To measure the similarity of documents, the Wasserstein distance is a powerful tool, but it requires a high computational cost. Recently, for fast computation of the Wasserstein distance, methods for approximating the Wasserstein distance using a tree metric have been proposed. These tree-based methods allow fast comparisons of a large number of documents; however, they are unsupervised and do not learn task-specific distances. In this work, we propose the Supervised Tree-Wasserstein (STW) distance, a fast, supervised metric learning method based on the tree metric. Specifically, we rewrite the Wasserstein distance on the tree metric by the parent-child relationships of a tree and formulate it as a continuous optimization problem using a contrastive loss. Experimentally, we show that the STW distance can be computed fast, and improves the accuracy of document classification tasks. Furthermore, the STW distance is formulated by matrix multiplications, runs on a GPU, and is suitable for batch processing. Therefore, we show that the STW distance is extremely efficient when comparing a large number of documents.
This paper proposes a deep learning-based method to identify the segments of a clinical note corresponding to ICD-9 broad categories which are further color-coded with respect to 17 ICD-9 categories. The proposed Medical Segment Colorer (MSC) architecture is a pipeline framework that works in three stages: (1) word categorization, (2) phrase allocation, and (3) document classification. MSC uses gated recurrent unit neural networks (GRUs) to map from an input document to word multi-labels to phrase allocations, and uses statistical median to map phrase allocation to document multi-label. We compute variable length segment coloring from overlapping phrase allocation probabilities. These cross-level bidirectional contextual links identify adaptive context and then produce segment coloring. We train and evaluate MSC using the document labeled MIMIC-III clinical notes. Training is conducted solely using document multi-labels without any information on phrases, segments, or words. In addition to coloring a clinical note, MSC generates as byproducts document multi-labeling and word tagging -- creation of ICD9 category keyword lists based on segment coloring. Performance comparison of MSC byproduct document multi-labels versus methods whose purpose is to produce justifiable document multi-labels is 64% vs 52.4% micro-average F1-score against the CAML (CNN attention multi label) method. For evaluation of MSC segment coloring results, medical practitioners independently assigned the colors to broad ICD9 categories given a sample of 40 colored notes and a sample of 50 words related to each category based on the word tags. Binary scoring of this evaluation has a median value of 83.3% and mean of 63.7%.
Semi-supervised learning through deep generative models and multi-lingual pretraining techniques have orchestrated tremendous success across different areas of NLP. Nonetheless, their development has happened in isolation, while the combination of both could potentially be effective for tackling task-specific labelled data shortage. To bridge this gap, we combine semi-supervised deep generative models and multi-lingual pretraining to form a pipeline for document classification task. Compared to strong supervised learning baselines, our semi-supervised classification framework is highly competitive and outperforms the state-of-the-art counterparts in low-resource settings across several languages.
Optimal transport (OT) theory provides powerful tools to compare probability measures. However, OT is limited to nonnegative measures having the same mass, and suffers serious drawbacks about its computation and statistics. This leads to several proposals of regularized variants of OT in the recent literature. In this work, we consider an \textit{entropy partial transport} (EPT) problem for nonnegative measures on a tree having different masses. The EPT is shown to be equivalent to a standard complete OT problem on a one-node extended tree. We derive its dual formulation, then leverage this to propose a novel regularization for EPT which admits fast computation and negative definiteness. To our knowledge, the proposed regularized EPT is the first approach that yields a \textit{closed-form} solution among available variants of unbalanced OT. For practical applications without priori knowledge about the tree structure for measures, we propose tree-sliced variants of the regularized EPT, computed by averaging the regularized EPT between these measures using random tree metrics, built adaptively from support data points. Exploiting the negative definiteness of our regularized EPT, we introduce a positive definite kernel, and evaluate it against other baselines on benchmark tasks such as document classification with word embedding and topological data analysis. In addition, we empirically demonstrate that our regularization also provides effective approximations.
The Multi-Task Learning (MTL) technique has been widely studied by word-wide researchers. The majority of current MTL studies adopt the hard parameter sharing structure, where hard layers tend to learn general representations over all tasks and specific layers are prone to learn specific representations for each task. Since the specific layers directly follow the hard layers, the MTL model needs to estimate this direct change (from general to specific) as well. To alleviate this problem, we introduce the novel cluster layer, which groups tasks into clusters during training procedures. In a cluster layer, the tasks in the same cluster are further required to share the same network. By this way, the cluster layer produces the general presentation for the same cluster, while produces relatively specific presentations for different clusters. As transitions the cluster layers are used between the hard layers and the specific layers. The MTL model thus learns general representations to specific representations gradually. We evaluate our model with MTL document classification and the results demonstrate the cluster layer is quite efficient in MTL.
