This paper addresses writer identification and writer retrieval which is considered as a challenging problem in the document analysis and recognition field. In this work, a novel pipeline is proposed for the problem at hand by employing a unified neural network architecture consisting of the ResNet-20 as a feature extractor and an integrated NetVLAD layer, inspired by the vector of locally aggregated descriptors (VLAD), in the head of the latter part. Having defined this architecture, the triplet semi-hard loss function is used to directly learn an embedding for individual input image patches. Subsequently, generalized max-pooling technique is employed for the aggregation of embedded descriptors of each handwritten image. Also, a novel re-ranking strategy is introduced for the task of identification and retrieval based on $k$-reciprocal nearest neighbors, and it is shown that the pipeline can benefit tremendously from this step. Experimental evaluation has been done on the three publicly available datasets: the ICDAR 2013, CVL, and KHATT datasets. Results indicate that while we perform comparably to the state-of-the-art on the KHATT, our writer identification and writer retrieval pipeline achieves superior performance on the ICDAR 2013 and CVL datasets in terms of mAP.
Prediction of protein-ligand complexes for flexible proteins remains still a challenging problem in computational structural biology and drug design. Here we present two novel deep neural network approaches with significant improvement in efficiency and accuracy of binding mode prediction on a large and diverse set of protein systems compared to standard docking. Whereas the first graph convolutional network is used for re-ranking poses the second approach aims to generate and rank poses independent of standard docking approaches. This novel approach relies on the prediction of distance matrices between ligand atoms and protein C_alpha atoms thus incorporating side-chain flexibility implicitly.
Named entity discovery (NED) is an important information retrieval problem that can be decomposed into two sub-problems. The first sub-problem, named entity recognition (NER), aims to tag pre-defined sets of words in a vocabulary (called "named entities": names, places, locations, ...) when they appear in natural language. The second subproblem, named entity linking/identification (NEL), considers these entity mentions as queries to be identified in a pre-existing database. In this paper, we consider the NEL problem, and assume a set of queries (or mentions) that have to be identified within a knowledge base. This knowledge base is represented by a text database paired with a semantic graph. We present state-of-the-art methods in NEL, and propose a 2-step method for individual identification of named entities. Our approach is well-motivated by the limitations brought by recent deep learning approaches that lack interpratability, and require lots of parameter tuning along with large volume of annotated data. First of all, we propose a filtering algorithm designed with information retrieval and text mining techniques, aiming to maximize precision at K (typically for 5 <= K <=20). Then, we introduce two graph-based methods for named entity identification to maximize precision at 1 by re-ranking the remaining top entity candidates. The first identification method is using parametrized graph mining, and the second similarity with graph kernels. Our approach capitalizes on a fine-grained classification of entities from annotated web data. We present our algorithms in details, and show experimentally on standard datasets (NIST TAC-KBP, CONLL/AIDA) their performance in terms of precision are better than any graph-based method reported, and competitive with state-of-the-art systems. Finally, we conclude on the advantages of our graph-based approach compared to recent deep learning methods.
The trade-off between relevance and fairness in personalized recommendations has been explored in recent works, with the goal of minimizing learned discrimination towards certain demographics while still producing relevant results. We present a fairness-aware variation of the Maximal Marginal Relevance (MMR) re-ranking method which uses representations of demographic groups computed using a labeled dataset. This method is intended to incorporate fairness with respect to these demographic groups. We perform an experiment on a stock photo dataset and examine the trade-off between relevance and fairness against a well known baseline, MMR, by using human judgment to examine the results of the re-ranking when using different fractions of a labeled dataset, and by performing a quantitative analysis on the ranked results of a set of query images. We show that our proposed method can incorporate fairness in the ranked results while obtaining higher precision than the baseline, while our case study shows that even a limited amount of labeled data can be used to compute the representations to obtain fairness. This method can be used as a post-processing step for recommender systems and search.
Personalized recommendation brings about novel challenges in ensuring fairness, especially in scenarios in which users are not the only stakeholders involved in the recommender system. For example, the system may want to ensure that items from different providers have a fair chance of being recommended. To solve this problem, we propose a Fairness-Aware Re-ranking algorithm (FAR) to balance the ranking quality and provider-side fairness. We iteratively generate the ranking list by trading off between accuracy and the coverage of the providers. Although fair treatment of providers is desirable, users may differ in their receptivity to the addition of this type of diversity. Therefore, personalized user tolerance towards provider diversification is incorporated. Experiments are conducted on both synthetic and real-world data. The results show that our proposed re-ranking algorithm can significantly promote fairness with a slight sacrifice in accuracy and can do so while being attentive to individual user differences.
