviews with local runners, we identified corresponding additional needs. Indeed, runners usually have specific implicit or explicit objectives or limits with respect to the elevation gain they want to target or to avoid with a tour. In consequence, we extend our first approach. We expose the algorithm we have defined to address elevation constraints during tour generation and the underlying intuitions. An important differentiator with prior art is that we address elevation constraints during the tour generation phase and not a posteriori. This means that our approach is able to efficiently generate new tours that match the user’s constraints, while prior art rather finds matching tours by searching a posteriori among a set of available tours, and only finds matching ones if they are already present.

Multimedia collections are more than ever growing in size and diversity. Effective multimedia retrieval systems are thus critical to access these datasets from the end-user perspective and in a scalable way. We are interested in repositories of image/text multimedia objects and we study multimodal information fusion techniques in the context of content-based multimedia information retrieval. We focus on graph-based methods, which have proven to provide state-of-the-art performances. We particularly examine two such methods: cross-media similarities and random-walk-based scores. From a theoretical viewpoint, we propose a unifying graph-based framework, which encompasses the two aforementioned approaches. Our proposal allows us to highlight the core features one should consider when using a graph-based technique for the combination of visual and textual information. We compare cross-media and random-walk-based results using three different real-world datasets. From a practical standpoint, our extended empirical analyses allow us to provide insights and guidelines about the use of graph-based methods for multimodal information fusion in content-based multimedia information retrieval.

Domain Adaptation (DA) exploits labeled data and models from similar domains in order to alleviate the annotation burden when learning a model in a new domain. Our contribution to the field is three-fold. First, we propose a new dataset, LandMarkDA, to study the adaptation between landmark place recognition models trained with different artistic image styles, such as photos, paintings and drawings. The new LandMarkDA proposes new adaptation challenges, where current deep architectures show their limits. Second, we propose an experimental study of recent shallow and deep adaptation networks, based on using Maximum Mean Discrepancy to bridge the domain gap. We study different design choices for these models by varying the network architectures and evaluate them on OFF31 and the new LandMarkDA collections. We show that shallow networks can still be competitive under an appropriate feature extraction. Finally, we also benchmark a new DA method that successfully combines the artistic image style-transfer with deep discrepancy-based networks.

A large majority of existing domain adaptation methods makes an assumption of freely available labeled source and unlabeled target data. They exploit the discrepancy between their distributions and build representations common to both target and source domains. In reality, such a simplifying assumption rarely holds, since source data are routinely a subject of legal and contractual constraints between data owners and data customers. Despite a limited access to source domain data, decision-making procedures might be available in the form of, e.g., classification rules trained on the source and made ready for a direct deployment and later reuse. In other cases, the owner of a source data is allowed to share a few representative examples such as class means. The aim of this chapter is therefore to address the domain adaptation problem in such constrained real world applications, i.e. where the reuse of source domain data is limited to classification rules or a few representative examples. As a solution, we extend recent techniques based on feature corruption and their marginalization, both considering supervised and unsupervised domain adaptation settings. The proposed models are tested and compared on private and publicly available source datasets showing significant performance gains despite the absence of the whole source data and shortage of labeled target data.

We propose an extended framework for marginalized domain adaptation, aimed at addressing unsupervised, supervised and semi-supervised scenarios. We argue that the denoising principle should be extended to explicitly promote domain-invariant features as well as help the classification task. Therefore we propose to jointly learn the data auto-encoders and the target classifiers. First, in order to make the denoised features domain-invariant, we propose a domain regularization that may be either a domain prediction loss or a maximum mean discrepancy between the source and target data. The noise marginalization in this case is reduced to solving the linear matrix system $AX=B$ which has a closed-form solution. Second, in order to help the classification, we include a class regularization term. Adding this component reduces the learning problem to solving a Sylvester linear matrix equation $AX+BX=C$, for which an efficient iterative procedure exists as well. We did an extensive study to assess how these regularization terms improve the baseline performance in the three domain adaptation scenarios and present experimental results on two image and one text benchmark datasets, conventionally used for validating domain adaptation methods. We report our findings and comparison with state-of-the-art methods.