The relationship between words in a sentence often tells us more about the underlying semantic content of a document than its actual words, individually. In this work, we propose two novel algorithms, called Flexible Lexical Chain II and Fixed Lexical Chain II. These algorithms combine the semantic relations derived from lexical chains, prior knowledge from lexical databases, and the robustness of the distributional hypothesis in word embeddings as building blocks forming a single system. In short, our approach has three main contributions: (i) a set of techniques that fully integrate word embeddings and lexical chains; (ii) a more robust semantic representation that considers the latent relation between words in a document; and (iii) lightweight word embeddings models that can be extended to any natural language task. We intend to assess the knowledge of pre-trained models to evaluate their robustness in the document classification task. The proposed techniques are tested against seven word embeddings algorithms using five different machine learning classifiers over six scenarios in the document classification task. Our results show the integration between lexical chains and word embeddings representations sustain state-of-the-art results, even against more complex systems.
The mushroom body of the fruit fly brain is one of the best studied systems in neuroscience. At its core it consists of a population of Kenyon cells, which receive inputs from multiple sensory modalities. These cells are inhibited by the anterior paired lateral neuron, thus creating a sparse high dimensional representation of the inputs. In this work we study a mathematical formalization of this network motif and apply it to learning the correlational structure between words and their context in a corpus of unstructured text, a common natural language processing (NLP) task. We show that this network can learn semantic representations of words and can generate both static and context-dependent word embeddings. Unlike conventional methods (e.g., BERT, GloVe) that use dense representations for word embedding, our algorithm encodes semantic meaning of words and their context in the form of sparse binary hash codes. The quality of the learned representations is evaluated on word similarity analysis, word-sense disambiguation, and document classification. It is shown that not only can the fruit fly network motif achieve performance comparable to existing methods in NLP, but, additionally, it uses only a fraction of the computational resources (shorter training time and smaller memory footprint).
From small screenshots to large videos, documents take up a bulk of space in a modern smartphone. Documents in a phone can accumulate from various sources, and with the high storage capacity of mobiles, hundreds of documents are accumulated in a short period. However, searching or managing documents remains an onerous task, since most search methods depend on meta-information or only text in a document. In this paper, we showcase that a single modality is insufficient for classification and present a novel pipeline to classify documents on-device, thus preventing any private user data transfer to server. For this task, we integrate an open-source library for Optical Character Recognition (OCR) and our novel model architecture in the pipeline. We optimise the model for size, a necessary metric for on-device inference. We benchmark our classification model with a standard multimodal dataset FOOD-101 and showcase competitive results with the previous State of the Art with 30% model compression.
In this work, we introduce BanglaBERT, a BERT-based Natural Language Understanding (NLU) model pretrained in Bangla, a widely spoken yet low-resource language in the NLP literature. To pretrain BanglaBERT, we collect 27.5 GB of Bangla pretraining data (dubbed `Bangla2B+') by crawling 110 popular Bangla sites. We introduce two downstream task datasets on natural language inference and question answering and benchmark on four diverse NLU tasks covering text classification, sequence labeling, and span prediction. In the process, we bring them under the first-ever Bangla Language Understanding Benchmark (BLUB). BanglaBERT achieves state-of-the-art results outperforming multilingual and monolingual models. We are making the models, datasets, and a leaderboard publicly available at https://github.com/csebuetnlp/banglabert to advance Bangla NLP.
