Recent advances in deep learning have facilitated the demand of neural models for real applications. In practice, these applications often need to be deployed with limited resources while keeping high accuracy. This paper touches the core of neural models in NLP, word embeddings, and presents a new embedding distillation framework that remarkably reduces the dimension of word embeddings without compromising accuracy. A novel distillation ensemble approach is also proposed that trains a high-efficient student model using multiple teacher models. In our approach, the teacher models play roles only during training such that the student model operates on its own without getting supports from the teacher models during decoding, which makes it eighty times faster and lighter than other typical ensemble methods. All models are evaluated on seven document classification datasets and show a significant advantage over the teacher models for most cases. Our analysis depicts insightful transformation of word embeddings from distillation and suggests a future direction to ensemble approaches using neural models.
Attention networks show promise for both vision and language tasks, by emphasizing relationships between constituent elements through weighting functions. Such elements could be regions in an image output by a region proposal network, or words in a sentence, represented by word embedding. Thus far the learning of attention weights has been driven solely by the minimization of task specific loss functions. We introduce a method for learning attention weights to better emphasize informative pair-wise relations between entities. The key component is a novel center-mass cross entropy loss, which can be applied in conjunction with the task specific ones. We further introduce a focused attention backbone to learn these attention weights for general tasks. We demonstrate that the focused supervision leads to improved attention distribution across meaningful entities, and that it enhances the representation by aggregating features from them. Our focused attention module leads to state-of-the-art recovery of relations in a relationship proposal task and boosts performance for various vision and language tasks.
Pretrained contextual representation models (Peters et al., 2018; Devlin et al., 2018) have pushed forward the state-of-the-art on many NLP tasks. A new release of BERT (Devlin, 2018) includes a model simultaneously pretrained on 104 languages with impressive performance for zero-shot cross-lingual transfer on a natural language inference task. This paper explores the broader cross-lingual potential of mBERT (multilingual) as a zero shot language transfer model on 5 NLP tasks covering a total of 39 languages from various language families: NLI, document classification, NER, POS tagging, and dependency parsing. We compare mBERT with the best-published methods for zero-shot cross-lingual transfer and find mBERT competitive on each task. Additionally, we investigate the most effective strategy for utilizing mBERT in this manner, determine to what extent mBERT generalizes away from language specific features, and measure factors that influence cross-lingual transfer.
With the development of information technology, we have witnessed an age of data explosion which produces a large variety of data filled with redundant information. Because dimension reduction is an essential tool which embeds high-dimensional data into a lower-dimensional subspace to avoid redundant information, it has attracted interests from researchers all over the world. However, facing with features from multiple views, it's difficult for most dimension reduction methods to fully comprehended multi-view features and integrate compatible and complementary information from these features to construct low-dimensional subspace directly. Furthermore, most multi-view dimension reduction methods cannot handle features from nonlinear spaces with high dimensions. Therefore, how to construct a multi-view dimension reduction methods which can deal with multi-view features from high-dimensional nonlinear space is of vital importance but challenging. In order to address this problem, we proposed a novel method named Co-regularized Multi-view Sparse Reconstruction Embedding (CMSRE) in this paper. By exploiting correlations of sparse reconstruction from multiple views, CMSRE is able to learn local sparse structures of nonlinear manifolds from multiple views and constructs significative low-dimensional representations for them. Due to the proposed co-regularized scheme, correlations of sparse reconstructions from multiple views are preserved by CMSRE as much as possible. Furthermore, sparse representation produces more meaningful correlations between features from each single view, which helps CMSRE to gain better performances. Various evaluations based on the applications of document classification, face recognition and image retrieval can demonstrate the effectiveness of the proposed approach on multi-view dimension reduction.
We present, to our knowledge, the first application of BERT to document classification. A few characteristics of the task might lead one to think that BERT is not the most appropriate model: syntactic structures matter less for content categories, documents can often be longer than typical BERT input, and documents often have multiple labels. Nevertheless, we show that a straightforward classification model using BERT is able to achieve the state of the art across four popular datasets. To address the computational expense associated with BERT inference, we distill knowledge from BERT-large to small bidirectional LSTMs, reaching BERT-base parity on multiple datasets using 30x fewer parameters. The primary contribution of our paper is improved baselines that can provide the foundation for future work.