Conventional multi-view re-ranking methods usually perform asymmetrical matching between the region of interest (ROI) in the query image and the whole target image for similarity computation. Due to the inconsistency in the visual appearance, this practice tends to degrade the retrieval accuracy particularly when the image ROI, which is usually interpreted as the image objectness, accounts for a smaller region in the image. Since Privileged Information (PI), which can be viewed as the image prior, enables well characterizing the image objectness, we are aiming at leveraging PI for further improving the performance of the multi-view re-ranking accuracy in this paper. Towards this end, we propose a discriminative multi-view re-ranking approach in which both the original global image visual contents and the local auxiliary PI features are simultaneously integrated into a unified training framework for generating the latent subspaces with sufficient discriminating power. For the on-the-fly re-ranking, since the multi-view PI features are unavailable, we only project the original multi-view image representations onto the latent subspace, and thus the re-ranking can be achieved by computing and sorting the distances from the multi-view embeddings to the separating hyperplane. Extensive experimental evaluations on the two public benchmarks Oxford5k and Paris6k reveal our approach provides further performance boost for accurate image re-ranking, whilst the comparative study demonstrates the advantage of our method against other multi-view re-ranking methods.
Vocabulary mismatch is a central problem in information retrieval (IR), i.e., the relevant documents may not contain the same (symbolic) terms of the query. Recently, neural representations have shown great success in capturing semantic relatedness, leading to new possibilities to alleviate the vocabulary mismatch problem in IR. However, most existing efforts in this direction have been devoted to the re-ranking stage. That is to leverage neural representations to help re-rank a set of candidate documents, which are typically obtained from an initial retrieval stage based on some symbolic index and search scheme (e.g., BM25 over the inverted index). This naturally raises a question: if the relevant documents have not been found in the initial retrieval stage due to vocabulary mismatch, there would be no chance to re-rank them to the top positions later. Therefore, in this paper, we study the problem how to employ neural representations to improve the recall of relevant documents in the initial retrieval stage. Specifically, to meet the efficiency requirement of the initial stage, we introduce a neural index for the neural representations of documents, and propose two hybrid search schemes based on both neural and symbolic indices, namely the parallel search scheme and the sequential search scheme. Our experiments show that both hybrid index and search schemes can improve the recall of the initial retrieval stage with small overhead.
We implemented and evaluated a two-stage retrieval method for personalized academic search in which the initial search results are re-ranked using an author-topic profile. In academic search tasks, the user's own data can help optimizing the ranking of search results to match the searcher's specific individual needs. The author-topic profile consists of topic-specific terms, stored in a graph. We re-rank the top-1000 retrieved documents using ten features that represent the similarity between the document and the author-topic graph. We found that the re-ranking gives a small but significant improvement over the reproduced best method from the literature. Storing the profile as a graph has a number of advantages: it is flexible with respect to node and relation types; it is a visualization of knowledge that is interpretable by the user, and it offers the possibility to view relational characteristics of individual nodes.
Contextualisation has proven to be effective in tailoring \linebreak search results towards the users' information need. While this is true for a basic query search, the usage of contextual session information during exploratory search especially on the level of browsing has so far been underexposed in research. In this paper, we present two approaches that contextualise browsing on the level of structured metadata in a Digital Library (DL), (1) one variant bases on document similarity and (2) one variant utilises implicit session information, such as queries and different document metadata encountered during the session of a users. We evaluate our approaches in a living lab environment using a DL in the social sciences and compare our contextualisation approaches against a non-contextualised approach. For a period of more than three months we analysed 47,444 unique retrieval sessions that contain search activities on the level of browsing. Our results show that a contextualisation of browsing significantly outperforms our baseline in terms of the position of the first clicked item in the result set. The mean rank of the first clicked document (measured as mean first relevant - MFR) was 4.52 using a non-contextualised ranking compared to 3.04 when re-ranking the result lists based on similarity to the previously viewed document. Furthermore, we observed that both contextual approaches show a noticeably higher click-through rate. A contextualisation based on document similarity leads to almost twice as many document views compared to the non-contextualised ranking.
