Many practical applications of AI in medicine consist of semi-supervised discovery: The investigator aims to identify features of interest at a resolution more fine-grained than that of the available human labels. This is often the scenario faced in healthcare applications as coarse, high-level labels (e.g., billing codes) are often the only sources that are readily available. These challenges are compounded for modalities such as text, where the feature space is very high-dimensional, and often contains considerable amounts of noise. In this work, we generalize a recently proposed zero-shot sequence labeling method, "binary labeling via a convolutional decomposition", to the case where the available document-level human labels are themselves relatively high-dimensional. The approach yields classification with "introspection", relating the fine-grained features of an inference-time prediction to their nearest neighbors from the training set, under the model. The approach is effective, yet parsimonious, as demonstrated on a well-studied MIMIC-III multi-label classification task of electronic health record data, and is useful as a tool for organizing the analysis of neural model predictions and high-dimensional datasets. Our proposed approach yields both a competitively effective classification model and an interrogation mechanism to aid healthcare workers in understanding the salient features that drive the model's predictions.
The training of deep-learning-based text classification models relies heavily on a huge amount of annotation data, which is difficult to obtain. When the labeled data is scarce, models tend to struggle to achieve satisfactory performance. However, human beings can distinguish new categories very efficiently with few examples. This is mainly due to the fact that human beings can leverage knowledge obtained from relevant tasks. Inspired by human intelligence, we propose to introduce external knowledge into few-shot learning to imitate human knowledge. A novel parameter generator network is investigated to this end, which is able to use the external knowledge to generate relation network parameters. Metrics can be transferred among tasks when equipped with these generated parameters, so that similar tasks use similar metrics while different tasks use different metrics. Through experiments, we demonstrate that our method outperforms the state-of-the-art few-shot text classification models.
Generating explanations for neural networks has become crucial for their applications in real-world with respect to reliability and trustworthiness. In natural language processing, existing methods usually provide important features which are words or phrases selected from an input text as an explanation, but ignore the interactions between them. It poses challenges for humans to interpret an explanation and connect it to model prediction. In this work, we build hierarchical explanations by detecting feature interactions. Such explanations visualize how words and phrases are combined at different levels of the hierarchy, which can help users understand the decision-making of black-box models. The proposed method is evaluated with three neural text classifiers (LSTM, CNN, and BERT) on two benchmark datasets, via both automatic and human evaluations. Experiments show the effectiveness of the proposed method in providing explanations that are both faithful to models and interpretable to humans.
Modern text classification models are susceptible to adversarial examples, perturbed versions of the original text indiscernible by humans which get misclassified by the model. Recent works in NLP use rule-based synonym replacement strategies to generate adversarial examples. These strategies can lead to out-of-context and unnaturally complex token replacements, which are easily identifiable by humans. We present BAE, a black box attack for generating adversarial examples using contextual perturbations from a BERT masked language model. BAE replaces and inserts tokens in the original text by masking a portion of the text and leveraging the BERT-MLM to generate alternatives for the masked tokens. Through automatic and human evaluations, we show that BAE performs a stronger attack, in addition to generating adversarial examples with improved grammaticality and semantic coherence as compared to prior work.
Text classification has become increasingly challenging due to the continuous refinement of classification label granularity and the expansion of classification label scale. To address that, some research has been applied onto strategies that exploit the hierarchical structure in problems with a large number of categories. At present, hierarchical text classification (HTC) has received extensive attention and has broad application prospects. Making full use of the relationship between parent category and child category in text classification task can greatly improve the performance of classification. In this paper, We propose a joint embedding of text and parent category based on hierarchical fine-tuning ordered neurons LSTM (HFT-ONLSTM) for HTC. Our method makes full use of the connection between the upper-level and lower-level labels. Experiments show that our model outperforms the state-of-the-art hierarchical model at a lower computation cost.
Machine learning-based multi-label medical text classifications can be used to enhance the understanding of the human body and aid the need for patient care. We present a broad study on clinical natural language processing techniques to maximise a feature representing text when predicting medical codes on patients with multi-morbidity. We present results of multi-label medical text classification problems with 18, 50 and 155 labels. We compare several variations to embeddings, text tagging, and pre-processing. For imbalanced data we show that labels which occur infrequently, benefit the most from additional features incorporated in embeddings. We also show that high dimensional embeddings pre-trained using health-related data present a significant improvement in a multi-label setting, similarly to the way they improve performance for binary classification. High dimensional embeddings from this research are made available for public use.