This paper presents a novel adaptively connected neural network (ACNet) to improve the traditional convolutional neural networks (CNNs) {in} two aspects. First, ACNet employs a flexible way to switch global and local inference in processing the internal feature representations by adaptively determining the connection status among the feature nodes (e.g., pixels of the feature maps) \footnote{In a computer vision domain, a node refers to a pixel of a feature map{, while} in {the} graph domain, a node denotes a graph node.}. We can show that existing CNNs, the classical multilayer perceptron (MLP), and the recently proposed non-local network (NLN) \cite{nonlocalnn17} are all special cases of ACNet. Second, ACNet is also capable of handling non-Euclidean data. Extensive experimental analyses on {a variety of benchmarks (i.e.,} ImageNet-1k classification, COCO 2017 detection and segmentation, CUHK03 person re-identification, CIFAR analysis, and Cora document categorization) demonstrate that {ACNet} cannot only achieve state-of-the-art performance but also overcome the limitation of the conventional MLP and CNN \footnote{Corresponding author: Liang Lin ([email omitted])}. The code is available at \url{https://github.com/wanggrun/Adaptively-Connected-Neural-Networks}.
In today's legal environment, lawsuits and regulatory investigations require companies to embark upon increasingly intensive data-focused engagements to identify, collect and analyze large quantities of data. When documents are staged for review the process can require companies to dedicate an extraordinary level of resources, both with respect to human resources, but also with respect to the use of technology-based techniques to intelligently sift through data. For several years, attorneys have been using a variety of tools to conduct this exercise, and most recently, they are accepting the use of machine learning techniques like text classification to efficiently cull massive volumes of data to identify responsive documents for use in these matters. In recent years, a group of AI and Machine Learning researchers have been actively researching Explainable AI. In an explainable AI system, actions or decisions are human understandable. In typical legal `document review' scenarios, a document can be identified as responsive, as long as one or more of the text snippets in a document are deemed responsive. In these scenarios, if predictive coding can be used to locate these responsive snippets, then attorneys could easily evaluate the model's document classification decision. When deployed with defined and explainable results, predictive coding can drastically enhance the overall quality and speed of the document review process by reducing the time it takes to review documents. The authors of this paper propose the concept of explainable predictive coding and simple explainable predictive coding methods to locate responsive snippets within responsive documents. We also report our preliminary experimental results using the data from an actual legal matter that entailed this type of document review.
Over the last few years, machine learning over graph structures has manifested a significant enhancement in text mining applications such as event detection, opinion mining, and news recommendation. One of the primary challenges in this regard is structuring a graph that encodes and encompasses the features of textual data for the effective machine learning algorithm. Besides, exploration and exploiting of semantic relations is regarded as a principal step in text mining applications. However, most of the traditional text mining methods perform somewhat poor in terms of employing such relations. In this paper, we propose a sentence-level graph-based text representation which includes stop words to consider semantic and term relations. Then, we employ a representation learning approach on the combined graphs of sentences to extract the latent and continuous features of the documents. Eventually, the learned features of the documents are fed into a deep neural network for the sentiment classification task. The experimental results demonstrate that the proposed method substantially outperforms the related sentiment analysis approaches based on several benchmark datasets. Furthermore, our method can be generalized on different datasets without any dependency on pre-trained word embeddings.
We propose a method to create document representations that reflect their internal structure. We modify Tree-LSTMs to hierarchically merge basic elements such as words and sentences into blocks of increasing complexity. Our Structure Tree-LSTM implements a hierarchical attention mechanism over individual components and combinations thereof. We thus emphasize the usefulness of Tree-LSTMs for texts larger than a sentence. We show that structure-aware encoders can be used to improve the performance of document classification. We demonstrate that our method is resilient to changes to the basic building blocks, as it performs well with both sentence and word embeddings. The Structure Tree-LSTM outperforms all the baselines on two datasets by leveraging structural clues. We show our model's interpretability by visualizing how our model distributes attention inside a document. On a third dataset from the medical domain, our model achieves competitive performance with the state of the art. This result shows the Structure Tree-LSTM can leverage dependency relations other than text structure, such as a set of reports on the same patient.
It is intuitive that NLP tasks for logographic languages like Chinese should benefit from the use of the glyph information in those languages. However, due to the lack of rich pictographic evidence in glyphs and the weak generalization ability of standard computer vision models on character data, an effective way to utilize the glyph information remains to be found. In this paper, we address this gap by presenting Glyce, the glyph-vectors for Chinese character representations. We make three major innovations: (1) We use historical Chinese scripts (e.g., bronzeware script, seal script, traditional Chinese, etc) to enrich the pictographic evidence in characters; (2) We design CNN structures (called tianzege-CNN) tailored to Chinese character image processing; and (3) We use image-classification as an auxiliary task in a multi-task learning setup to increase the model's ability to generalize. We show that glyph-based models are able to consistently outperform word/char ID-based models in a wide range of Chinese NLP tasks. We are able to set new state-of-the-art results for a variety of Chinese NLP tasks, including tagging (NER, CWS, POS), sentence pair classification, single sentence classification tasks, dependency parsing, and semantic role labeling. For example, the proposed model achieves an F1 score of 80.6 on the OntoNotes dataset of NER, +1.5 over BERT; it achieves an almost perfect accuracy of 99.8\% on the Fudan corpus for text classification. Code found at https://github.com/ShannonAI/glyce.