Objective: Systematic reviews of scholarly documents often provide complete and exhaustive summaries of literature relevant to a research question. However, well-done systematic reviews are expensive, time-demanding, and labor-intensive. Here, we propose an automatic document classification approach to significantly reduce the effort in reviewing documents. Methods: We first describe a manual document classification procedure that is used to curate a pertinent training dataset and then propose three classifiers: a keyword-guided method, a cluster analysis-based refined method, and a random forest approach that utilizes a large set of feature tokens. As an example, this approach is used to identify documents studying female sex workers that are assumed to contain content relevant to either HIV or violence. We compare the performance of the three classifiers by cross-validation and conduct a sensitivity analysis on the portion of data utilized in training the model. Results: The random forest approach provides the highest area under the curve (AUC) for both receiver operating characteristic (ROC) and precision/recall (PR). Analyses of precision and recall suggest that random forest could facilitate manually reviewing 20\% of the articles while containing 80\% of the relevant cases. Finally, we found a good classifier could be obtained by using a relatively small training sample size. Conclusions: In sum, the automated procedure of document classification presented here could improve both the precision and efficiency of systematic reviews, as well as facilitating live reviews, where reviews are updated regularly.
Topic models have been widely used to learn text representations and gain insight into document corpora. To perform topic discovery, most existing neural models either take document bag-of-words (BoW) or sequence of tokens as input followed by variational inference and BoW reconstruction to learn topic-word distribution. However, leveraging topic-word distribution for learning better features during document encoding has not been explored much. To this end, we develop a framework TAN-NTM, which processes document as a sequence of tokens through a LSTM whose contextual outputs are attended in a topic-aware manner. We propose a novel attention mechanism which factors in topic-word distribution to enable the model to attend on relevant words that convey topic related cues. The output of topic attention module is then used to carry out variational inference. We perform extensive ablations and experiments resulting in ~9-15 percentage improvement over score of existing SOTA topic models in NPMI coherence on several benchmark datasets - 20Newsgroups, Yelp Review Polarity and AGNews. Further, we show that our method learns better latent document-topic features compared to existing topic models through improvement on two downstream tasks: document classification and topic guided keyphrase generation.
In this paper, we investigate the performance and success rates of Na\"ive Bayes Classification Algorithm for automatic classification of Turkish news into predetermined categories like economy, life, health etc. We use Apache Big Data technologies such as Hadoop, HDFS, Spark and Mahout, and apply these distributed technologies to Machine Learning.
Optimization methods (optimizers) get special attention for the efficient training of neural networks in the field of deep learning. In literature there are many papers that compare neural models trained with the use of different optimizers. Each paper demonstrates that for a particular problem an optimizer is better than the others but as the problem changes this type of result is no longer valid and we have to start from scratch. In our paper we propose to use the combination of two very different optimizers but when used simultaneously they can overcome the performances of the single optimizers in very different problems. We propose a new optimizer called MAS (Mixing ADAM and SGD) that integrates SGD and ADAM simultaneously by weighing the contributions of both through the assignment of constant weights. Rather than trying to improve SGD or ADAM we exploit both at the same time by taking the best of both. We have conducted several experiments on images and text document classification, using various CNNs, and we demonstrated by experiments that the proposed MAS optimizer produces better performance than the single SGD or ADAM optimizers. The source code and all the results of the experiments are available online at the following link https://gitlab.com/nicolalandro/multi\_optimizer
The interaction of conversational systems with users poses an exciting opportunity for improving them after deployment, but little evidence has been provided of its feasibility. In most applications, users are not able to provide the correct answer to the system, but they are able to provide binary (correct, incorrect) feedback. In this paper we propose feedback-weighted learning based on importance sampling to improve upon an initial supervised system using binary user feedback. We perform simulated experiments on document classification (for development) and Conversational Question Answering datasets like QuAC and DoQA, where binary user feedback is derived from gold annotations. The results show that our method is able to improve over the initial supervised system, getting close to a fully-supervised system that has access to the same labeled examples in in-domain experiments (QuAC), and even matching in out-of-domain experiments (DoQA). Our work opens the prospect to exploit interactions with real users and improve conversational systems after deployment.
Prior research notes that BERT's computational cost grows quadratically with sequence length thus leading to longer training times, higher GPU memory constraints and carbon emissions. While recent work seeks to address these scalability issues at pre-training, these issues are also prominent in fine-tuning especially for long sequence tasks like document classification. Our work thus focuses on optimizing the computational cost of fine-tuning for document classification. We achieve this by complementary learning of both topic and language models in a unified framework, named TopicBERT. This significantly reduces the number of self-attention operations - a main performance bottleneck. Consequently, our model achieves a 1.4x ($\sim40\%$) speedup with $\sim40\%$ reduction in $CO_2$ emission while retaining $99.9\%$ performance over 5 datasets.