The overwhelming majority of existing domain adaptation methods makes an assumption of freely available source domain data. An equal access to both source and target data makes it possible to measure the discrepancy between their distributions and to build representations common to both target and source domains. In reality, such a simplifying assumption rarely holds, since source data are routinely a subject of legal and contractual constraints between data owners and data customers. When source domain data can not be accessed, decision making procedures are often available for adaptation nevertheless. These procedures are often presented in the form of classification, identification, ranking etc. rules trained on source data and made ready for a direct deployment and later reuse. In other cases, the owner of a source data is allowed to share a few representative examples such as class means. In this paper we address the domain adaptation problem in real world applications, where the reuse of source domain data is limited to classification rules or a few representative examples. We extend the recent techniques of feature corruption and their marginalization, both in supervised and unsupervised settings. We test and compare them on private and publicly available source datasets and show that significant performance gains can be achieved despite the absence of source data and shortage of labeled target data.

Background: Community children’s nursing services (CCNS) provides nursing and supportive care, ranging from relatively simple to highly technological interventions, to children and young people (0Y18 years) within the family home. Objectives: The aim of this study was to elicit the perspectives on and experiences about CCNS in England in relation to things that are working well or that could be improved and the vision for services. Methods: Using a qualitative approach, underpinned by an Appreciative Inquiry philosophy, researchers worked closely with advisory groups (six children and young people, four parents, and five professionals) to design and implement the study. Arts-based participatory appreciative workshops were held in nine regional locations in England; shared activities were undertaken to elicit participants’ experiences. Participants unable to attend the workshops were able to contribute via semistructured interview (face to face or by telephone) or by a dedicated blog or e-mail. Results: Thematic analysis was used, and 214 people participated: families (n = 82), children (n = 27), and professionals or stakeholders (n = 105). Things that were working well were effective communication, robust leadership, actively enabling the child’s care to be sustained at home, and partnerships based on mutual trust. Problems relating to feelings of marginalization, ineffective commissioning, underprovision of services and hours of service availability, lack of equipment or resources, and poor communication between services and settings were areas of concern. The vision for CCNS was for a flexible, equitable, and accessible service that supported children’s and families’ needs and choices and which enabled parents to be parents first rather thancaregivers. Discussion: Care at home reduces the disruption to children’s and families’ lives and can empower them to make decisions and control routines and practices. Having CCNS situated within larger teams with strong interdisciplinary andtransdisciplinary practices seems to offer the best benefits for children and families. Findings from this study have directly influenced government policy and practice. Further research is needed to determine efficacy of particular models and practices.

Nous proposons dans ce papier une analyse theorique des modeles de Pseudo-Relevance Feedback (PRF), expliquant pourquoi certains modeles comme les modeles d’information sont plus performants. Pour ce faire, nous proposons un ensemble de proprietes pour les fonctions de PRF, dont une portant sur la frequence documentaire (DF) qui est validee experimentalement. Cette etude theorique revele que plusieurs modeles de reference ne respectent pas la propriete sur l’effet IDF ou sur l’effet DF. Les modeles satisfaisant toutes les proprietes surpassent ces modeles en partie deficients.

