Training a linear classifier or lightweight model on top of pretrained vision model outputs, socalled ‘frozen features’, leads to impressive performance on a number of downstream few-shot tasks. Currently, frozen features are not modified during training. On the other hand, when networks are trained directly on images, data augmentation is a standard recipe that improves performance with no substantial overhead. In this paper, we conduct an extensive pilot study on few-shot image classification that explores applying data augmentations in the frozen feature space, dubbed 'frozen feature augmentation (FroFA)‘, cov-ering twenty augmentations in total. Our study demonstrates that adopting a deceptively simple pointwise FroFA, such as brightness, can improve few-shot performance consistently across three network architectures, three large pre-training datasets, and eight transfer datasets.

We explore the boundaries of scaling up a multilingual vision and language model, both in terms of size of the components and the breadth of its training task mixture. Our model achieves new levels of performance on a wide-range of varied and complex tasks, including multiple image-based captioning and question-answering tasks, image-based document understanding and few-shot (in-context) learning, as well as object detection, video question answering, and video captioning. Our model advances the state-of-the-art on most vision-and-language benchmarks considered (20+ of them). Finally, we observe emerging capabilities, such as complex counting and multilingual object detection, tasks that are not explicitly in the training mix.

The most performant spatio-temporal action localisation models use external person proposals and complex external memory banks. We propose a fully end-to-end, purely-transformer based model that directly ingests an input video, and outputs tubelets - a sequence of bounding boxes and the action classes at each frame. Our flexible model can be trained with either sparse bounding-box supervision on individual frames, or full tubelet annotations. And in both cases, it predicts coherent tubelets as the output. Moreover, our end-to-end model requires no additional pre-processing in the form of proposals, or post-processing in terms of non-maximal suppression. We perform extensive ablation experiments, and significantly advance the state-of-the-art on five different spatio-temporal action localisation benchmarks with both sparse keyframes and full tubelet annotations.

We study the fractal structure of language, aiming to provide a precise formalism for quantifying properties that may have been previously suspected but not formally shown. We establish that language is: (1) self-similar, exhibiting complexities at all levels of granularity, with no particular characteristic context length, and (2) long-range dependent (LRD), with a Hurst parameter of approximately 0.7. Based on these findings, we argue that short-term patterns/dependencies in language, such as in paragraphs, mirror the patterns/dependencies over larger scopes, like entire documents. This may shed some light on how next-token prediction can capture the structure of text across multiple levels of granularity, from words and clauses to broader contexts and intents. In addition, we carry out an extensive analysis across different domains and architectures, showing that fractal parameters are robust. Finally, we demonstrate that the tiny variations in fractal parameters seen across LLMs improve upon perplexity-based bits-per-byte (BPB) in predicting their downstream performance. We hope these findings offer a fresh perspective on language and the mechanisms underlying the success of LLMs.

When learning to rank from user interactions, search and recommender systems must address biases in user behavior to provide a high-quality ranking. One type of bias that has recently been studied in the ranking literature is when sensitive attributes, such as gender, have an impact on a user's judgment about an item's utility. For example, in a search for an expertise area, some users may be biased towards clicking on male candidates over female candidates. We call this type of bias group membership bias.

Finetuning language models on a collection of datasets phrased as instructions has been shown to improve model performance and generalization to unseen tasks. In this paper we explore instruction finetuning with a particular focus on (1) scaling the number of tasks, (2) scaling the model size, and (3) finetuning on chain-of-thought data. We find that instruction finetuning with the above aspects dramatically improves performance on a variety of model classes (PaLM, T5, U-PaLM), prompting setups (zero-shot, few-shot, CoT), and evaluation benchmarks (MMLU, BBH, TyDiQA, MGSM, open-ended generation, RealToxicityPrompts). For instance, Flan-PaLM 540B instruction-finetuned on 1.8K tasks outperforms PaLM 540B by a large margin (+9.4% on average). Flan-PaLM 540B achieves state-of-the-art performance on several benchmarks (at time of release), such as 75.2% on five-shot MMLU. We also publicly release Flan-T5 checkpoints,1 which achieve strong few-shot performance even compared to much larger models, such as PaLM 62B. Overall, instruction finetuning is a general method for improving the performance and usability of pretrained language models.

