We propose an extensible deep learning method that uses reinforcement learning to train neural networks for offline ranking in information retrieval (IR). We call our method BanditRank as it treats ranking as a contextual bandit problem. In the domain of learning to rank for IR, current deep learning models are trained on objective functions different from the measures they are evaluated on. Since most evaluation measures are discrete quantities, they cannot be leveraged by directly using gradient descent algorithms without an approximation. BanditRank bridges this gap by directly optimizing a task-specific measure, such as mean average precision (MAP), using gradient descent. Specifically, a contextual bandit whose action is to rank input documents is trained using a policy gradient algorithm to directly maximize the reward. The reward can be a single measure, such as MAP, or a combination of several measures. The notion of ranking is also inherent in BanditRank, similar to the current \textit{listwise} approaches. To evaluate the effectiveness of BanditRank, we conducted a series of experiments on datasets related to three different tasks, i.e., web search, community, and factoid question answering. We found that it performs better than state-of-the-art methods when applied on the question answering datasets. On the web search dataset, we found that BanditRank performed better than four strong listwise baselines including LambdaMART, AdaRank, ListNet and Coordinate Ascent.
How to leverage cross-document interactions to improve ranking performance is an important topic in information retrieval (IR) research. However, this topic has not been well-studied in the learning-to-rank setting and most of the existing work still treats each document independently while scoring. The recent development of deep learning shows strength in modeling complex relationships across sequences and sets. It thus motivates us to study how to leverage cross-document interactions for learning-to-rank in the deep learning framework. In this paper, we formally define the permutation-equivariance requirement for a scoring function that captures cross-document interactions. We then propose a self-attention based document interaction network and show that it satisfies the permutation-equivariant requirement, and can generate scores for document sets of varying sizes. Our proposed methods can automatically learn to capture document interactions without any auxiliary information, and can scale across large document sets. We conduct experiments on three ranking datasets: the benchmark Web30k, a Gmail search, and a Google Drive Quick Access dataset. Experimental results show that our proposed methods are both more effective and efficient than baselines.
We present a systematic investigation of layer-wise BERT activations for general-purpose text representations to understand what linguistic information they capture and how transferable they are across different tasks. Sentence-level embeddings are evaluated against two state-of-the-art models on downstream and probing tasks from SentEval, while passage-level embeddings are evaluated on four question-answering (QA) datasets under a learning-to-rank problem setting. Embeddings from the pre-trained BERT model perform poorly in semantic similarity and sentence surface information probing tasks. Fine-tuning BERT on natural language inference data greatly improves the quality of the embeddings. Combining embeddings from different BERT layers can further boost performance. BERT embeddings outperform BM25 baseline significantly on factoid QA datasets at the passage level, but fail to perform better than BM25 on non-factoid datasets. For all QA datasets, there is a gap between embedding-based method and in-domain fine-tuned BERT (we report new state-of-the-art results on two datasets), which suggests deep interactions between question and answer pairs are critical for those hard tasks.
We present a recurrent neural network based system for automatic quality estimation of natural language generation (NLG) outputs, which jointly learns to assign numerical ratings to individual outputs and to provide pairwise rankings of two different outputs. The latter is trained using pairwise hinge loss over scores from two copies of the rating network. We use learning to rank and synthetic data to improve the quality of ratings assigned by our system: we synthesise training pairs of distorted system outputs and train the system to rank the less distorted one higher. This leads to a 12% increase in correlation with human ratings over the previous benchmark. We also establish the state of the art on the dataset of relative rankings from the E2E NLG Challenge (Du\v{s}ek et al., 2019), where synthetic data lead to a 4% accuracy increase over the base model.