Task specific fine-tuning of a pre-trained neural language model using a custom softmax output layer is the de facto approach of late when dealing with document classification problems. This technique is not adequate when labeled examples are not available at training time and when the metadata artifacts in a document must be exploited. We address these challenges by generating document representations that capture both text and metadata artifacts in a task agnostic manner. Instead of traditional auto-regressive or auto-encoding based training, our novel self-supervised approach learns a soft-partition of the input space when generating text embeddings. Specifically, we employ a pre-learned topic model distribution as surrogate labels and construct a loss function based on KL divergence. Our solution also incorporates metadata explicitly rather than just augmenting them with text. The generated document embeddings exhibit compositional characteristics and are directly used by downstream classification tasks to create decision boundaries from a small number of labeled examples, thereby eschewing complicated recognition methods. We demonstrate through extensive evaluation that our proposed cross-model fusion solution outperforms several competitive baselines on multiple datasets.
Categorizing documents into a given label hierarchy is intuitively appealing due to the ubiquity of hierarchical topic structures in massive text corpora. Although related studies have achieved satisfying performance in fully supervised hierarchical document classification, they usually require massive human-annotated training data and only utilize text information. However, in many domains, (1) annotations are quite expensive where very few training samples can be acquired; (2) documents are accompanied by metadata information. Hence, this paper studies how to integrate the label hierarchy, metadata, and text signals for document categorization under weak supervision. We develop HiMeCat, an embedding-based generative framework for our task. Specifically, we propose a novel joint representation learning module that allows simultaneous modeling of category dependencies, metadata information and textual semantics, and we introduce a data augmentation module that hierarchically synthesizes training documents to complement the original, small-scale training set. Our experiments demonstrate a consistent improvement of HiMeCat over competitive baselines and validate the contribution of our representation learning and data augmentation modules.
In recent years, there has been an increased interest in the application of Natural Language Processing (NLP) to legal documents. The use of convolutional and recurrent neural networks along with word embedding techniques have presented promising results when applied to textual classification problems, such as sentiment analysis and topic segmentation of documents. This paper proposes the use of NLP techniques for textual classification, with the purpose of categorizing the descriptions of the services provided by the Public Prosecutor's Office of the State of Paran\'a to the population in one of the areas of law covered by the institution. Our main goal is to automate the process of assigning petitions to their respective areas of law, with a consequent reduction in costs and time associated with such process while allowing the allocation of human resources to more complex tasks. In this paper, we compare different approaches to word representations in the aforementioned task: including document-term matrices and a few different word embeddings. With regards to the classification models, we evaluated three different families: linear models, boosted trees and neural networks. The best results were obtained with a combination of Word2Vec trained on a domain-specific corpus and a Recurrent Neural Network (RNN) architecture (more specifically, LSTM), leading to an accuracy of 90\% and F1-Score of 85\% in the classification of eighteen categories (law areas).
In this work we present the experiments which lead to the creation of our BERT and ELECTRA based German language models, GBERT and GELECTRA. By varying the input training data, model size, and the presence of Whole Word Masking (WWM) we were able to attain SoTA performance across a set of document classification and named entity recognition (NER) tasks for both models of base and large size. We adopt an evaluation driven approach in training these models and our results indicate that both adding more data and utilizing WWM improve model performance. By benchmarking against existing German models, we show that these models are the best German models to date. Our trained models will be made publicly available to the research community.
Several works in computer vision have demonstrated the effectiveness of active learning for adapting the recognition model when new unlabeled data becomes available. Most of these works consider that labels obtained from the annotator are correct. However, in a practical scenario, as the quality of the labels depends on the annotator, some of the labels might be wrong, which results in degraded recognition performance. In this paper, we address the problems of i) how a system can identify which of the queried labels are wrong and ii) how a multi-class active learning system can be adapted to minimize the negative impact of label noise. Towards solving the problems, we propose a noisy label filtering based learning approach where the inter-relationship (context) that is quite common in natural data is utilized to detect the wrong labels. We construct a graphical representation of the unlabeled data to encode these relationships and obtain new beliefs on the graph when noisy labels are available. Comparing the new beliefs with the prior relational information, we generate a dissimilarity score to detect the incorrect labels and update the recognition model with correct labels which result in better recognition performance. This is demonstrated in three different applications: scene classification, activity classification, and document classification.