Modern transformer-based information retrieval models achieve state-of-the-art performance across various benchmarks. The self-attention of the transformer models is a powerful mechanism to contextualize terms over the whole input but quickly becomes prohibitively expensive for long input as required in document retrieval. Instead of focusing on the model itself to improve efficiency, this paper explores different bag of words document representations that encode full documents by only a fraction of their characteristic terms, allowing us to control and reduce the input length. We experiment with various models for document retrieval on MS MARCO data, as well as zero-shot document retrieval on Robust04, and show large gains in efficiency while retaining reasonable effectiveness. Inference time efficiency gains are both lowering the time and memory complexity in a controllable way, allowing for further trading off memory footprint and query latency. More generally, this line of research connects traditional IR models with neural “NLP” models and offers novel ways to explore the space between (efficient, but less effective) traditional rankers and (effective, but less efficient) neural rankers elegantly.

Sanket Mehta, Jai Gupta, Yi Tay, Mostafa Dehghani, Vinh Tran, Jinfeng Rao, Marc Najork, Emma Strubell, Donald Metzler. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. 2023.

Yi Tay, Jason Wei, Hyung Chung, Vinh Tran, David So, Siamak Shakeri, Xavier Garcia, Steven Zheng, Jinfeng Rao, Aakanksha Chowdhery, Denny Zhou, Donald Metzler, Slav Petrov, Neil Houlsby, Quoc Le, Mostafa Dehghani. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. 2023.

The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image and video modelling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters (Chen et al., 2022). We present a recipe for highly efficient and stable training of a 22B-parameter ViT (ViT-22B) and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features), ViT-22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between fairness and performance, state-of-the-art alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT-22B demonstrates the potential for "LLM-like" scaling in vision, and provides key steps towards getting there.

Humans have the ability to adapt the type of information they use, the procedure they employ, and the amount of time they spend when solving problems. However, most standard neural networks have a fixed function type and computation budget regardless of the sample’s nature or difficulty. Adaptivity is a powerful paradigm as it not only imbues practitioners with flexibility pertaining to the downstream usage of these models but can also serve as a powerful inductive bias for solving certain challenging classes of problems. In this work, we introduce a new approach called AdaTape, which allows for dynamic computation in neural networks through adaptive tape tokens. AdaTape utilizes an elastic input sequence by equipping an architecture with a dynamic read-and-write tape. Specifically, we adaptively generate input sequences using tape tokens obtained from a tape bank which can be either trainable or derived from input data. We examine the challenges and requirements to obtain dynamic sequence content and length, and propose the Adaptive Tape Reading (ATR) algorithm to achieve both goals. Through extensive experiments on image recognition tasks, we show that AdaTape can achieve better performance while maintaining the computational cost. To facilitate further research, we have released code at https://github.com/google-research/scenic/tree/main/scenic/projects/adatape.

The ubiquitous and demonstrably suboptimal choice of resizing images to a fixed resolution before processing them with computer vision models has not yet been successfully challenged. However, models such as the Vision Transformer (ViT) offer flexible sequence-based modeling, and hence varying input sequence lengths. We take advantage of this with NaViT (Native Resolution ViT) which uses sequence packing during training to process inputs of arbitrary resolutions and aspect ratios. Alongside flexible model usage, we demonstrate improved training efficiency for large-scale supervised and contrastive image-text pretraining. NaViT can be efficiently transferred to standard tasks such as image and video classification, object detection, and semantic segmentation and leads to improved results on robustness and fairness benchmarks. At inference time, the input resolution flexibility can be used to smoothly navigate the test-time cost-performance trade-off. We believe that NaViTmarks a departure from the standard, CNN-designed, input and modelling pipeline used by most computer vision models, and represents a promising direction for ViTs.