The improvement in Information Technology has encouraged the use of Igbo in the creation of text such as resources and news articles online. Text similarity is of great importance in any text-based applications. This paper presents a comparative analysis of n-gram text representation on Igbo text document similarity. It adopted Euclidean similarity measure to determine the similarities between Igbo text documents represented with two word-based n-gram text representation (unigram and bigram) models. The evaluation of the similarity measure is based on the adopted text representation models. The model is designed with Object-Oriented Methodology and implemented with Python programming language with tools from Natural Language Toolkits (NLTK). The result shows that unigram represented text has highest distance values whereas bigram has the lowest corresponding distance values. The lower the distance value, the more similar the two documents and better the quality of the model when used for a task that requires similarity measure. The similarity of two documents increases as the distance value moves down to zero (0). Ideally, the result analyzed revealed that Igbo text document similarity measured on bigram represented text gives accurate similarity result. This will give better, effective and accurate result when used for tasks such as text classification, clustering and ranking on Igbo text.
Automatic emotion categorization has been predominantly formulated as text classification in which textual units are assigned to an emotion from a predefined inventory, for instance following the fundamental emotion classes proposed by Paul Ekman (fear, joy, anger, disgust, sadness, surprise) or Robert Plutchik (adding trust, anticipation). This approach ignores existing psychological theories to some degree, which provide explanations regarding the perception of events. For instance, the description that somebody discovers a snake is associated with fear, based on the appraisal as being an unpleasant and non-controllable situation. This emotion reconstruction is even possible without having access to explicit reports of a subjective feeling (for instance expressing this with the words "I am afraid."). Automatic classification approaches therefore need to learn properties of events as latent variables (for instance that the uncertainty and the mental or physical effort associated with the encounter of a snake leads to fear). With this paper, we propose to make such interpretations of events explicit, following theories of cognitive appraisal of events, and show their potential for emotion classification when being encoded in classification models. Our results show that high quality appraisal dimension assignments in event descriptions lead to an improvement in the classification of discrete emotion categories. We make our corpus of appraisal-annotated emotion-associated event descriptions publicly available.
While the strength of Topological Data Analysis has been explored in many studies on high dimensional numeric data, it is still a challenging task to apply it to text. As the primary goal in topological data analysis is to define and quantify the shapes in numeric data, defining shapes in the text is much more challenging, even though the geometries of vector spaces and conceptual spaces are clearly relevant for information retrieval and semantics. In this paper, we examine two different methods of extraction of topological features from text, using as the underlying representations of words the two most popular methods, namely word embeddings and TF-IDF vectors. To extract topological features from the word embedding space, we interpret the embedding of a text document as high dimensional time series, and we analyze the topology of the underlying graph where the vertices correspond to different embedding dimensions. For topological data analysis with the TF-IDF representations, we analyze the topology of the graph whose vertices come from the TF-IDF vectors of different blocks in the textual document. In both cases, we apply homological persistence to reveal the geometric structures under different distance resolutions. Our results show that these topological features carry some exclusive information that is not captured by conventional text mining methods. In our experiments we observe adding topological features to the conventional features in ensemble models improves the classification results (up to 5\%). On the other hand, as expected, topological features by themselves may be not sufficient for effective classification. It is an open problem to see whether TDA features from word embeddings might be sufficient, as they seem to perform within a range of few points from top results obtained with a linear support vector classifier.
In recent years, topological data analysis has been utilized for a wide range of problems to deal with high dimensional noisy data. While text representations are often high dimensional and noisy, there are only a few work on the application of topological data analysis in natural language processing. In this paper, we introduce a novel algorithm to extract topological features from word embedding representation of text that can be used for text classification. Working on word embeddings, topological data analysis can interpret the embedding high-dimensional space and discover the relations among different embedding dimensions. We will use persistent homology, the most commonly tool from topological data analysis, for our experiment. Examining our topological algorithm on long textual documents, we will show our defined topological features may outperform conventional text mining features.
