Entity Linking (EL) is the task of automatically identifying entity mentions in a piece of text and resolving them to a corresponding entity in a reference knowledge base like Wikipedia. There is a large number of EL tools available for different types of documents and domains, yet EL remains a challenging task where the lack of precision on particularly ambiguous mentions often spoils the usefulness of automated disambiguation results in real applications. A priori approximations of the difficulty to link a particular entity mention can facilitate flagging of critical cases as part of semi-automated EL systems, while detecting latent factors that affect the EL performance, like corpus-specific features, can provide insights on how to improve a system based on the special characteristics of the underlying corpus. In this paper, we first introduce a consensus-based method to generate difficulty labels for entity mentions on arbitrary corpora. The difficulty labels are then exploited as training data for a supervised classification task able to predict the EL difficulty of entity mentions using a variety of features. Experiments over a corpus of news articles show that EL difficulty can be estimated with high accuracy, revealing also latent features that affect EL performance. Finally, evaluation results demonstrate the effectiveness of the proposed method to inform semi-automated EL pipelines.

Classifiers used in the wild, in particular for safety-critical systems, should not only have good generalization properties but also should know when they don't know, in particular make low confidence predictions far away from the training data. We show that ReLU type neural networks which yield a piecewise linear classifier function fail in this regard as they produce almost always high confidence predictions far away from the training data. For bounded domains like images we propose a new robust optimization technique similar to adversarial training which enforces low confidence predictions far away from the training data. We show that this technique is surprisingly effective in reducing the confidence of predictions far away from the training data while maintaining high confidence predictions and similar test error on the original classification task compared to standard training.

Transcribed datasets typically contain speaker identity for each instance in the data. We investigate two ways to incorporate this information during training: Multi-Task Learning and Adversarial Learning. In multi-task learning, the goal is speaker prediction; we expect a performance improvement with this joint training if the two tasks of speech recognition and speaker recognition share a common set of underlying features. In contrast, adversarial learning is a means to learn representations invariant to the speaker. We then expect better performance if this learnt invariance helps generalizing to new speakers. While the two approaches seem natural in the context of speech recognition, they are incompatible because they correspond to opposite gradients back-propagated to the model. In order to better understand the effect of these approaches in terms of error rates, we compare both strategies in controlled settings. Moreover, we explore the use of additional untranscribed data in a semi-supervised, adversarial learning manner to improve error rates. Our results show that deep models trained on big datasets already develop invariant representations to speakers without any auxiliary loss. When considering adversarial learning and multi-task learning, the impact on the acoustic model seems minor. However, models trained in a semi-supervised manner can improve error-rates.

Autism spectrum condition (ASC) or autism spectrum disorder (ASD) is primarily identified with the help of behavioral indications encompassing social, sensory and motor characteristics. Although categorized, recurring motor actions are measured during diagnosis, quantifiable measures that ascertain kinematic physiognomies in the movement configurations of autistic persons are not adequately studied, hindering the advances in understanding the etiology of motor mutilation. Subject aspects such as behavioral characters that influences ASD need further exploration. Presently, limited autism datasets concomitant with screening ASD are available, and a majority of them are genetic. Hence, in this study, we used a dataset related to autism screening enveloping ten behavioral and ten personal attributes that have been effective in diagnosing ASD cases from controls in behavior science. ASD diagnosis is time exhaustive and uneconomical. The burgeoning ASD cases worldwide mandate a need for the fast and economical screening tool. Our study aimed to implement an artificial neural network with the Levenberg-Marquardt algorithm to detect ASD and examine its predictive accuracy. Consecutively, develop a clinical decision support system for early ASD identification.