A common way to avoid overfitting in supervised learning is early stopping, where a held-out set is used for iterative evaluation during training to find a sweet spot in the number of training steps that gives maximum generalization. However, such a method requires a disjoint validation set, thus part of the labeled data from the training set is usually left out for this purpose, which is not ideal when training data is scarce. Furthermore, when the training labels are noisy, the performance of the model over a validation set may not be an accurate proxy for generalization. In this paper, we propose a method to spot an early stopping point in the training iterations of an overparameterized (NN) without the need for a validation set. We first show that in the overparameterized regime the randomly initialized weights of a linear model converge to the same direction during training. Using this result, we propose to train two parallel instances of a linear model, initialized with different random seeds, and use their intersection as a signal to detect overfitting. In order to detect intersection, we use the cosine distance between the weights of the parallel models during training iterations. Noticing that the final layer of a NN is a linear map of pre-last layer activations to output logits, we build on our criterion for linear models and propose an extension to multi-layer networks, using the new notion of counterfactual weights. We conduct experiments on two areas that early stopping has noticeable impact on preventing overfitting of a NN: (i) learning from noisy labels; and (ii) learning to rank in information retrieval. Our experiments on four widely used datasets confirm the effectiveness of our method for generalization. For a wide range of learning rates, our method, called Cosine-Distance Criterion (CDC), leads to better generalization on average than all the methods that we compare against in almost all of the tested cases.

Scenic is an open-source11https://github.com/google-research/scenic JAX library with a focus on transformer-based models for computer vision research and beyond. The goal of this toolkit is to facilitate rapid experimentation, prototyping, and research of new architectures and models. Scenic supports a diverse range of tasks (e.g., classification, segmentation, detection) and facilitates working on multi-modal problems, along with GPU/TPU support for large-scale, multi-host and multi-device training. Scenic also offers optimized implementations of state-of-the-art research models spanning a wide range of modalities. Scenic has been successfully used for numerous projects and published papers and continues serving as the library of choice for rapid prototyping and publication of new research ideas.

Recent advances in Transformer-based large language models (LLMs) have led to significant performance improvements across many tasks. These gains come with a drastic increase in the models' size, potentially leading to slow and costly use at inference time. In practice, however, the series of generations made by LLMs is composed of varying levels of difficulty. While certain predictions truly benefit from the models' full capacity, other continuations are more trivial and can be solved with reduced compute. In this work, we introduce Confident Adaptive Language Modeling (CALM), a framework for dynamically allocating different amounts of compute per input and generation timestep. Early exit decoding involves several challenges that we address here, such as: (1) what confidence measure to use; (2) connecting sequence-level constraints to local per-token exit decisions; and (3) attending back to missing hidden representations due to early exits in previous tokens. Through theoretical analysis and empirical experiments on three diverse text generation tasks, we demonstrate the efficacy of our framework in reducing compute---potential speedup of up to $\times 3$---while provably maintaining high performance.

In this paper, we demonstrate that information retrieval can be accomplished with a single Transformer, in which all information about the corpus is encoded in the parameters of the model. To this end, we introduce the Differentiable Search Index (DSI), a new paradigm that learns a text-to-text model that maps string queries directly to relevant docids; in other words, a DSI model answers queries directly using only its parameters, dramatically simplifying the whole retrieval process. We study variations in how documents and their identifiers are represented, variations in training procedures, and the interplay between models and corpus sizes. Experiments demonstrate that given appropriate design choices, DSI significantly outperforms strong baselines such as dual encoder models. Moreover, DSI demonstrates strong generalization capabilities, outperforming a BM25 baseline in a zero-shot setup.

Although information access systems have long supportedpeople in accomplishing a wide range of tasks, we propose broadening the scope of users of information access systems to include task-driven machines, such as machine learning models. In this way, the core principles of indexing, representation, retrieval, and ranking can be applied and extended to substantially improve model generalization, scalability, robustness, and interpretability. We describe a generic retrieval-enhanced machine learning (REML) framework, which includes a number of existing models as special cases. REML challenges information retrieval conventions, presenting opportunities for novel advances in core areas, including optimization. The REML research agenda lays a foundation for a new style of information access research and paves a path towards advancing machine learning and artificial intelligence.