Nous presentons dans cet article une vue analytique des contraintes heuristiques recemment proposees pour les fonctions d’ordonnancement (retrieval function). Ces caracterisations permettent ainsi de tester simplement si un modele de recherche d’information (RI) respecte ces contraintes ou non. De plus, nous examinons un certain nombre de resultats empiriques sur les distributions de frequences de mots et le role central joue par le phenomene de rafale, pour lequel nous proposons une definition formelle. Nous introduisons ensuite une nouvelle famille de modeles probabilistes pour la RI, fondee sur la notion d’information. Lorsque la loi de probabilite sous-jacente est capable de modeliser le phenomene de rafale, alors le modele devient naturellement valide au sens des contraintes heuristiques. Les distributions log-logistique et SPL sont presentees dans ce contexte et les experiences, menees sur trois collections differentes, illustrent le comportement adequat de ces modeles ; ils surpassent Okapi BM25 et les modeles de langues, avec lissage de Jelinek-Mercer ou de Dirichlet, a la fois pour la precision moyenne et la precision en tete de liste, et fournissent des resultats similaires aux modeles DFR (Divergence from Randomness) tout en les simplifiant.

This paper deals with multimedia information access. We propose two new approaches for hybrid text-image information processing that can be straightforwardly generalized to the more general multimodal scenario. Both approaches fall in the trans-media pseudo-relevance feedback category. Our first method proposes using a mixture model of the aggregate components, considering them as a single relevance concept. In our second approach, we define trans-media similarities as an aggregation of monomodal similarities between the elements of the aggregate and the new multimodal object. We also introduce the monomodal similarity measures for text and images that serve as basic components for both proposed trans-media similarities. We show how one can frame a large variety of problem in order to address them with the proposed techniques: image annotation or captioning, text illustration and multimedia retrieval and clustering. Finally, we present how these methods can be integrated in two applications: a travel blog assistant system and a tool for browsing the Wikipedia taking into account the multimedia nature of its content.

Access to data is critical to any machine learning component aimed at training an accurate predictive model. In reality, data is often a subject of technical and legal constraints. Data may contain sensitive topics and data owners are often reluctant to share them. Instead of access to data, they make available decision making procedures to enable predictions on new data. Under the black box classifier constraint, we build an effective domain adaptation technique which adapts classifier predictions in a transductive setting. We run experiments on text categorization datasets and show that significant gains can be achieved, especially in the unsupervised case where no labels are available in the target domain.

Finding domain invariant features is critical for successful domain adaptation and transfer learning. However, in the case of unsupervised adaptation, there is a significant risk of overfitting on source training data. Recently, a regularization for domain adaptation was proposed for deep models by (Ganin and Lempitsky, 2015). We build on their work by suggesting a more appropriate regularization for denoising autoencoders. Our model remains unsupervised and can be computed in a closed form. On standard text classification adaptation tasks, our approach yields the state of the art results, with an important reduction of the learning cost.

We propose to extend the marginalized denoising autoencoder (MDA) framework with a domain regularization whose aim is to denoise both the source and target data in such a way that the features become domain invariant and the adaptation gets easier. The domain regularization, based either on the maximum mean discrepancy (MMD) measure or on the domain prediction, aims to reduce the distance between the source and the target data. We also exploit the source class labels as another way to regularize the loss, by using a domain classifier regularizer. We show that in these cases, the noise marginalization gets reduced to solving either the linear matrix system \(\mathbf{A}\mathbf{X}=\mathbf{B}\), for which there exists a closed-form solution, or to a Sylvester linear matrix equation \(\mathbf{A}\mathbf{X}+\mathbf{X}\mathbf{B}=\mathbf{C}\) that can be solved efficiently using the Bartels-Stewart algorithm. We did an extensive study on how these regularization terms improve the baseline performance and we present experiments on three image benchmark datasets, conventionally used for domain adaptation methods. We report our findings and comparisons with state-of-the-art methods.

We address the problem of domain adaptation (DA) from one or multiple source domains to a target domain. Most of the existing DA methods assume that source data is largely available. Such an assumption rarely holds in real applications, for both technical and legal reasons. More realistic are situations where source domain observations become quickly unavailable, but only some domain representatives can be retained, either as source instances or as their aggregation. In this paper therefore we focus on the Domain Specific Class Means (DSCM) classifier [5] that can handle such scenario and we combine it with the sMDA framework [4]. We show, on a variety of datasets and tasks, that the method can be applied successfully even when no labeled target is available and also that it can provide performance comparable to the case where dense knowledge (all source data) is available.