Document classification is a challenging task with important applications. The deep learning approaches to the problem have gained much attention recently. Despite the progress, the proposed models do not incorporate the knowledge of the document structure in the architecture efficiently and not take into account the contexting importance of words and sentences. In this paper, we propose a new approach based on a combination of convolutional neural networks, gated recurrent units, and attention mechanisms for document classification tasks. The main contribution of this work is the use of convolution layers to extract more meaningful, generalizable and abstract features by the hierarchical representation. The proposed method in this paper improves the results of the current attention-based approaches for document classification.
This paper deals with the automatic analysis of conversations between a customer and an agent in a call centre of a customer care service. The purpose of the analysis is to hypothesize themes about problems and complaints discussed in the conversation. Themes are defined by the application documentation topics. A conversation may contain mentions that are irrelevant for the application purpose and multiple themes whose mentions may be interleaved portions of a conversation that cannot be well defined. Two methods are proposed for multiple theme hypothesization. One of them is based on a cosine similarity measure using a bag of features extracted from the entire conversation. The other method introduces the concept of thematic density distributed around specific word positions in a conversation. In addition to automatically selected words, word bi-grams with possible gaps between successive words are also considered and selected. Experimental results show that the results obtained with the proposed methods outperform the results obtained with support vector machines on the same data. Furthermore, using the theme skeleton of a conversation from which thematic densities are derived, it will be possible to extract components of an automatic conversation report to be used for improving the service performance. Index Terms: multi-topic audio document classification, hu-man/human conversation analysis, speech analytics, distance bigrams
Text classification must sometimes be applied in a low-resource language with no labeled training data. However, training data may be available in a related language. We investigate whether character-level knowledge transfer from a related language helps text classification. We present a cross-lingual document classification framework (CACO) that exploits cross-lingual subword similarity by jointly training a character-based embedder and a word-based classifier. The embedder derives vector representations for input words from their written forms, and the classifier makes predictions based on the word vectors. We use a joint character representation for both the source language and the target language, which allows the embedder to generalize knowledge about source language words to target language words with similar forms. We propose a multi-task objective that can further improve the model if additional cross-lingual or monolingual resources are available. Experiments confirm that character-level knowledge transfer is more data-efficient than word-level transfer between related languages.
We introduce an architecture to learn joint multilingual sentence representations for 93 languages, belonging to more than 30 different families and written in 28 different scripts. Our system uses a single BiLSTM encoder with a shared BPE vocabulary for all languages, which is coupled with an auxiliary decoder and trained on publicly available parallel corpora. This enables us to learn a classifier on top of the resulting embeddings using English annotated data only, and transfer it to any of the 93 languages without any modification. Our experiments in cross-lingual natural language inference (XNLI dataset), cross-lingual document classification (MLDoc dataset) and parallel corpus mining (BUCC dataset) show the effectiveness of our approach. We also introduce a new test set of aligned sentences in 112 languages, and show that our sentence embeddings obtain strong results in multilingual similarity search even for low-resource languages. Our implementation, the pre-trained encoder and the multilingual test set are available at https://github.com/facebookresearch/LASER
In this paper, a novel approach is proposed for the recognition of Persian phonemes in the Persian Consonant-Vowel Combination (PCVC) speech dataset. Nowadays, deep neural networks play a crucial role in classification tasks. However, the best results in speech recognition are not yet as perfect as human recognition rate. Deep learning techniques show outstanding performance over many other classification tasks like image classification, document classification, etc. Furthermore, the performance is sometimes better than a human. The reason why automatic speech recognition (ASR) systems are not as qualified as the human speech recognition system, mostly depends on features of data which is fed to deep neural networks. Methods: In this research, firstly, the sound samples are cut for the exact extraction of phoneme sounds in 50ms samples. Then, phonemes are divided into 30 groups, containing 23 consonants, 6 vowels, and a silence phoneme. Results: The short-time Fourier transform (STFT) is conducted on them, and the results are given to PPNet (A new deep convolutional neural network architecture) classifier and a total average of 75.87% accuracy is reached which is the best result ever compared to other algorithms on separated Persian phonemes (Like in PCVC speech dataset). Conclusion: This method can be used not only for recognizing mono-phonemes but also it can be adopted as an input to the selection of the best words in speech transcription.