Rabeeh Karimi Mahabadi, Sebastian Ruder, Mostafa Dehghani, James Henderson. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

Yi Tay, Mostafa Dehghani, Jai Prakash Gupta, Vamsi Aribandi, Dara Bahri, Zhen Qin, Donald Metzler. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

We present pure-transformer based models for video classification, drawing upon the recent success of such models in image classification. Our model extracts spatiotemporal tokens from the input video, which are then encoded by a series of transformer layers. In order to handle the long sequences of tokens encountered in video, we propose several, efficient variants of our model which factorise the spatial- and temporal-dimensions of the input. Although transformer-based models are known to only be effective when large training datasets are available, we show how we can effectively regularise the model during training and leverage pretrained image models to be able to train on comparatively small datasets. We conduct thorough ablation studies, and achieve state-of-the-art results on multiple video classification benchmarks including Kinetics 400 and 600, Epic Kitchens, Something-Something v2 and Moments in Time, outperforming prior methods based on deep 3D convolutional networks.

This paper proposes Omnidirectional Representations from Transformers (OMNINET). In OmniNet, instead of maintaining a strictly horizon-tal receptive field, each token is allowed to attend to all tokens in the entire network. This process can also be interpreted as a form of extreme or intensive attention mechanism that has the receptive field of the entire width and depth of the network. To this end, the omnidirectional attention is learned via a meta-learner, which is essentially another self-attention based model. In order to mitigate the computationally expensive costs of full receptive field attention, we leverage efficient self-attention models such as kernel-based, low-rank attention and/or Big Bird as the meta-learner. Extensive experiments are conducted on autoregressive language modeling(LM1B, C4), Machine Translation, Long Range Arena (LRA), and Image Recognition.The experiments show that OmniNet achieves considerable improvements across these tasks, including achieving state-of-the-art performance on LM1B,WMT’14 En-De/En-Fr, and Long Range Arena.Moreover, using omnidirectional representation in Vision Transformers leads to significant improvements on image recognition tasks on both few-shot learning and fine-tuning setups.

In this paper, we introduce a novel visual representation learning which relies on a handful of adaptively learned tokens, and which is applicable to both image and video understanding tasks. Instead of relying on hand-designed splitting strategies to obtain visual tokens and processing a large number of densely sampled patches for attention, our approach learns to mine important tokens in visual data. This results in efficiently and effectively finding a few important visual tokens and enables modeling of pairwise attention between such tokens, over a longer temporal horizon for videos, or the spatial content in image frames. Our experiments demonstrate strong performance on several challenging benchmarks for video recognition tasks. Importantly, due to our tokens being adaptive, we accomplish competitive results at significantly reduced computational cost. We establish new state-of-the-arts on multiple video datasets, including Kinetics-400, Kinetics-600, Charades, and AViD.

Abstract Efficiently exploiting all sources of information such as labeled instances, classes’ representation, and relations of them has a high impact on the performance of Multi-Label Text Classification (MLTC) systems. Most of the current approaches use labeled documents as the primary source of information for MLTC. We investigate the effectiveness of different sources of information— such as the labeled training data, textual labels of classes, and taxonomy relations of classes— for MLTC. More specifically, first, for each document–class pair, different features are extracted using different sources of information. The features reflect the similarity of classes and documents. Then, MLTC is considered to be a ranking problem, and a learning to rank (LTR) approach is used for ranking classes regarding documents and selecting labels of documents. An important characteristic of many MLTC instances is that documents can belong to multiple classes and there are implicit relations between classes. We apply score propagation on top of LTR to incorporate co-occurrence patterns of classes in labeled documents. Our main findings are the following. First, using an LTR approach integrating all features, we observe significantly better performance than previous systems for MLTC. Specifically, we show that simple classification approaches fail when there is a high number of classes. Second, the analysis of feature weights reveals the relative importance of various sources of evidence, also giving insight into the underlying classification problem. Interestingly, the results indicate that the titles of documents are more informative than all other sources of information. Third, a lean-and-mean system using only four features is able to perform at 96% of the large LTR model that we propose in this paper. Fourth, using the co-occurrence information of classes helps in classifying documents more accurately. Our results show that the co-occurrence information is more helpful when the underlying classifier has a poor performance.

Users seek direct answers to complex questions from large open-domain knowledge sources like the Web. Open-domain question answering has become a critical task to be solved for building systems that help address users' complex information needs. Most open-domain question answering systems use a search engine to retrieve a set of candidate documents, select one or a few of them as context, and then apply reading comprehension models to extract answers. Some questions, however, require taking a broader context into account, e.g., by considering low-ranked documents that are not immediately relevant, combining information from multiple documents, and reasoning over multiple facts from these documents to infer the answer. In this paper, we propose a model based on the Transformer architecture that is able to efficiently operate over a larger set of candidate documents by effectively combining the evidence from these documents during multiple steps of reasoning, while it is robust against noise from low-ranked non-relevant documents included in the set. We use our proposed model, called TraCRNet, on two public open-domain question answering datasets, SearchQA and Quasar-T, and achieve results that meet or exceed the state-of-the-art.