In this paper we address the problem of domain adaptation using multiple source domains. We extend the XRCE contribution to Clef'14 Domain Adaptation challengei¾?[6] with the new methods and new datasets. We describe a new class of domain adaptation technique based on stacked marginalized denoising autoencoders sMDA. It aims at extracting and denoising features common to both source and target domains in the unsupervised mode. Noise marginalization allows to obtain a closed form solution and to considerably reduce the training time. We build a classification system which compares sMDA combined with SVM or with Domain Specific Class Mean classifiers to the state-of-the art in both unsupervised and semi-supervised settings. We report the evaluation results for a number of image and text datasets.

On-street parking, just as any publicly owned utility, is used inefficiently if access is free or priced very far from market rates. This paper introduces a novel demand management solution: using data from dedicated occupancy sensors an iteration scheme updates parking rates to better match demand. The new rates encourage parkers to avoid peak hours and peak locations and reduce congestion and underuse. The solution is deliberately simple so that it is easy to understand, easily seen to be fair and leads to parking policies that are easy to remember and act upon. We study the convergence properties of the iteration scheme and prove that it converges to a reasonable distribution for a very large class of models. The algorithm is in use to change parking rates in over 6000 spaces in downtown Los Angeles since June 2012 as part of the LA Express Park project. Initial results are encouraging with a reduction of congestion and underuse, while in more locations rates were decreased than increased.

Multimedia collections are more than ever growing in size and diversity. Effective multimedia retrieval systems are thus critical to access these datasets from the end-user perspective and in a scalable way. We are interested in repositories of image/text multimedia objects and we study multimodal information fusion techniques in the context of content based multimedia information retrieval. We focus on graph based methods which have proven to provide state-of-the-art performances. We particularly examine two of such methods : cross-media similarities and random walk based scores. From a theoretical viewpoint, we propose a unifying graph based framework which encompasses the two aforementioned approaches. Our proposal allows us to highlight the core features one should consider when using a graph based technique for the combination of visual and textual information. We compare cross-media and random walk based results using three different real-world datasets. From a practical standpoint, our extended empirical analysis allow us to provide insights and guidelines about the use of graph based methods for multimodal information fusion in content based multimedia information retrieval., Comment: An extended version of the paper: Visual and Textual Information Fusion in Multimedia Retrieval using Semantic Filtering and Graph based Methods, by J. Ah-Pine, G. Csurka and S. Clinchant, submitted to ACM Transactions on Information Systems

The field of parking is going through a period of extreme innovation. Cities in the United States are now exploring new technology to improve on-street parking. One such innovation is dynamic pricing based on sensors and smart meters. This paper presents the results of two surveys and an ethnographic study in the context of LA Express ParkTM to understand users’ behaviors, knowledge and perceptions around parking. Survey results demonstrated that a high number of users misunderstood one of three tested stickers that convey time of day pricing. Furthermore, after discovering the availability of cheaper parking spots nearby, people expressed willingness to change their future behavior to park in those places. Ethnographic field studies found that it is common for many parkers to use handicapped placards for over eight hours in one parking session. A percentage of these parkers may be using placards illegally. We propose that increasing some parking restrictions during the day may curb placard use by making it more difficult to park for long periods.

Social monitoring platforms are software services that enable the rapid analysis of massive amounts of information (mostly text messages) expressed in social networks. Their usage today mostly focuses on marketing industries that need to reach their customers in a proper way. In the e-government domain, however, such tools have mostly been used by political parties as a support to election campaigns, but few have been actually used to provide a comprehensive understanding of the citizens' needs in everyday life. In this chapter, the authors present a set of data analytics tools that can help public authorities to extract and summarize textual content from Internet forums and social media feeds. There are many potential applications of these tools, such as the visualization of the main political discussion in the city, early detection of disagreement with the local politics, and city services connected to social media.