We propose a new framework for entity and event extraction based on generative adversarial imitation learning—an inverse reinforcement learning method using a generative adversarial network (GAN). We assume that instances and labels yield to various extents of difficulty and the gains and penalties (rewards) are expected to be diverse. We utilize discriminators to estimate proper rewards according to the difference between the labels committed by the ground-truth (expert) and the extractor (agent). Our experiments demonstrate that the proposed framework outperforms state-of-the-art methods.
Current systems of fine-grained entity typing use distant supervision in conjunction with existing knowledge bases to assign categories (type labels) to entity mentions. However, the type labels so obtained from knowledge bases are often noisy (i.e., incorrect for the entity mention's local context). We define a new task, Label Noise Reduction in Entity Typing (LNR), to be the automatic identification of correct type labels (type-paths) for training examples, given the set of candidate type labels obtained by distant supervision with a given type hierarchy. The unknown type labels for individual entity mentions and the semantic similarity between entity types pose unique challenges for solving the LNR task. We propose a general framework, called PLE, to jointly embed entity mentions, text features and entity types into the same low-dimensional space where, in that space, objects whose types are semantically close have similar representations. Then we estimate the type-path for each training example in a top-down manner using the learned embeddings. We formulate a global objective for learning the embeddings from text corpora and knowledge bases, which adopts a novel margin-based loss that is robust to noisy labels and faithfully models type correlation derived from knowledge bases. Our experiments on three public typing datasets demonstrate the effectiveness and robustness of PLE, with an average of 25% improvement in accuracy compared to next best method.
25 November 2019 in IEEE Transactions on Mobile Computing
In mobile crowdsourcing, the accuracy of the collected data is usually hard to ensure. Researchers have proposed techniques to identify truth from noisy data by inferring and utilizing the reliability of mobile users, and allocate tasks to users with higher reliability. However, they neglect the fact that a user may only have expertise on some problems (in some domains), but not others, and hence causing two problems: low estimation accuracy in truth analysis and ineffective task allocation. To address these problems, we propose Expertise-aware Truth Analysis and Task Allocation (ETA2), which can effectively infer user expertise, and then estimate truth and allocate tasks based on the inferred expertise. ETA2 relies on a novel semantic analysis method to identify the expertise, and an expertise-aware truth analysis method to find the truth. For expertise-aware task allocation in ETA2, we formalize and solve two problems based on the optimization objectives: max-quality task allocation which maximizes the probability for tasks to be allocated to users with high expertise and min-cost task allocation which minimizes the cost of task allocation while ensuring high-quality data are collected. Experimental results based on two real-world datasets and one synthetic dataset demonstrate that ETA2 significantly outperforms existing solutions.
20 May 2015 in BMC Medical Informatics and Decision Making
The Entity Linking (EL) task links entity mentions from an unstructured document to entities in a knowledge base. Although this problem is well-studied in news and social media, this problem has not received much attention in the life science domain. One outcome of tackling the EL problem in the life sciences domain is to enable scientists to build computational models of biological processes with more efficiency. However, simply applying a news-trained entity linker produces inadequate results.
When integrating information from multiple sources, it is common to encounter conflicting answers to the same question. Truth discovery is to infer the most accurate and complete integrated answers from conflicting sources. In some cases, there exist questions for which the true answers are excluded from the candidate answers provided by all sources. Without any prior knowledge, these questions, named no-truth questions, are difficult to be distinguished from the questions that have true answers, named has-truth questions. In particular, these no-truth questions degrade the precision of the answer integration system. We address such a challenge by introducing source quality, which is made up of three fine-grained measures: silent rate, false spoken rate and true spoken rate. By incorporating these three measures, we propose a probabilistic graphical model, which simultaneously infers truth as well as source quality without any a priori training involving ground truth answers. Moreover, since inferring this graphical model requires parameter tuning of the prior of truth, we propose an initialization scheme based upon a quantity named truth existence score, which synthesizes two indicators, namely, participation rate and consistency rate. Compared with existing methods, our method can effectively filter out no-truth questions, which results in more accurate source quality estimation. Consequently, our method provides more accurate and complete answers to both has-truth and no-truth questions. Experiments on three real-world datasets illustrate the notable advantage of our method over existing state-of-the-art truth discovery methods.
23 August 2013 in Knowledge and Information Systems