Manual coding of text data from open-ended questions into different categories is time consuming and expensive. Automated coding uses statistical/machine learning to train on a small subset of manually coded text answers. Recently, pre-training a general language model on vast amounts of unrelated data and then adapting the model to the specific application has proven effective in natural language processing. Using two data sets, we empirically investigate whether BERT, the currently dominant pre-trained language model, is more effective at automated coding of answers to open-ended questions than other non-pre-trained statistical learning approaches. We found fine-tuning the pre-trained BERT parameters is essential as otherwise BERT's is not competitive. Second, we found fine-tuned BERT barely beats the non-pre-trained statistical learning approaches in terms of classification accuracy when trained on 100 manually coded observations. However, BERT's relative advantage increases rapidly when more manually coded observations (e.g. 200-400) are available for training. We conclude that for automatically coding answers to open-ended questions BERT is preferable to non-pretrained models such as support vector machines and boosting.
相關內容
One-class classification (OCC) is the problem of deciding whether an observed sample belongs to a target class or not. We consider the problem of learning an OCC model when the dataset available at the learning stage contains only samples from the target class. We aim at obtaining a classifier that performs as the generalized likelihood ratio test (GLRT), which is a well-known and provably optimal (under specific assumptions) classifier when the statistic of the target class is available. To this end, we consider both the multilayer perceptron neural network (NN) and the support vector machine (SVM) models. They are trained as two-class classifiers using an artificial dataset for the alternative class, obtained by generating random samples, uniformly over the domain of the target-class dataset. We prove that, under suitable assumptions, the models converge (with a large dataset) to the GLRT. Moreover, we show that the one-class least squares SVM (OCLSSVM) at convergence performs as the GLRT, with a suitable transformation function. Lastly, we compare the obtained solutions with the autoencoder (AE) classifier, which does not in general provide the GLRT
Pre-trained language models (LMs) are shown to easily generate toxic language. In this work, we systematically explore domain-adaptive training to reduce the toxicity of language models. We conduct this study on three dimensions: training corpus, model size, and parameter efficiency. For the training corpus, we propose to leverage the generative power of LMs and generate nontoxic datasets for domain-adaptive training, which mitigates the exposure bias and is shown to be more data-efficient than using a curated pre-training corpus. We demonstrate that the self-generation method consistently outperforms the existing baselines across various model sizes on both automatic and human evaluations, even when it uses a 1/3 smaller training corpus. We then comprehensively study detoxifying LMs with parameter sizes ranging from 126M up to 530B (3x larger than GPT-3), a scale that has never been studied before. We find that i) large LMs have similar toxicity levels as smaller ones given the same pre-training corpus, and ii) large LMs require more endeavor to detoxify. We also explore parameter-efficient training methods for detoxification. We demonstrate that adding and training adapter-only layers in LMs not only saves a lot of parameters but also achieves a better trade-off between toxicity and perplexity than whole model adaptation for the large-scale models.
Transfer learning (TL) from pretrained deep models is a standard practice in modern medical image classification (MIC). However, what levels of features to be reused are problem-dependent, and uniformly finetuning all layers of pretrained models may be suboptimal. This insight has partly motivated the recent \emph{differential} TL strategies, such as TransFusion (TF) and layer-wise finetuning (LWFT), which treat the layers in the pretrained models differentially. In this paper, we add one more strategy into this family, called \emph{TruncatedTL}, which reuses and finetunes appropriate bottom layers and directly discards the remaining layers. This yields not only superior MIC performance but also compact models for efficient inference, compared to other differential TL methods. We validate the performance and model efficiency of TruncatedTL on three MIC tasks covering both 2D and 3D images. For example, on the BIMCV COVID-19 classification dataset, we obtain improved performance with around $1/4$ model size and $2/3$ inference time compared to the standard full TL model. Code is available at //github.com/sun-umn/Transfer-Learning-in-Medical-Imaging.
Implicit discourse relations bind smaller linguistic units into coherent texts. Automatic sense prediction for implicit relations is hard, because it requires understanding the semantics of the linked arguments. Furthermore, annotated datasets contain relatively few labeled examples, due to the scale of the phenomenon: on average each discourse relation encompasses several dozen words. In this paper, we explore the utility of pre-trained sentence embeddings as base representations in a neural network for implicit discourse relation sense classification. We present a series of experiments using both supervised end-to-end trained models and pre-trained sentence encoding techniques - SkipThought, Sent2vec and Infersent. The pre-trained embeddings are competitive with the end-to-end model, and the approaches are complementary, with combined models yielding significant performance improvements on two of the three evaluations.