The automatic identification of propaganda has gained significance in recent years due to technological and social changes in the way news is generated and consumed. That this task can be addressed effectively using BERT, a powerful new architecture which can be fine-tuned for text classification tasks, is not surprising. However, propaganda detection, like other tasks that deal with news documents and other forms of decontextualized social communication (e.g. sentiment analysis), inherently deals with data whose categories are simultaneously imbalanced and dissimilar. We show that BERT, while capable of handling imbalanced classes with no additional data augmentation, does not generalise well when the training and test data are sufficiently dissimilar (as is often the case with news sources, whose topics evolve over time). We show how to address this problem by providing a statistical measure of similarity between datasets and a method of incorporating cost-weighting into BERT when the training and test sets are dissimilar. We test these methods on the Propaganda Techniques Corpus (PTC) and achieve the second-highest score on sentence-level propaganda classification.
In Multi-Label Text Classification (MLTC), one sample can belong to more than one class. It is observed that most MLTC tasks, there are dependencies or correlations among labels. Existing methods tend to ignore the relationship among labels. In this paper, a graph attention network-based model is proposed to capture the attentive dependency structure among the labels. The graph attention network uses a feature matrix and a correlation matrix to capture and explore the crucial dependencies between the labels and generate classifiers for the task. The generated classifiers are applied to sentence feature vectors obtained from the text feature extraction network (BiLSTM) to enable end-to-end training. Attention allows the system to assign different weights to neighbor nodes per label, thus allowing it to learn the dependencies among labels implicitly. The results of the proposed model are validated on five real-world MLTC datasets. The proposed model achieves similar or better performance compared to the previous state-of-the-art models.
The dominant approaches to text representation in natural language rely on learning embeddings on massive corpora which have convenient properties such as compositionality and distance preservation. In this paper, we develop a novel method to learn a heavy-tailed embedding with desirable regularity properties regarding the distributional tails, which allows to analyze the points far away from the distribution bulk using the framework of multivariate extreme value theory. In particular, a classifier dedicated to the tails of the proposed embedding is obtained which performance outperforms the baseline. This classifier exhibits a scale invariance property which we leverage by introducing a novel text generation method for label preserving dataset augmentation. Numerical experiments on synthetic and real text data demonstrate the relevance of the proposed framework and confirm that this method generates meaningful sentences with controllable attribute, e.g. positive or negative sentiment.
Multi-class text classification is one of the key problems in machine learning and natural language processing. Emerging neural networks deal with the problem using a multi-output softmax layer and achieve substantial progress, but they do not explicitly learn the correlation among classes. In this paper, we use a multi-task framework to address multi-class classification, where a multi-class classifier and multiple binary classifiers are trained together. Moreover, we employ adversarial training to distinguish the class-specific features and the class-agnostic features. The model benefits from better feature representation. We conduct experiments on two large-scale multi-class text classification tasks and demonstrate that the proposed architecture outperforms baseline approaches.
Today text classification models have been widely used. However, these classifiers are found to be easily fooled by adversarial examples. Fortunately, standard attacking methods generate adversarial texts in a pair-wise way, that is, an adversarial text can only be created from a real-world text by replacing a few words. In many applications, these texts are limited in numbers, therefore their corresponding adversarial examples are often not diverse enough and sometimes hard to read, thus can be easily detected by humans and cannot create chaos at a large scale. In this paper, we propose an end to end solution to efficiently generate adversarial texts from scratch using generative models, which are not restricted to perturbing the given texts. We call it unrestricted adversarial text generation. Specifically, we train a conditional variational autoencoder (VAE) with an additional adversarial loss to guide the generation of adversarial examples. Moreover, to improve the validity of adversarial texts, we utilize discrimators and the training framework of generative adversarial networks (GANs) to make adversarial texts consistent with real data. Experimental results on sentiment analysis demonstrate the scalability and efficiency of our method. It can attack text classification models with a higher success rate than existing methods, and provide acceptable quality for humans in the meantime.
Summarizing data samples by quantitative measures has a long history, with descriptive statistics being a case in point. However, as natural language processing methods flourish, there are still insufficient characteristic metrics to describe a collection of texts in terms of the words, sentences, or paragraphs they comprise. In this work, we propose metrics of diversity, density, and homogeneity that quantitatively measure the dispersion, sparsity, and uniformity of a text collection. We conduct a series of simulations to verify that each metric holds desired properties and resonates with human intuitions. Experiments on real-world datasets demonstrate that the proposed characteristic metrics are highly correlated with text classification performance of a renowned model, BERT, which could inspire future applications.