Influenza-like illness (ILI) estimation from web search data is an important web analytics task. The basic idea is to use the frequencies of queries in web search logs that are correlated with past ILI activity as features when estimating current ILI activity. It has been noted that since influenza is seasonal, this approach can lead to spurious correlations with features/queries that also exhibit seasonality, but have no relationship with ILI. Spurious correlations can, in turn, degrade performance. To address this issue, we propose modeling the seasonal variation in ILI activity and selecting queries that are correlated with the residual of the seasonal model and the observed ILI signal. Experimental results show that re-ranking queries obtained by Google Correlate based on their correlation with the residual strongly favours ILI-related queries.
Currently, most people use PubMed to search the MEDLINE database, an important bibliographical information source for life science and biomedical information. However, PubMed has some drawbacks that make it difficult to find relevant publications pertaining to users' individual intentions, especially for non-expert users. To ameliorate the disadvantages of PubMed, we developed G-Bean, a graph based biomedical search engine, to search biomedical articles in MEDLINE database more efficiently.G-Bean addresses PubMed's limitations with three innovations: parallel document index creation,ontology-graph based query expansion, and retrieval and re-ranking of documents based on user's search intention.Performance evaluation with 106 OHSUMED benchmark queries shows that G-Bean returns more relevant results than PubMed does when using these queries to search the MEDLINE database. PubMed could not even return any search result for some OHSUMED queries because it failed to form the appropriate Boolean query statement automatically from the natural language query strings. G-Bean is available at http://bioinformatics.clemson.edu/G-Bean/index.php.G-Bean addresses PubMed's limitations with ontology-graph based query expansion, automatic document indexing, and user search intention discovery. It shows significant advantages in finding relevant articles from the MEDLINE database to meet the information need of the user.
Rank fusion is a powerful technique that allows multiple sources of information to be combined into a single result set. However, to date fusion has not been regarded as being cost-effective in cases where strict per-query efficiency guarantees are required, such as in web search. In this work we propose a novel solution to rank fusion by splitting the computation into two parts -- one phase that is carried out offline to generate pre-computed centroid answers for queries with broadly similar information needs, and then a second online phase that uses the corresponding topic centroid to compute a result page for each query. We explore efficiency improvements to classic fusion algorithms whose costs can be amortized as a pre-processing step, and can then be combined with re-ranking approaches to dramatically improve effectiveness in multi-stage retrieval systems with little efficiency overhead at query time. Experimental results using the ClueWeb12B collection and the UQV100 query variations demonstrate that centroid-based approaches allow improved retrieval effectiveness at little or no loss in query throughput or latency, and with reasonable pre-processing requirements. We additionally show that queries that do not match any of the pre-computed clusters can be accurately identified and efficiently processed in our proposed ranking pipeline.
InfoSeeking Lab's FATE (Fairness Accountability Transparency Ethics) group at University of Washington participated in 2020 TREC Fairness Ranking Track. This report describes that track, assigned data and tasks, our group definitions, and our results. Our approach to bringing fairness in retrieval and re-ranking tasks with Semantic Scholar data was to extract various dimensions of author identity. These dimensions included gender and location. We developed modules for these extractions in a way that allowed us to plug them in for either of the tasks as needed. After trying different combinations of relative weights assigned to relevance, gender, and location information, we chose five runs for retrieval and five runs for re-ranking tasks. The results showed that our runs performed below par for re-ranking task, but above average for retrieval.
This paper describes Tilde's submission to the WMT2020 shared task on news translation for both directions of the English-Polish language pair in both the constrained and the unconstrained tracks. We follow our submissions from the previous years and build our baseline systems to be morphologically motivated sub-word unit-based Transformer base models that we train using the Marian machine translation toolkit. Additionally, we experiment with different parallel and monolingual data selection schemes, as well as sampled back-translation. Our final models are ensembles of Transformer base and Transformer big models that feature right-to-left re-ranking.