The availability of massive data and computing power allowing for effective data driven neural approaches is having a major impact on machine learning and information retrieval research, but these models have a basic problem with efficiency. Current neural ranking models are implemented as multistage rankers: for efficiency reasons, the neural model only re-ranks the top ranked documents retrieved by a first-stage efficient ranker in response to a given query. Neural ranking models learn dense representations causing essentially every query term to match every document term, making it highly inefficient or intractable to rank the whole collection. The reliance on a first stage ranker creates a dual problem: First, the interaction and combination effects are not well understood. Second, the first stage ranker serves as a "gate-keeper" or filter, effectively blocking the potential of neural models to uncover new relevant documents. In this work, we propose a standalone neural ranking model (SNRM) by introducing a sparsity property to learn a latent sparse representation for each query and document. This representation captures the semantic relationship between the query and documents, but is also sparse enough to enable constructing an inverted index for the whole collection. We parameterize the sparsity of the model to yield a retrieval model as efficient as conventional term based models. Our model gains in efficiency without loss of effectiveness: it not only outperforms the existing term matching baselines, but also performs similarly to the recent re-ranking based neural models with dense representations. Our model can also take advantage of pseudo-relevance feedback for further improvements. More generally, our results demonstrate the importance of sparsity in neural IR models and show that dense representations can be pruned effectively, giving new insights about essential semantic features and their distributions.

In recent years, machine learning approaches, and in particular deep neural networks, have yielded significant improvements on several natural language processing and computer vision tasks; however, such breakthroughs have not yet been observed in the area of information retrieval. Besides the complexity of IR tasks, such as understanding the user's information needs, a main reason is the lack of high-quality and/or large-scale training data for many IR tasks. This necessitates studying how to design and train machine learning algorithms where there is no large-scale or high-quality data in hand. Therefore, considering the quick progress in development of machine learning models, this is an ideal time for a workshop that especially focuses on learning in such an important and challenging setting for IR tasks. The goal of this workshop is to bring together researchers from industry---where data is plentiful but noisy---with researchers from academia---where data is sparse but clean to discuss solutions to these related problems.

Abstract The expertise of human experts can be formally extracted from their written documents, research projects, and everyday activities. The process whereby experts are recognized according to their activities is called expert finding. In this paper, we propose an approach to identify the experts in a given field according to the content of 3 easily accessible sources of information: (a) “Publications,” (b) “Social interactions,” and (c) “Scientometric information.” We employed the Dempster‐Shafer theory to combine the results obtained from individual sources to find a final unified ranking. In Digital Bibliography & Library Project standard data, it is shown that the Dempster‐Shafer combination creates a desired synergy between 2 bodies of knowledge, which improves the precision of the top‐ranked results.

Supporting exploratory search tasks with the help of structured data is an effective way to go beyond keyword search, as it provides an overview of the data, enables users to zoom in on their intent, and provides assistance during their navigation trails. However, finding a good starting point for a search episode in the given structure can still pose a considerable challenge, as users tend to be unfamiliar with exact, complex hierarchical structure. Thus, providing lookahead clues can be of great help and allow users to make better decisions on their search trajectory.

Recently, researchers started to pay attention to the detection of temporal shifts in the meaning of words. However, most (if not all) of these approaches restricted their efforts to uncovering change over time, thus neglecting other valuable dimensions such as social or political variability. We propose an approach for detecting semantic shifts between different viewpoints---broadly defined as a set of texts that share a specific metadata feature, which can be a time-period, but also a social entity such as a political party. For each viewpoint, we learn a semantic space in which each word is represented as a low dimensional neural embedded vector. The challenge is to compare the meaning of a word in one space to its meaning in another space and measure the size of the semantic shifts. We compare the effectiveness of a measure based on optimal transformations between the two spaces with a measure based on the similarity of the neighbors of the word in the respective spaces. Our experiments demonstrate that the combination of these two performs best. We show that the semantic shifts not only occur over time but also along different viewpoints in a short period of time. For evaluation, we demonstrate how this approach captures meaningful semantic shifts and can help improve other tasks such as the contrastive viewpoint summarization and ideology detection (measured as classification accuracy) in political texts. We also show that the two laws of semantic change which were empirically shown to hold for temporal shifts also hold for shifts across viewpoints. These laws state that frequent words are less likely to shift meaning while words with many senses are more likely to do so.