Islamophobic hate speech on social media inflicts considerable harm on both targeted individuals and wider society, and also risks reputational damage for the host platforms. Accordingly, there is a pressing need for robust tools to detect and classify Islamophobic hate speech at scale. Previous research has largely approached the detection of Islamophobic hate speech on social media as a binary task. However, the varied nature of Islamophobia means that this is often inappropriate for both theoretically-informed social science and effectively monitoring social media. Drawing on in-depth conceptual work we build a multi-class classifier which distinguishes between non-Islamophobic, weak Islamophobic and strong Islamophobic content. Accuracy is 77.6% and balanced accuracy is 83%. We apply the classifier to a dataset of 109,488 tweets produced by far right Twitter accounts during 2017. Whilst most tweets are not Islamophobic, weak Islamophobia is considerably more prevalent (36,963 tweets) than strong (14,895 tweets). Our main input feature is a gloVe word embeddings model trained on a newly collected corpus of 140 million tweets. It outperforms a generic word embeddings model by 5.9 percentage points, demonstrating the importan4ce of context. Unexpectedly, we also find that a one-against-one multi class SVM outperforms a deep learning algorithm.

Using the outputs obtained from the online testing, it is not so difficult to collect a large-scale of learning data. We may be able to actively tackle the collected data to find the optimal strategies for better learning methods. It is also important to analyze the data theoretically (see [23]). This paper is aimed at obtaining effective learning strategies for students at risk for failing courses and/or dropping out, using a large-scale of learning data collected from the online testings. In this paper, unlike the conventional methods using the correct answer rate (CAR) to identify the ability of a student (e.g., see [13]), we use the ability obtained from the item response theory (IRT, e.g., see [1], [4], [17]), and we show a new method to identify students at risk as early as possible using the IRT results.

With the emergence of the Hospital Readmission Reduction Program of the Center for Medicare and Medicaid Services on October 1, 2012, forecasting unplanned patient readmission risk became crucial to the healthcare domain. There are tangible works in the literature emphasizing on developing readmission risk prediction models; However, the models are not accurate enough to be deployed in an actual clinical setting. Our study considers patient readmission risk as the objective for optimization and develops a useful risk prediction model to address unplanned readmissions. Furthermore, Genetic Algorithm and Greedy Ensemble is used to optimize the developed model constraints.

Cervical cancer is the leading gynecological malignancy worldwide. This paper presents diverse classification techniques and shows the advantage of feature selection approaches to the best predicting of cervical cancer disease. There are thirty-two attributes with eight hundred and fifty-eight samples. Besides, this data suffers from missing values and imbalance data. Therefore, over-sampling, under-sampling and embedded over and under sampling have been used. Furthermore, dimensionality reduction techniques are required for improving the accuracy of the classifier. Therefore, feature selection methods have been studied as they divided into two distinct categories, filters and wrappers. The results show that age, first sexual intercourse, number of pregnancies, smokes, hormonal contraceptives, and STDs: genital herpes are the main predictive features with high accuracy with 97.5%. Decision Tree classifier is shown to be advantageous in handling classification assignment with excellent performance.

The burgeoning need for kidney transplantation mandates immediate attention. Mismatch of deceased donor-recipient kidney leads to post-transplant death. To ensure ideal kidney donor-recipient match and minimize post-transplant deaths, the paper develops a prediction model that identifies factors that determine the probability of success of renal transplantation, that is, if the kidney procured from the deceased donor can be transplanted or discarded. The paper conducts a study enveloping data for 584 imported kidneys collected from 12 transplant centers associated with an organ procurement organization located in New York City, NY. The predicting model yielding best performance measures can be beneficial to the healthcare industry. Transplant centers and organ procurement organizations can take advantage of the prediction model to efficiently predict the outcome of kidney transplantation. Consequently, it will reduce the mortality rate caused by mismatching of donor-recipient kidney transplantation during the surgery.

Treelets, introduced by Lee, Nadler, and Wasserman [1, 2], is a method to produce a multiscale, hierarchicaldecomposition of unordered data. The central premise of Treelets is to exploit sparsity and capture intrinsic localized structures with only a few features, represented interms of an orthonormal basis. The hierarchical tree constructed by the treelet algorithm provides a scale-based partition of the data that can be used for classification, specially for cluster analysis [3]. Cluster analysis, also called clustering, is concerned with finding a partition of a set such that its corresponding equivalence class captures similarity of its elements. The Treelet approach is an example of hierarchical clustering (HC) [4], which is a type of methods that provides a nested and multiscale clustering.