Computerized document classification already orders the news articles that Apple's "News" app or Google's "personalized search" feature groups together to match a reader's interests. The invisible and therefore illegible decisions that go into these tailored searches have been the subject of a critique by scholars who emphasize that our intelligence about documents is only as good as our ability to understand the criteria of search. This article will attempt to unpack the procedures used in computational classification of texts, translating them into term legible to humanists, and examining opportunities to render the computational text classification process subject to expert critique and improvement.
Reviewing radiology reports in emergency departments is an essential but laborious task. Timely follow-up of patients with abnormal cases in their radiology reports may dramatically affect the patient's outcome, especially if they have been discharged with a different initial diagnosis. Machine learning approaches have been devised to expedite the process and detect the cases that demand instant follow up. However, these approaches require a large amount of labeled data to train reliable predictive models. Preparing such a large dataset, which needs to be manually annotated by health professionals, is costly and time-consuming. This paper investigates a semi-supervised learning framework for radiology report classification across three hospitals. The main goal is to leverage clinical unlabeled data in order to augment the learning process where limited labeled data is available. To further improve the classification performance, we also integrate a transfer learning technique into the semi-supervised learning pipeline . Our experimental findings show that (1) convolutional neural networks (CNNs), while being independent of any problem-specific feature engineering, achieve significantly higher effectiveness compared to conventional supervised learning approaches, (2) leveraging unlabeled data in training a CNN-based classifier reduces the dependency on labeled data by more than 50% to reach the same performance of a fully supervised CNN, and (3) transferring the knowledge gained from available labeled data in an external source hospital significantly improves the performance of a semi-supervised CNN model over their fully supervised counterparts in a target hospital.
The Brazilian court system is currently the most clogged up judiciary system in the world. Thousands of lawsuit cases reach the supreme court every day. These cases need to be analyzed in order to be associated to relevant tags and allocated to the right team. Most of the cases reach the court as raster scanned documents with widely variable levels of quality. One of the first steps for the analysis is to classify these documents. In this paper we present a Bidirectional Long Short-Term Memory network (Bi-LSTM) to classify these pieces of legal document.
The abundance of text data being produced in the modern age makes it increasingly important to intuitively group, categorize, or classify text data by theme for efficient retrieval and search. Yet, the high dimensionality and imprecision of text data, or more generally language as a whole, prove to be challenging when attempting to perform unsupervised document clustering. In this thesis, we present two novel methods for improving unsupervised document clustering/classification by theme. The first is to improve document representations. We look to exploit "term neighborhoods" and "blur" semantic weight across neighboring terms. These neighborhoods are located in the semantic space afforded by "word embeddings." The second method is for cluster revision, based on what we deem as "stochastic barcoding", or "S- Barcode" patterns. Text data is inherently high dimensional, yet clustering typically takes place in a low dimensional representation space. Our method utilizes lower dimension clustering results as initial cluster configurations, and iteratively revises the configuration in the high dimensional space. We show with experimental results how both of the two methods improve the quality of document clustering. While this thesis elaborates on the two new conceptual contributions, a joint thesis by David Yan details the feature transformation and software architecture we developed for unsupervised document classification.
Fake news are nowadays an issue of pressing concern, given their recent rise as a potential threat to high-quality journalism and well-informed public discourse. The Fake News Challenge (FNC-1) was organized in 2017 to encourage the development of machine learning-based classification systems for stance detection (i.e., for identifying whether a particular news article agrees, disagrees, discusses, or is unrelated to a particular news headline), thus helping in the detection and analysis of possible instances of fake news. This article presents a new approach to tackle this stance detection problem, based on the combination of string similarity features with a deep neural architecture that leverages ideas previously advanced in the context of learning efficient text representations, document classification, and natural language inference. Specifically, we use bi-directional Recurrent Neural Networks, together with max-pooling over the temporal/sequential dimension and neural attention, for representing (i) the headline, (ii) the first two sentences of the news article, and (iii) the entire news article. These representations are then combined/compared, complemented with similarity features inspired on other FNC-1 approaches, and passed to a final layer that predicts the stance of the article towards the headline. We also explore the use of external sources of information, specifically large datasets of sentence pairs originally proposed for training and evaluating natural language inference methods, in order to pre-train specific components of the neural network architecture (e.g., the RNNs used for encoding sentences). The obtained results attest to the effectiveness of the proposed ideas and show that our model, particularly when considering pre-training and the combination of neural representations together with similarity features, slightly outperforms the previous state-of-the-art.