Learning continuous representations of discrete objects such as text, users, movies, and URLs lies at the heart of many applications including language and user modeling. When using discrete objects as input to neural networks, we often ignore the underlying structures (e.g., natural groupings and similarities) and embed the objects independently into individual vectors. As a result, existing methods do not scale to large vocabulary sizes. In this paper, we design a simple and efficient embedding algorithm that learns a small set of anchor embeddings and a sparse transformation matrix. We call our method Anchor & Transform (ANT) as the embeddings of discrete objects are a sparse linear combination of the anchors, weighted according to the transformation matrix. ANT is scalable, flexible, and end-to-end trainable. We further provide a statistical interpretation of our algorithm as a Bayesian nonparametric prior for embeddings that encourages sparsity and leverages natural groupings among objects. By deriving an approximate inference algorithm based on Small Variance Asymptotics, we obtain a natural extension that automatically learns the optimal number of anchors instead of having to tune it as a hyperparameter. On text classification, language modeling, and movie recommendation benchmarks, we show that ANT is particularly suitable for large vocabulary sizes and demonstrates stronger performance with fewer parameters (up to 40x compression) as compared to existing compression baselines.
Many text classification tasks are domain-dependent, and various domain adaptation approaches have been proposed to predict unlabeled data in a new domain. Domain-adversarial neural networks (DANN) and their variants have been used widely recently and have achieved promising results for this problem. However, most of these approaches assume that the label proportions of the source and target domains are similar, which rarely holds in most real-world scenarios. Sometimes the label shift can be large and the DANN fails to learn domain-invariant features. In this study, we focus on unsupervised domain adaptation of text classification with label shift and introduce a domain adversarial network with label proportions estimation (DAN-LPE) framework. The DAN-LPE simultaneously trains a domain adversarial net and processes label proportions estimation by the confusion of the source domain and the predictions of the target domain. Experiments show the DAN-LPE achieves a good estimate of the target label distributions and reduces the label shift to improve the classification performance.
For reliability, it is important that the predictions made by machine learning methods are interpretable by human. In general, deep neural networks (DNNs) can provide accurate predictions, although it is difficult to interpret why such predictions are obtained by DNNs. On the other hand, interpretation of linear models is easy, although their predictive performance would be low since real-world data is often intrinsically non-linear. To combine both the benefits of the high predictive performance of DNNs and high interpretability of linear models into a single model, we propose neural generators of sparse local linear models (NGSLLs). The sparse local linear models have high flexibility as they can approximate non-linear functions. The NGSLL generates sparse linear weights for each sample using DNNs that take original representations of each sample (e.g., word sequence) and their simplified representations (e.g., bag-of-words) as input. By extracting features from the original representations, the weights can contain rich information to achieve high predictive performance. Additionally, the prediction is interpretable because it is obtained by the inner product between the simplified representations and the sparse weights, where only a small number of weights are selected by our gate module in the NGSLL. In experiments with real-world datasets, we demonstrate the effectiveness of the NGSLL quantitatively and qualitatively by evaluating prediction performance and visualizing generated weights on image and text classification tasks.
Bidirectional Encoder Representations from Transformers (BERT) has recently achieved state-of-the-art performance on a broad range of NLP tasks including sentence classification, machine translation, and question answering. The BERT model architecture is derived primarily from the transformer. Prior to the transformer era, bidirectional Long Short-Term Memory (BLSTM) has been the dominant modeling architecture for neural machine translation and question answering. In this paper, we investigate how these two modeling techniques can be combined to create a more powerful model architecture. We propose a new architecture denoted as Transformer with BLSTM (TRANS-BLSTM) which has a BLSTM layer integrated to each transformer block, leading to a joint modeling framework for transformer and BLSTM. We show that TRANS-BLSTM models consistently lead to improvements in accuracy compared to BERT baselines in GLUE and SQuAD 1.1 experiments. Our TRANS-BLSTM model obtains an F1 score of 94.01% on the SQuAD 1.1 development dataset, which is comparable to the state-of-the-art result.