The unevenness importance of criminal activities in the onion domains of the Tor Darknet and the different levels of their appeal to the end-user make them tangled to measure their influence. To this end, this paper presents a novel content-based ranking framework to detect the most influential onion domains. Our approach comprises a modeling unit that represents an onion domain using forty features extracted from five different resources: user-visible text, HTML markup, Named Entities, network topology, and visual content. And also, a ranking unit that, using the Learning-to-Rank (LtR) approach, automatically learns a ranking function by integrating the previously obtained features. Using a case-study based on drugs-related onion domains, we obtained the following results. (1) Among the explored LtR schemes, the listwise approach outperforms the benchmarked methods with an NDCG of 0.95 for the top-10 ranked domains. (2) We proved quantitatively that our framework surpasses the link-based ranking techniques. Also, (3) with the selected feature, we observed that the textual content, composed by text, NER, and HTML features, is the most balanced approach, in terms of efficiency and score obtained. The proposed framework might support Law Enforcement Agencies in detecting the most influential domains related to possible suspicious activities.
In this work we describe the system from Natural Language Processing group at Arizona State University for the TextGraphs 2019 Shared Task. The task focuses on Explanation Regeneration, an intermediate step towards general multi-hop inference on large graphs. Our approach consists of modeling the explanation regeneration task as a \textit{learning to rank} problem, for which we use state-of-the-art language models and explore dataset preparation techniques. We utilize an iterative re-ranking based approach to further improve the rankings. Our system secured 2nd rank in the task with a mean average precision (MAP) of 41.3\% on the test set.
When estimating the relevancy between a query and a document, ranking models largely neglect the mutual information among documents. A common wisdom is that if two documents are similar in terms of the same query, they are more likely to have similar relevance score. To mitigate this problem, in this paper, we propose a multi-agent reinforced ranking model, named MarlRank. In particular, by considering each document as an agent, we formulate the ranking process as a multi-agent Markov Decision Process (MDP), where the mutual interactions among documents are incorporated in the ranking process. To compute the ranking list, each document predicts its relevance to a query considering not only its own query-document features but also its similar documents features and actions. By defining reward as a function of NDCG, we can optimize our model directly on the ranking performance measure. Our experimental results on two LETOR benchmark datasets show that our model has significant performance gains over the state-of-art baselines. We also find that the NDCG shows an overall increasing trend along with the step of interactions, which demonstrates that the mutual information among documents helps improve the ranking performance.
List-wise based learning to rank methods are generally supposed to have better performance than point- and pair-wise based. However, in real-world applications, state-of-the-art systems are not from list-wise based camp. In this paper, we propose a new non-linear algorithm in the list-wise based framework called ListMLE, which uses the Plackett-Luce (PL) loss. Our experiments are conducted on the two largest publicly available real-world datasets, Yahoo challenge 2010 and Microsoft 30K. This is the first time in the single model level for a list-wise based system to match or overpass state-of-the-art systems in real-world datasets.
In learning to rank area, industry-level applications have been dominated by gradient boosting framework, which fits a tree using least square error principle. While in classification area, another tree fitting principle, weighted least square error, has been widely used, such as LogitBoost and its variants. However, there is a lack of analysis on the relationship between the two principles in the scenario of learning to rank. We propose a new principle named least objective loss based error that enables us to analyze the issue above as well as several important learning to rank models. We also implement two typical and strong systems and conduct our experiments in two real-world datasets. Experimental results show that our proposed method brings moderate improvements over least square error principle.
We present a pairwise learning to rank approach based on a neural net, called DirectRanker, that generalizes the RankNet architecture. We show mathematically that our model is reflexive, antisymmetric, and transitive allowing for simplified training and improved performance. Experimental results on the LETOR MSLR-WEB10K, MQ2007 and MQ2008 datasets show that our model outperforms numerous state-of-the-art methods, while being inherently simpler in structure and using a pairwise approach only.
We introduce Uncertain Natural Language Inference (UNLI), a refinement of Natural Language Inference (NLI) that shifts away from categorical labels, targeting instead the direct prediction of subjective probability assessments. We demonstrate the feasibility of collecting annotations for UNLI by relabeling a portion of the SNLI dataset under a probabilistic scale, where items even with the same categorical label differ in how likely people judge them to be true given a premise. We describe a direct scalar regression modeling approach, and find that existing categorically labeled NLI data can be used in pre-training. Our best models approach human performance, demonstrating models may be capable of more subtle inferences than the categorical bin assignment employed in current NLI tasks.
