Recently deeplearning models have been shown to be capable of making remarkable performance in sentences and documents classification tasks. In this work, we propose a novel framework called AC-BLSTM for modeling sentences and documents, which combines the asymmetric convolution neural network (ACNN) with the Bidirectional Long Short-Term Memory network (BLSTM). Experiment results demonstrate that our model achieves state-of-the-art results on five tasks, including sentiment analysis, question type classification, and subjectivity classification. In order to further improve the performance of AC-BLSTM, we propose a semi-supervised learning framework called G-AC-BLSTM for text classification by combining the generative model with AC-BLSTM.
The skip-thought model has been proven to be effective at learning sentence representations and capturing sentence semantics. In this paper, we propose a suite of techniques to trim and improve it. First, we validate a hypothesis that, given a current sentence, inferring the previous and inferring the next sentence provide similar supervision power, therefore only one decoder for predicting the next sentence is preserved in our trimmed skip-thought model. Second, we present a connection layer between encoder and decoder to help the model to generalize better on semantic relatedness tasks. Third, we found that a good word embedding initialization is also essential for learning better sentence representations. We train our model unsupervised on a large corpus with contiguous sentences, and then evaluate the trained model on 7 supervised tasks, which includes semantic relatedness, paraphrase detection, and text classification benchmarks. We empirically show that, our proposed model is a faster, lighter-weight and equally powerful alternative to the original skip-thought model.
This paper presents the analysis of the impact of a floating-point number precision reduction on the quality of text classification. The precision reduction of the vectors representing the data (e.g. TF-IDF representation in our case) allows for a decrease of computing time and memory footprint on dedicated hardware platforms. The impact of precision reduction on the classification quality was performed on 5 corpora, using 4 different classifiers. Also, dimensionality reduction was taken into account. Results indicate that the precision reduction improves classification accuracy for most cases (up to 25% of error reduction). In general, the reduction from 64 to 4 bits gives the best scores and ensures that the results will not be worse than with the full floating-point representation.
Electronic health records (EHRs) contain important clinical information about patients. Efficient and effective use of this information could supplement or even replace manual chart review as a means of studying and improving the quality and safety of healthcare delivery. However, some of these clinical data are in the form of free text and require pre-processing before use in automated systems. A common free text data source is radiology reports, typically dictated by radiologists to explain their interpretations. We sought to demonstrate machine learning classification of computed tomography (CT) imaging reports into binary outcomes, i.e. positive and negative for fracture, using regular text classification and classifiers based on topic modeling. Topic modeling provides interpretable themes (topic distributions) in reports, a representation that is more compact than the commonly used bag-of-words representation and can be processed faster than raw text in subsequent automated processes. We demonstrate new classifiers based on this topic modeling representation of the reports. Aggregate topic classifier (ATC) and confidence-based topic classifier (CTC) use a single topic that is determined from the training dataset based on different measures to classify the reports on the test dataset. Alternatively, similarity-based topic classifier (STC) measures the similarity between the reports' topic distributions to determine the predicted class. Our proposed topic modeling-based classifier systems are shown to be competitive with existing text classification techniques and provides an efficient and interpretable representation.
Text preprocessing is often the first step in the pipeline of a Natural Language Processing (NLP) system, with potential impact in its final performance. Despite its importance, text preprocessing has not received much attention in the deep learning literature. In this paper we investigate the impact of simple text preprocessing decisions (particularly tokenizing, lemmatizing, lowercasing and multiword grouping) on the performance of a standard neural text classifier. We perform an extensive evaluation on standard benchmarks from text categorization and sentiment analysis. While our experiments show that a simple tokenization of input text is generally adequate, they also highlight significant degrees of variability across preprocessing techniques. This reveals the importance of paying attention to this usually-overlooked step in the pipeline, particularly when comparing different models. Finally, our evaluation provides insights into the best preprocessing practices for training word embeddings.
While natural languages are compositional, how state-of-the-art neural models achieve compositionality is still unclear. We propose a deep network, which not only achieves competitive accuracy for text classification, but also exhibits compositional behavior. That is, while creating hierarchical representations of a piece of text, such as a sentence, the lower layers of the network distribute their layer-specific attention weights to individual words. In contrast, the higher layers compose meaningful phrases and clauses, whose lengths increase as the networks get deeper until fully composing the sentence.
Multi-task learning leverages potential correlations among related tasks to extract common features and yield performance gains. However, most previous works only consider simple or weak interactions, thereby failing to model complex correlations among three or more tasks. In this paper, we propose a multi-task learning architecture with four types of recurrent neural layers to fuse information across multiple related tasks. The architecture is structurally flexible and considers various interactions among tasks, which can be regarded as a generalized case of many previous works. Extensive experiments on five benchmark datasets for text classification show that our model can significantly improve performances of related tasks with additional information from others.
We study in this work the importance of depth in convolutional models for text classification, either when character or word inputs are considered. We show on 5 standard text classification and sentiment analysis tasks that deep models indeed give better performances than shallow networks when the text input is represented as a sequence of characters. However, a simple shallow-and-wide network outperforms deep models such as DenseNet with word inputs. Our shallow word model further establishes new state-of-the-art performances on two datasets: Yelp Binary (95.9\%) and Yelp Full (64.9\%).