Generative models are now capable of producing highly realistic images that look nearly indistinguishable from the data on which they are trained. This raises the question: if we have good enough generative models, do we still need datasets? We investigate this question in the setting of learning general-purpose visual representations from a black-box generative model rather than directly from data. Given an off-the-shelf image generator without any access to its training data, we train representations from the samples output by this generator. We compare several representation learning methods that can be applied to this setting, using the latent space of the generator to generate multiple "views" of the same semantic content. We show that for contrastive methods, this multiview data can naturally be used to identify positive pairs (nearby in latent space) and negative pairs (far apart in latent space). We find that the resulting representations rival those learned directly from real data, but that good performance requires care in the sampling strategy applied and the training method. Generative models can be viewed as a compressed and organized copy of a dataset, and we envision a future where more and more "model zoos" proliferate while datasets become increasingly unwieldy, missing, or private. This paper suggests several techniques for dealing with visual representation learning in such a future. Code is released on our project page: //ali-design.github.io/GenRep/
Graph classification aims to perform accurate information extraction and classification over graphstructured data. In the past few years, Graph Neural Networks (GNNs) have achieved satisfactory performance on graph classification tasks. However, most GNNs based methods focus on designing graph convolutional operations and graph pooling operations, overlooking that collecting or labeling graph-structured data is more difficult than grid-based data. We utilize meta-learning for fewshot graph classification to alleviate the scarce of labeled graph samples when training new tasks.More specifically, to boost the learning of graph classification tasks, we leverage GNNs as graph embedding backbone and meta-learning as training paradigm to capture task-specific knowledge rapidly in graph classification tasks and transfer them to new tasks. To enhance the robustness of meta-learner, we designed a novel step controller driven by Reinforcement Learning. The experiments demonstrate that our framework works well compared to baselines.
Pre-trained deep neural network language models such as ELMo, GPT, BERT and XLNet have recently achieved state-of-the-art performance on a variety of language understanding tasks. However, their size makes them impractical for a number of scenarios, especially on mobile and edge devices. In particular, the input word embedding matrix accounts for a significant proportion of the model's memory footprint, due to the large input vocabulary and embedding dimensions. Knowledge distillation techniques have had success at compressing large neural network models, but they are ineffective at yielding student models with vocabularies different from the original teacher models. We introduce a novel knowledge distillation technique for training a student model with a significantly smaller vocabulary as well as lower embedding and hidden state dimensions. Specifically, we employ a dual-training mechanism that trains the teacher and student models simultaneously to obtain optimal word embeddings for the student vocabulary. We combine this approach with learning shared projection matrices that transfer layer-wise knowledge from the teacher model to the student model. Our method is able to compress the BERT_BASE model by more than 60x, with only a minor drop in downstream task metrics, resulting in a language model with a footprint of under 7MB. Experimental results also demonstrate higher compression efficiency and accuracy when compared with other state-of-the-art compression techniques.
Extreme multi-label text classification (XMC) aims to tag each input text with the most relevant labels from an extremely large label set, such as those that arise in product categorization and e-commerce recommendation. Recently, pretrained language representation models such as BERT achieve remarkable state-of-the-art performance across a wide range of NLP tasks including sentence classification among small label sets (typically fewer than thousands). Indeed, there are several challenges in applying BERT to the XMC problem. The main challenges are: (i) the difficulty of capturing dependencies and correlations among labels, whose features may come from heterogeneous sources, and (ii) the tractability to scale to the extreme label setting as the model size can be very large and scale linearly with the size of the output space. To overcome these challenges, we propose X-BERT, the first feasible attempt to finetune BERT models for a scalable solution to the XMC problem. Specifically, X-BERT leverages both the label and document text to build label representations, which induces semantic label clusters in order to better model label dependencies. At the heart of X-BERT is finetuning BERT models to capture the contextual relations between input text and the induced label clusters. Finally, an ensemble of the different BERT models trained on heterogeneous label clusters leads to our best final model. Empirically, on a Wiki dataset with around 0.5 million labels, X-BERT achieves new state-of-the-art results where the precision@1 reaches 67:80%, a substantial improvement over 32.58%/60.91% of deep learning baseline fastText and competing XMC approach Parabel, respectively. This amounts to a 11.31% relative improvement over Parabel, which is indeed significant since the recent approach SLICE only has 5.53% relative improvement.
Language model pre-training has proven to be useful in learning universal language representations. As a state-of-the-art language model pre-training model, BERT (Bidirectional Encoder Representations from Transformers) has achieved amazing results in many language understanding tasks. In this paper, we conduct exhaustive experiments to investigate different fine-tuning methods of BERT on text classification task and provide a general solution for BERT fine-tuning. Finally, the proposed solution obtains new state-of-the-art results on eight widely-studied text classification datasets.
High spectral dimensionality and the shortage of annotations make hyperspectral image (HSI) classification a challenging problem. Recent studies suggest that convolutional neural networks can learn discriminative spatial features, which play a paramount role in HSI interpretation. However, most of these methods ignore the distinctive spectral-spatial characteristic of hyperspectral data. In addition, a large amount of unlabeled data remains an unexploited gold mine for efficient data use. Therefore, we proposed an integration of generative adversarial networks (GANs) and probabilistic graphical models for HSI classification. Specifically, we used a spectral-spatial generator and a discriminator to identify land cover categories of hyperspectral cubes. Moreover, to take advantage of a large amount of unlabeled data, we adopted a conditional random field to refine the preliminary classification results generated by GANs. Experimental results obtained using two commonly studied datasets demonstrate that the proposed framework achieved encouraging classification accuracy using a small number of data for training.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191