Answering multiple-choice questions in a setting in which no supporting documents are explicitly provided continues to stand as a core problem in natural language processing. The contribution of this article is two-fold. First, it describes a method which can be used to semantically rank documents extracted from Wikipedia or similar natural language corpora. Second, we propose a model employing the semantic ranking that holds the first place in two of the most popular leaderboards for answering multiple-choice questions: ARC Easy and Challenge. To achieve this, we introduce a self-attention based neural network that latently learns to rank documents by their importance related to a given question, whilst optimizing the objective of predicting the correct answer. These documents are considered relevant contexts for the underlying question. We have published the ranked documents so that they can be used off-the-shelf to improve downstream decision models.
Today's Internet Services are undergoing fundamental changes and shifting to an intelligent computing era where AI is widely employed to augment services. In this context, many innovative AI algorithms, systems, and architectures are proposed, and thus the importance of benchmarking and evaluating them rises. However, modern Internet services adopt a microservice-based architecture and consist of various modules. The diversity of these modules and complexity of execution paths, the massive scale and complex hierarchy of datacenter infrastructure, the confidential issues of data sets and workloads pose great challenges to benchmarking. In this paper, we present the first industry-standard Internet service AI benchmark suite---AIBench with seventeen industry partners, including several top Internet service providers. AIBench provides a highly extensible, configurable, and flexible benchmark framework that contains loosely coupled modules. We identify sixteen prominent AI problem domains like learning to rank, each of which forms an AI component benchmark, from three most important Internet service domains: search engine, social network, and e-commerce, which is by far the most comprehensive AI benchmarking effort. On the basis of the AIBench framework, abstracting the real-world data sets and workloads from one of the top e-commerce providers, we design and implement the first end-to-end Internet service AI benchmark, which contains the primary modules in the critical paths of an industry scale application and is scalable to deploy on different cluster scales. The specifications, source code, and performance numbers are publicly available from the benchmark council web site http://www.benchcouncil.org/AIBench/index.html.
This paper explores entity embedding effectiveness in ad-hoc entity retrieval, which introduces distributed representation of entities into entity retrieval. The knowledge graph contains lots of knowledge and models entity semantic relations with the well-formed structural representation. Entity embedding learns lots of semantic information from the knowledge graph and represents entities with a low-dimensional representation, which provides an opportunity to establish interactions between query related entities and candidate entities for entity retrieval. Our experiments demonstrate the effectiveness of entity embedding based model, which achieves more than 5\% improvement than the previous state-of-the-art learning to rank based entity retrieval model. Our further analysis reveals that the entity semantic match feature effective, especially for the scenario which needs more semantic understanding.
Road transportation is of critical importance for a nation, having profound effects in the economy, the health and life style of its people. With the growth of cities and populations come bigger demands for mobility and safety, creating new problems and magnifying those of the past. New tools are needed to face the challenge, to keep roads in good conditions, their users safe, and minimize the impact on the environment. This dissertation is concerned with road quality assessment and aggressive driving, two important problems in road transportation, approached in the context of Intelligent Transportation Systems by using Machine Learning techniques to analyze acceleration time series acquired with smartphone-based opportunistic sensing to automatically detect, classify, and characterize events of interest. Two aspects of road quality assessment are addressed: the detection and the characterization of road anomalies. For the first, the most widely cited works in the literature are compared and proposals capable of equal or better performance are presented, removing the reliance on threshold values and reducing the computational cost and dimensionality of previous proposals. For the second, new approaches for the estimation of pothole depth and the functional condition of speed reducers are showed. The new problem of pothole depth ranking is introduced, using a learning-to-rank approach to sort acceleration signals by the depth of the potholes that they reflect. The classification of aggressive driving maneuvers is done with automatic feature extraction, finding characteristically shaped subsequences in the signals as more effective discriminants than conventional descriptors calculated over time windows. Finally, all the previously mentioned tasks are combined to produce a robust road transport evaluation platform.