Users try to articulate their complex information needs during search sessions by reformulating their queries. To make this process more effective, search engines provide related queries to help users in specifying the information need in their search process. In this paper, we propose a customized sequence-to-sequence model for session-based query suggestion. In our model, we employ a query-aware attention mechanism to capture the structure of the session context. is enables us to control the scope of the session from which we infer the suggested next query, which helps not only handle the noisy data but also automatically detect session boundaries. Furthermore, we observe that, based on the user query reformulation behavior, within a single session a large portion of query terms is retained from the previously submitted queries and consists of mostly infrequent or unseen terms that are usually not included in the vocabulary. We therefore empower the decoder of our model to access the source words from the session context during decoding by incorporating a copy mechanism. Moreover, we propose evaluation metrics to assess the quality of the generative models for query suggestion. We conduct an extensive set of experiments and analysis. e results suggest that our model outperforms the baselines both in terms of the generating queries and scoring candidate queries for the task of query suggestion.

Despite the impressive improvements achieved by unsupervised deep neural networks in computer vision and NLP tasks, such improvements have not yet been observed in ranking for information retrieval. The reason may be the complexity of the ranking problem, as it is not obvious how to learn from queries and documents when no supervised signal is available. Hence, in this paper, we propose to train a neural ranking model using weak supervision, where labels are obtained automatically without human annotators or any external resources (e.g., click data). To this aim, we use the output of an unsupervised ranking model, such as BM25, as a weak supervision signal. We further train a set of simple yet effective ranking models based on feed-forward neural networks. We study their effectiveness under various learning scenarios (point-wise and pair-wise models) and using different input representations (i.e., from encoding query-document pairs into dense/sparse vectors to using word embedding representation). We train our networks using tens of millions of training instances and evaluate it on two standard collections: a homogeneous news collection (Robust) and a heterogeneous large-scale web collection (ClueWeb). Our experiments indicate that employing proper objective functions and letting the networks to learn the input representation based on weakly supervised data leads to impressive performance, with over 13% and 35% MAP improvements over the BM25 model on the Robust and the ClueWeb collections. Our findings also suggest that supervised neural ranking models can greatly benefit from pre-training on large amounts of weakly labeled data that can be easily obtained from unsupervised IR models.

Machine learning plays a role in many aspects of modern IR systems, and deep learning is applied in all of them. The fast pace of modern-day research has given rise to many approaches to many IR problems. The amount of information available can be overwhelming both for junior students and for experienced researchers looking for new research topics and directions. The aim of this full- day tutorial is to give a clear overview of current tried-and-trusted neural methods in IR and how they benefit IR.

There is an ongoing debate on personalization, adapting results to the unique user exploiting a user's personal history, versus customization, adapting results to a group profile sharing one or more characteristics with the user at hand. Personal profiles are often sparse, due to cold start problems and the fact that users typically search for new items or information, necessitating to back-off to customization, but group profiles often suffer from accidental features brought in by the unique individual contributing to the group. In this paper we propose a generalized group profiling approach that teases apart the exact contribution of the individual user level and the `abstract' group level by extracting a latent model that captures all, and only, the essential features of the whole group. Our main findings are the followings.