In this paper, we propose a novel approach for text classification based on clustering word embeddings, inspired by the bag of visual words model, which is widely used in computer vision. After each word in a collection of documents is represented as word vector using a pre-trained word embeddings model, a k-means algorithm is applied on the word vectors in order to obtain a fixed-size set of clusters. The centroid of each cluster is interpreted as a super word embedding that embodies all the semantically related word vectors in a certain region of the embedding space. Every embedded word in the collection of documents is then assigned to the nearest cluster centroid. In the end, each document is represented as a bag of super word embeddings by computing the frequency of each super word embedding in the respective document. We also diverge from the idea of building a single vocabulary for the entire collection of documents, and propose to build class-specific vocabularies for better performance. Using this kind of representation, we report results on two text mining tasks, namely text categorization by topic and polarity classification. On both tasks, our model yields better performance than the standard bag of words.
There have been some works that learn a lexicon together with the corpus to improve the word embeddings. However, they either model the lexicon separately but update the neural networks for both the corpus and the lexicon by the same likelihood, or minimize the distance between all of the synonym pairs in the lexicon. Such methods do not consider the relatedness and difference of the corpus and the lexicon, and may not be the best optimized. In this paper, we propose a novel method that considers the relatedness and difference of the corpus and the lexicon. It trains word embeddings by learning the corpus to predicate a word and its corresponding synonym under the context at the same time. For polysemous words, we use a word sense disambiguation filter to eliminate the synonyms that have different meanings for the context. To evaluate the proposed method, we compare the performance of the word embeddings trained by our proposed model, the control groups without the filter or the lexicon, and the prior works in the word similarity tasks and text classification task. The experimental results show that the proposed model provides better embeddings for polysemous words and improves the performance for text classification.
One of the major hurdles preventing the full exploitation of information from online communities is the widespread concern regarding the quality and credibility of user-contributed content. Prior works in this domain operate on a static snapshot of the community, making strong assumptions about the structure of the data (e.g., relational tables), or consider only shallow features for text classification. To address the above limitations, we propose probabilistic graphical models that can leverage the joint interplay between multiple factors in online communities --- like user interactions, community dynamics, and textual content --- to automatically assess the credibility of user-contributed online content, and the expertise of users and their evolution with user-interpretable explanation. To this end, we devise new models based on Conditional Random Fields for different settings like incorporating partial expert knowledge for semi-supervised learning, and handling discrete labels as well as numeric ratings for fine-grained analysis. This enables applications such as extracting reliable side-effects of drugs from user-contributed posts in healthforums, and identifying credible content in news communities. Online communities are dynamic, as users join and leave, adapt to evolving trends, and mature over time. To capture this dynamics, we propose generative models based on Hidden Markov Model, Latent Dirichlet Allocation, and Brownian Motion to trace the continuous evolution of user expertise and their language model over time. This allows us to identify expert users and credible content jointly over time, improving state-of-the-art recommender systems by explicitly considering the maturity of users. This also enables applications such as identifying helpful product reviews, and detecting fake and anomalous reviews with limited information.
Computer vision has benefited from initializing multiple deep layers with weights pretrained on large supervised training sets like ImageNet. Natural language processing (NLP) typically sees initialization of only the lowest layer of deep models with pretrained word vectors. In this paper, we use a deep LSTM encoder from an attentional sequence-to-sequence model trained for machine translation (MT) to contextualize word vectors. We show that adding these context vectors (CoVe) improves performance over using only unsupervised word and character vectors on a wide variety of common NLP tasks: sentiment analysis (SST, IMDb), question classification (TREC), entailment (SNLI), and question answering (SQuAD). For fine-grained sentiment analysis and entailment, CoVe improves performance of our baseline models to the state of the art.
In this paper, we investigate the application of text classification methods to predict the law area and the decision of cases judged by the French Supreme Court. We also investigate the influence of the time period in which a ruling was made over the textual form of the case description and the extent to which it is necessary to mask the judge's motivation for a ruling to emulate a real-world test scenario. We report results of 96% f1 score in predicting a case ruling, 90% f1 score in predicting the law area of a case, and 75.9% f1 score in estimating the time span when a ruling has been issued using a linear Support Vector Machine (SVM) classifier trained on lexical features.
Deep neural networks have become ubiquitous for applications related to visual recognition and language understanding tasks. However, it is often prohibitive to use typical neural networks on devices like mobile phones or smart watches since the model sizes are huge and cannot fit in the limited memory available on such devices. While these devices could make use of machine learning models running on high-performance data centers with CPUs or GPUs, this is not feasible for many applications because data can be privacy sensitive and inference needs to be performed directly "on" device. We introduce a new architecture for training compact neural networks using a joint optimization framework. At its core lies a novel objective that jointly trains using two different types of networks--a full trainer neural network (using existing architectures like Feed-forward NNs or LSTM RNNs) combined with a simpler "projection" network that leverages random projections to transform inputs or intermediate representations into bits. The simpler network encodes lightweight and efficient-to-compute operations in bit space with a low memory footprint. The two networks are trained jointly using backpropagation, where the projection network learns from the full network similar to apprenticeship learning. Once trained, the smaller network can be used directly for inference at low memory and computation cost. We demonstrate the effectiveness of the new approach at significantly shrinking the memory requirements of different types of neural networks while preserving good accuracy on visual recognition and text classification tasks. We also study the question "how many neural bits are required to solve a given task?" using the new framework and show empirical results contrasting model predictive capacity (in bits) versus accuracy on several datasets.