Our objective is to introduce to the NLP community an existing k-NN search library NMSLIB, a new retrieval toolkit FlexNeuART, as well as their integration capabilities. NMSLIB, while being one the fastest k-NN search libraries, is quite generic and supports a variety of distance/similarity functions. Because the library relies on the distance-based structure-agnostic algorithms, it can be further extended by adding new distances. FlexNeuART is a modular, extendible and flexible toolkit for candidate generation in IR and QA applications, which supports mixing of classic and neural ranking signals. FlexNeuART can efficiently retrieve mixed dense and sparse representations (with weights learned from training data), which is achieved by extending NMSLIB. In that, other retrieval systems work with purely sparse representations (e.g., Lucene), purely dense representations (e.g., FAISS and Annoy), or only perform mixing at the re-ranking stage.
State-of-the-art Machine Reading Comprehension (MRC) models for Open-domain Question Answering (QA) are typically trained for span selection using distantly supervised positive examples and heuristically retrieved negative examples. This training scheme possibly explains empirical observations that these models achieve a high recall amongst their top few predictions, but a low overall accuracy, motivating the need for answer re-ranking. We develop a simple and effective re-ranking approach (RECONSIDER) for span-extraction tasks, that improves upon the performance of large pre-trained MRC models. RECONSIDER is trained on positive and negative examples extracted from high confidence predictions of MRC models, and uses in-passage span annotations to perform span-focused re-ranking over a smaller candidate set. As a result, RECONSIDER learns to eliminate close false positive passages, and achieves a new state of the art on four QA tasks, including 45.5% Exact Match accuracy on Natural Questions with real user questions, and 61.7% on TriviaQA.
Many recent approaches towards neural information retrieval mitigate their computational costs by using a multi-stage ranking pipeline. In the first stage, a number of potentially relevant candidates are retrieved using an efficient retrieval model such as BM25. Although BM25 has proven decent performance as a first-stage ranker, it tends to miss relevant passages. In this context we propose CoRT, a simple neural first-stage ranking model that leverages contextual representations from pretrained language models such as BERT to complement term-based ranking functions while causing no significant delay at query time. Using the MS MARCO dataset, we show that CoRT significantly increases the candidate recall by complementing BM25 with missing candidates. Consequently, we find subsequent re-rankers achieve superior results with less candidates. We further demonstrate that passage retrieval using CoRT can be realized with surprisingly low latencies.
This paper describes DiDi AI Labs' submission to the WMT2020 news translation shared task. We participate in the translation direction of Chinese->English. In this direction, we use the Transformer as our baseline model, and integrate several techniques for model enhancement, including data filtering, data selection, back-translation, fine-tuning, model ensembling, and re-ranking. As a result, our submission achieves a BLEU score of $36.6$ in Chinese->English.
Holistic person re-identification (Re-ID) and partial person re-identification have achieved great progress respectively in recent years. However, scenarios in reality often include both holistic and partial pedestrian images, which makes single holistic or partial person Re-ID hard to work. In this paper, we propose a robust coarse granularity part-level person Re-ID network (CGPN), which not only extracts robust regional level body features, but also integrates supervised global features for both holistic and partial person images. CGPN gains two-fold benefit toward higher accuracy for person Re-ID. On one hand, CGPN learns to extract effective body part features for both holistic and partial person images. On the other hand, compared with extracting global features directly by backbone network, CGPN learns to extract more accurate global features with a supervision strategy. The single model trained on three Re-ID datasets including Market-1501, DukeMTMC-reID and CUHK03 achieves state-of-the-art performances and outperforms any existing approaches. Especially on CUHK03, which is the most challenging dataset for person Re-ID, in single query mode, we obtain a top result of Rank-1/mAP=87.1\%/83.6\% with this method without re-ranking, outperforming the current best method by +7.0\%/+6.7\%.
With worldwide concerns surrounding the Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2), there is a rapidly growing body of scientific literature on the virus. Clinicians, researchers, and policy-makers need to be able to search these articles effectively. In this work, we present a zero-shot ranking algorithm that adapts to COVID-related scientific literature. Our approach filters training data from another collection down to medical-related queries, uses a neural re-ranking model pre-trained on scientific text (SciBERT), and filters the target document collection. This approach ranks top among zero-shot methods on the TREC COVID Round 1 leaderboard, and exhibits a P@5 of 0.80 and an nDCG@10 of 0.68 when evaluated on both Round 1 and 2 judgments. Despite not relying on TREC-COVID data, our method outperforms models that do. As one of the first search methods to thoroughly evaluate COVID-19 search, we hope that this serves as a strong baseline and helps in the global crisis.