We study the use of BERT for non-factoid question-answering, focusing on the passage re-ranking task under varying passage lengths. To this end, we explore the fine-tuning of BERT in different learning-to-rank setups, comprising both point-wise and pair-wise methods, resulting in substantial improvements over the state-of-the-art. We then analyze the effectiveness of BERT for different passage lengths and suggest how to cope with large passages.
Link prediction (LP) algorithms propose to each node a ranked list of nodes that are currently non-neighbors, as the most likely candidates for future linkage. Owing to increasing concerns about privacy, users (nodes) may prefer to keep some of their connections protected or private. Motivated by this observation, our goal is to design a differentially private LP algorithm, which trades off between privacy of the protected node-pairs and the link prediction accuracy. More specifically, we first propose a form of differential privacy on graphs, which models the privacy loss only of those node-pairs which are marked as protected. Next, we develop DPLP , a learning to rank algorithm, which applies a monotone transform to base scores from a non-private LP system, and then adds noise. DPLP is trained with a privacy induced ranking loss, which optimizes the ranking utility for a given maximum allowed level of privacy leakage of the protected node-pairs. Under a recently-introduced latent node embedding model, we present a formal trade-off between privacy and LP utility. Extensive experiments with several real-life graphs and several LP heuristics show that DPLP can trade off between privacy and predictive performance more effectively than several alternatives.
Surfacing a ranked list of items for a search query to help buyers discover inventory and make purchase decisions is a critical problem in eCommerce search. Typically, items are independently predicted with a probability of sale with respect to a given search query. But in a dynamic marketplace like eBay, even for a single product, there are various different factors distinguishing one item from another which can influence the purchase decision for the user. Users have to make a purchase decision by considering all of these options. Majority of the existing learning to rank algorithms model the relative relevance between labeled items only at the loss functions like pairwise or list-wise losses. But they are limited to point-wise scoring functions where items are ranked independently based on the features of the item itself. In this paper, we study the influence of an item's neighborhood to its purchase decision. Here, we consider the neighborhood as the items ranked above and below the current item in search results. By adding delta features comparing items within a neighborhood and learning a ranking model, we are able to experimentally show that the new ranker with delta features outperforms our baseline ranker in terms of Mean Reciprocal Rank (MRR). The ranking models with proposed delta features result in $3-5\%$ improvement in MRR over the baseline model. We also study impact of different sizes for neighborhood. Experimental results show that neighborhood size $3$ perform the best based on MRR with an improvement of $4-5\%$ over the baseline model.
Users increasingly rely on social media feeds for consuming daily information. The items in a feed, such as news, questions, songs, etc., usually result from the complex interplay of a user's social contacts, her interests and her actions on the platform. The relationship of the user's own behavior and the received feed is often puzzling, and many users would like to have a clear explanation on why certain items were shown to them. Transparency and explainability are key concerns in the modern world of cognitive overload, filter bubbles, user tracking, and privacy risks. This paper presents FAIRY, a framework that systematically discovers, ranks, and explains relationships between users' actions and items in their social media feeds. We model the user's local neighborhood on the platform as an interaction graph, a form of heterogeneous information network constructed solely from information that is easily accessible to the concerned user. We posit that paths in this interaction graph connecting the user and her feed items can act as pertinent explanations for the user. These paths are scored with a learning-to-rank model that captures relevance and surprisal. User studies on two social platforms demonstrate the practical viability and user benefits of the FAIRY method.
Document summarisation can be formulated as a sequential decision-making problem, which can be solved by Reinforcement Learning (RL) algorithms. The predominant RL paradigm for summarisation learns a cross-input policy, which requires considerable time, data and parameter tuning due to the huge search spaces and the delayed rewards. Learning input-specific RL policies is a more efficient alternative but so far depends on handcrafted rewards, which are difficult to design and yield poor performance. We propose RELIS, a novel RL paradigm that learns a reward function with Learning-to-Rank (L2R) algorithms at training time and uses this reward function to train an input-specific RL policy at test time. We prove that RELIS guarantees to generate near-optimal summaries with appropriate L2R and RL algorithms. Empirically, we evaluate our approach on extractive multi-document summarisation. We show that RELIS reduces the training time by two orders of magnitude compared to the state-of-the-art models while performing on par with them.