This article offers an empirical study on the different ways of encoding Chinese, Japanese, Korean (CJK) and English languages for text classification. Different encoding levels are studied, including UTF-8 bytes, characters, words, romanized characters and romanized words. For all encoding levels, whenever applicable, we provide comparisons with linear models, fastText and convolutional networks. For convolutional networks, we compare between encoding mechanisms using character glyph images, one-hot (or one-of-n) encoding, and embedding. In total there are 473 models, using 14 large-scale text classification datasets in 4 languages including Chinese, English, Japanese and Korean. Some conclusions from these results include that byte-level one-hot encoding based on UTF-8 consistently produces competitive results for convolutional networks, that word-level n-grams linear models are competitive even without perfect word segmentation, and that fastText provides the best result using character-level n-gram encoding but can overfit when the features are overly rich.
In this era of digitization, knowing the user's sociolect aspects have become essential features to build the user specific recommendation systems. These sociolect aspects could be found by mining the user's language sharing in the form of text in social media and reviews. This paper describes about the experiment that was performed in PAN Author Profiling 2017 shared task. The objective of the task is to find the sociolect aspects of the users from their tweets. The sociolect aspects considered in this experiment are user's gender and native language information. Here user's tweets written in a different language from their native language are represented as Document - Term Matrix with document frequency as the constraint. Further classification is done using the Support Vector Machine by taking gender and native language as target classes. This experiment attains the average accuracy of 73.42% in gender prediction and 76.26% in the native language identification task.
Social media datasets, especially Twitter tweets, are popular in the field of text classification. Tweets are a valuable source of micro-text (sometimes referred to as "micro-blogs"), and have been studied in domains such as sentiment analysis, recommendation systems, spam detection, clustering, among others. Tweets often include keywords referred to as "Hashtags" that can be used as labels for the tweet. Using tweets encompassing 50 labels, we studied the impact of word versus character-level feature selection and extraction on different learners to solve a multi-class classification task. We show that feature extraction of simple character-level groups performs better than simple word groups and pre-processing methods like normalizing using Porter's Stemming and Part-of-Speech ("POS")-Lemmatization.
In recent years, many deep-learning based models are proposed for text classification. This kind of models well fits the training set from the statistical point of view. However, it lacks the capacity of utilizing instance-level information from individual instances in the training set. In this work, we propose to enhance neural network models by allowing them to leverage information from $k$-nearest neighbor (kNN) of the input text. Our model employs a neural network that encodes texts into text embeddings. Moreover, we also utilize $k$-nearest neighbor of the input text as an external memory, and utilize it to capture instance-level information from the training set. The final prediction is made based on features from both the neural network encoder and the kNN memory. Experimental results on several standard benchmark datasets show that our model outperforms the baseline model on all the datasets, and it even beats a very deep neural network model (with 29 layers) in several datasets. Our model also shows superior performance when training instances are scarce, and when the training set is severely unbalanced. Our model also leverages techniques such as semi-supervised training and transfer learning quite well.
Effectively making sense of short texts is a critical task for many real world applications such as search engines, social media services, and recommender systems. The task is particularly challenging as a short text contains very sparse information, often too sparse for a machine learning algorithm to pick up useful signals. A common practice for analyzing short text is to first expand it with external information, which is usually harvested from a large collection of longer texts. In literature, short text expansion has been done with all kinds of heuristics. We propose an end-to-end solution that automatically learns how to expand short text to optimize a given learning task. A novel deep memory network is proposed to automatically find relevant information from a collection of longer documents and reformulate the short text through a gating mechanism. Using short text classification as a demonstrating task, we show that the deep memory network significantly outperforms classical text expansion methods with comprehensive experiments on real world data sets.
Various text analysis techniques exist, which attempt to uncover unstructured information from text. In this work, we explore using statistical dependence measures for textual classification, representing text as word vectors. Student satisfaction scores on a 3-point scale and their free text comments written about university subjects are used as the dataset. We have compared two textual representations: a frequency word representation and term frequency relationship to word vectors, and found that word vectors provide a greater accuracy. However, these word vectors have a large number of features which aggravates the burden of computational complexity. Thus, we explored using a non-linear dependency measure for feature selection by maximizing the dependence between the text reviews and corresponding scores. Our quantitative and qualitative analysis on a student satisfaction dataset shows that our approach achieves comparable accuracy to the full feature vector, while being an order of magnitude faster in testing. These text analysis and feature reduction techniques can be used for other textual data applications such as sentiment analysis.