This paper provides insightful analysis into what decision processes are actually implemented by a trained recurrent network for sentiment classification, and uncover simple line attractor dynamics. All reviewers agree that this is interesting and illuminating, and that this work shows a good example of what can be done to open the black box of deep systems.

Originality: This is a very interesting algorithmic contribution. The introduced method gets state-of-the-art results under reasonable computation resources. I was reviewing a former version of this paper for some other conference and have to admit that the new version is significantly improved, mainly because the authors have succeeded to decrease the computational costs of the attention-based deep network by using the probabilistic label trees. Quality: The method is sound and the empirical analysis is of high quality. The paper does not have any theoretical contribution, but it is unnecessary for this kind of contribution.

The paper improves the SOTA in extreme classification achieving the difficult feat of outperforming one-vs-all techniques. The authors should follow the reviewers suggestions to improve the clarity of the paper, especially the description of the algorithm. They should also add a discussion as to why their technique is able to improve on the SOTA and provide the additional experimental results they included in the rebuttal.

Weaknesses: The biggest limitation of this work for me, is the experimental setup, specifically (1) the lack of comparison to existing models (2) poor results on text classification and speech act classification when compared to existing work and (3) the choice of benchmarks. I would recommend reporting results presented in previous work on POS tagging, speech act classification and text classification. This is particularly important since you run your own BERT baselines, it would be for the reader to know how these baselines compare with numbers reported in other papers. For example, [1] reports results on 20Newsgroups and [2,3] on the switchboard dialog act classification dataset and [4,5] on POS tagging. For example [1] reports 86.8% accuracy on 20newsgroups while you report only 32.21% for BERT and 51.01 for BERT Prism.

This paper studies a text classification algorithm based on an improved Transformer to improve the performance and efficiency of the model in text classification tasks. Aiming at the shortcomings of the traditional Transformer model in capturing deep semantic relationships and optimizing computational complexity, this paper introduces a multi-level attention mechanism and a contrastive learning strategy. The multi-level attention mechanism effectively models the global semantics and local features in the text by combining global attention with local attention; the contrastive learning strategy enhances the model's ability to distinguish between different categories by constructing positive and negative sample pairs while improving the classification effect. In addition, in order to improve the training and inference efficiency of the model on large-scale text data, this paper designs a lightweight module to optimize the feature transformation process and reduce the computational cost. Experimental results on the dataset show that the improved Transformer model outperforms the comparative models such as BiLSTM, CNN, standard Transformer, and BERT in terms of classification accuracy, F1 score, and recall rate, showing stronger semantic representation ability and generalization performance. The method proposed in this paper provides a new idea for algorithm optimization in the field of text classification and has good application potential and practical value. Future work will focus on studying the performance of this model in multi-category imbalanced datasets and cross-domain tasks and explore the integration wi

While the ImageNet dataset has been driving computer vision research over the past decade, significant label noise and ambiguity have made top-1 accuracy an insufficient measure of further progress. To address this, new label-sets and evaluation protocols have been proposed for ImageNet showing that state-of-the-art models already achieve over 95% accuracy and shifting the focus on investigating why the remaining errors persist.Recent work in this direction employed a panel of experts to manually categorize all remaining classification errors for two selected models. However, this process is time-consuming, prone to inconsistencies, and requires trained experts, making it unsuitable for regular model evaluation thus limiting its utility. To overcome these limitations, we propose the first automated error classification framework, a valuable tool to study how modeling choices affect error distributions. We use our framework to comprehensively evaluate the error distribution of over 900 models.

This paper presents a text classification algorithm inspired by the notion of superposition of states in quantum physics. By regarding text as a superposition of words, we derive the wave function of a document and we compute the transition probability of the document to a target class according to Born's rule. Two complementary implementations are presented. In the first one, wave functions are calculated explicitly. Through analysis of three benchmark datasets, we illustrate several aspects of the proposed method, such as classification performance, explainability, and computational efficiency.

In this paper we propose a general framework for learning distributed representations of attributes: characteristics of text whose representations can be jointly learned with word embeddings. Attributes can correspond to a wide variety of concepts, such as document indicators (to learn sentence vectors), language indicators (to learn distributed language representations), meta-data and side information (such as the age, gender and industry of a blogger) or representations of authors. We describe a third-order model where word context and attribute vectors interact multiplicatively to predict the next word in a sequence. This leads to the notion of conditional word similarity: how meanings of words change when conditioned on different attributes. We perform several experimental tasks including sentiment classification, cross-lingual document classification, and blog authorship attribution.

We propose a novel lightweight generative adversarial network for efficient image manipulation using natural language descriptions. To achieve this, a new word-level discriminator is proposed, which provides the generator with fine-grained training feedback at word-level, to facilitate training a lightweight generator that has a small number of parameters, but can still correctly focus on specific visual attributes of an image, and then edit them without affecting other contents that are not described in the text. Furthermore, thanks to the explicit training signal related to each word, the discriminator can also be simplified to have a lightweight structure. Compared with the state of the art, our method has a much smaller number of parameters, but still achieves a competitive manipulation performance. Extensive experimental results demonstrate that our method can better disentangle different visual attributes, then correctly map them to corresponding semantic words, and thus achieve a more accurate image modification using natural language descriptions.

Recent success of pre-trained language models crucially hinges on fine-tuning them on large amounts of labeled data for the downstream task, that are typically expensive to acquire or difficult to access for many applications. We study self-training as one of the earliest semi-supervised learning approaches to reduce the annotation bottleneck by making use of large-scale unlabeled data for the target task. Standard self-training mechanism randomly samples instances from the unlabeled pool to generate pseudo-labels and augment labeled data. We propose an approach to improve self-training by incorporating uncertainty estimates of the underlying neural network leveraging recent advances in Bayesian deep learning. Specifically, we propose (i) acquisition functions to select instances from the unlabeled pool leveraging Monte Carlo (MC) Dropout, and (ii) learning mechanism leveraging model confidence for self-training.