The use of feedback information is an effective approach to address the vocabulary gap between a user's query and the relevant documents. It has been shown that some relevant documents act like "poison pills," i.e. they hurt the performance of feedback systems despite the fact that they are relevant. In this paper, we study the positive counterpart of this by investigating the helpfulness of nonrelevant documents in feedback. In general, we find that although documents that are explicitly judged as non-relevant are normally assumed to be poisonous for feedback systems, sometimes considering high-scored non-relevant documents as a positive feedback helps to improve the performance of retrieval. In our experimental data, we observe a considerable fraction of non-relevant documents in higher ranked positions of the initial retrieval run, for most of the topics. Hence, by ignoring the potential value of non-relevant documents, we may loose a lot of useful information. We investigate the potential contribution of non-relevant documents using existing state-of-the-art feedback methods. Our main findings are the following. First, we find that some of the nonrelevant documents are exclusively helpful, they improve retrieval on their own, and others are complementary helpful, they lead to further improvement when added to a set of relevant documents. Second, we discover that, on average, exclusively helpful non-relevant documents have a higher contribution to the performance improvement, compared to the complementary ones. Third, we show that non-relevant documents in topics with poor average precision in the initial retrieval are more likely to help in the feedback.

Users tend to articulate their complex information needs in only a few keywords, making underspecified statements of request the main bottleneck for retrieval effectiveness. Taking advantage of feedback information is one of the best ways to enrich the query representation, but can also lead to loss of query focus and harm performance in particular when the initial query retrieves only little relevant information when overfitting to accidental features of the particular observed feedback documents. Inspired by the early work of Luhn [23], we propose significant words language models of feedback documents that capture all, and only, the significant shared terms from feedback documents. We adjust the weights of common terms that are already well explained by the document collection as well as the weight of rare terms that are only explained by specific feedback documents, which eventually results in having only the significant terms left in the feedback model.

Transforming the data into a suitable representation is the first key step of data analysis, and the performance of any data oriented method is heavily depending on it. We study questions on how we can best learn representations for textual entities that are: 1) precise, 2) robust against noisy terms, 3) transferable over time, and 4) interpretable by human inspection. Inspired by the early work of Luhn, we propose significant words language models of a set of documents that capture all, and only, the significant shared terms from them. We adjust the weights of common terms that are already well explained by the document collection as well as the weight of incidental rare terms that are only explained by specific documents, which eventually results in having only the significant terms left in the model.

Abstract Comparable corpora are key translation resources for both languages and domains with limited linguistic resources. The existing approaches for building comparable corpora are mostly based on ranking candidate documents in the target language for each source document using a cross-lingual retrieval model. These approaches also exploit other evidence of document similarity, such as proper names and publication dates, to build more reliable alignments. However, the importance of each evidence in the scores of candidate target documents is determined heuristically. In this paper, we employ a learning to rank method for ranking candidate target documents with respect to each source document. The ranking model is constructed by defining each evidence for similarity of bilingual documents as a feature whose weight is learned automatically. Learning feature weights can significantly improve the quality of alignments, because the reliability of features depends on the characteristics of both source and target languages of a comparable corpus. We also propose a method to generate appropriate training data for the task of building comparable corpora. We employed the proposed learning-based approach to build a multi-domain English–Persian comparable corpus which covers twelve different domains obtained from Open Directory Project. Experimental results show that the created alignments have high degrees of comparability. Comparison with existing approaches for building comparable corpora shows that our learning-based approach improves both quality and coverage of alignments.

Identifying authors of short texts on Internet or social media based communication systems is an important tool against fraud and cybercrimes. Besides the challenges raised by the limited length of these short messages, evolving language and writing styles of authors of these texts makes authorship attribution difficult. Most current short text authorship attribution approaches only address the challenge of limited text length. However, neglecting the second challenge may lead to poor performance of authorship attribution for authors who change their writing styles.

Email is a type of Web data which is produced in enormous quantities. It is beneficial to detect conversation threads contained in the email corpora for various applications, including discussion search, expert finding and even email clustering and classification. Conversation thread in email corpora can be defined as a cluster of exchanged emails among the same group of people by reply or forwarding on the same topic. According to this definition, we can define parent-child relation between emails, so email conversation threads seem to demonstrate tree structure. This paper presents a new approach based on genetic programming for reconstruction of conversation threads in emails data. This approach considers finding email conversation threads as an optimization problem, and exploits genetic programming to search intelligently in the space of possible solutions. Rather than several studies that have been conducted on this problem, this work concentrates on detecting accurate structure of conversation threads in high recall. This paper provides a comprehensive evaluation on the BC3 data set. Preliminary results suggest that our method provides acceptable precision and higher recall than existing methods.