This paper restores interpretability to adversarial training in text by restricting perturbations to existing words.
problem Lack of interpretability in adversarial training methods for text.
method Restrict perturbations to existing words in the input embedding space.
result Maintains or improves task performance while maintaining interpretability.
StoryGen uses images to generate coherent text from multiple images.
problem Generating coherent text from multiple images.
method Designing a Relational Text Data Generator algorithm to relate features from plural images.
result The model can generate meaningful paragraphs containing extracted features from input images.
Proposes FCNs for text classification with size-invariant inputs.
problem Classifying text of varying sizes.
method Uses fully convolutional networks with modifications to attention mechanisms.
result Suboptimal results on ITAmoji task, proposes fixes.
Scene text magnifier enhances readability for visually impaired.
problem Helps visually impaired read natural scene text.
method Four CNN-based networks: character erasing, extraction, magnify, synthesis.
result Effective text magnification without background alteration.
Adversaries reprogram text classification models without changing the original network.
problem Reprogramming neural networks trained on discrete input spaces like text classification.
method Context-based vocabulary remapping model for white-box and black-box settings.
result Successfully repurposed various text-classification models for new tasks.
Paper proposes A4NT to anonymize authors in text.
problem Revealing author attributes in text analysis.
method Combines neural machine translation and GANs for obfuscation.
result Successfully fools author attribute classifiers.
This research discovers model architecture and training dataset characteristics through strategic input probing.
problem Discovering model architecture and training dataset characteristics in black box models.
method Structured input probes and model outputs are used to train a deep classifier for image and text classification.
result The approach successfully distinguishes between different image and text datasets and architectures.
Study shows noisy historical data can still predict future text classification well.
problem Challenges in text classification with noisy, historical data.
method Examined how performance metrics on noisy data reflect future model performance.
result Noisy training data can be used to build effective prediction models for cleaner inputs.
A new method for semi-supervised text classification using layer partitioning.
problem Adapting neural semi-supervised learning to discrete text inputs.
method Decompose neural network into feature extractor F and update layer U for training. Use dropout for systematic noise. result Improves text classification especially on short texts compared to state-of-the-art methods.
Improves text-to-image translation by using GANs and captioning networks.
problem Generating images that accurately reflect the meaning of a sentence.
method Uses cycle consistent adversarial networks and captioning networks to improve image generation.
result Significantly improved image quality compared to existing methods.
AirDraw uses motion sensors to write text in the air on smart watches.
problem Limited text input on smart watch screens.
method Motion sensor data and machine learning for gesture recognition.
result 71% accuracy in text entry using AirDraw system.
New method reduces text classification errors by learning writing style instead of content.
problem Deep neural networks learn superficial patterns specific to training data.
method Adversarial training to unlearn confounding features.
result Model generalizes better and learns writing style features.
Structural-Jump-LSTM speeds up reading by skipping and jumping text.
problem Sequential inference in RNNs makes reading time linearly dependent on input length.
method Introduces a novel LSTM model with agents for skipping and jumping text.
result Structural-Jump-LSTM achieves best FLOP reduction and maintains or improves accuracy.
Modified GAN-CLS generates more plausible images from text descriptions.
problem Inference issues with GAN-CLS algorithm.
method Modified GAN-CLS algorithm with a corrected objective function.
result Generated images are more plausible and match input texts better.
Adversarial training provides a means of regularizing supervised learning algorithms while virtual adversarial training is able to extend supervised learning algorithms to the semi-supervised setting. However, both methods require making small perturbations to numerous entries of the input vector, which is inappropriat…
Paper develops a method to generate discrete adversarial attacks on text classification models using submodular optimization.
problem Generating adversarial examples for discrete structures like text is challenging.
method Formulated adversarial attacks as an optimization task on submodular set functions, guided by gradient information.
result Achieved a 1-1/e approximation factor for attacks using the greedy algorithm.
End-to-end speaker verification framework reduces text dependency.
problem Improving text-independent speaker verification.
method Jointly trains SE and ASR networks with triplet loss and adversarial gradient.
result Lower equal error rate and better text-independency compared to other approaches.
EDUCE model explains text predictions by identifying key concepts.
problem Providing clear explanations for text predictions.
method Uses unsupervised binary concept representations to explain model decisions.
result Model automatically defines relevant concepts for prediction tasks.
Project classifies malware using a multi-input LSTM model.
problem Classifying malware with high accuracy and handling class imbalance.
method Supervised multi-class LSTM model trained on annotated malware files.
result Multi-input model using both text and hexadecimal inputs outperformed single formats.
Bayesian approach quantifies uncertainty in LLM evaluations.
problem Statistical uncertainty in evaluating LLM behavior.
method Bayesian evaluation of LLM behavior using probabilistic text generation strategies.
result Bayesian approach provides useful uncertainty quantification about LLM behavior.
Paper addresses shortcomings in pointer generator networks for summarization.
problem Extractive summaries and factual inaccuracies in generated text.
method Appends traditional linguistic information to teach networks on text structure.
result Feasibility and potential of additional cues for improved generation.
Our work proves robustness of embedding schemes to discrete changes in text.
problem Discrete changes in text, like replacing a word, affect model robustness.
method Formal proofs and quantitative bounds for embedding schemes (concatenation, TF-IDF, Paragraph Vector).
result Embedding schemes are robust to discrete changes in text with Hölder or Lipschitz properties.
This work reduces the dimensionality of text data using SVD, improving performance and computational efficiency.
problem High-dimensional input spaces in text classification lead to excessive parameter count and computational infeasibility.
method Singular Value Decomposition (SVD) is applied to transform the input space into a lower-dimensional latent space.
result Neural networks trained on the lower-dimensional latent space achieve comparable or better performance with reduced computational complexity.
Joint training model for TTS and VC tasks using Tacotron and WaveNet.
problem Training a shared model for text-to-speech and voice conversion.
method Extended Tacotron model with dual attention mechanism for shared tasks, WaveNet for waveform generation.
result Joint training of a shared model achieves both TTS and VC tasks efficiently.
Sparse text alignments learned via optimal transport improve model explainability.
problem Building self-explaining models by selecting relevant text pieces.
method Employing optimal transport to find minimal cost alignments, introducing constrained variants for sparsity.
result Sparse and interpretable alignments achieved, preserving prediction accuracy.
SCROLLS benchmarks long text NLP tasks, improving existing models.
problem Short NLP benchmarks ignore long texts; SCROLLS addresses this.
method Handpicked long-text datasets for summarization, QA, and inference tasks.
result Improvement potential on SCROLLS tasks, as indicated by initial baselines.
LLMs' explanations are often insufficient and vary with input distribution.
problem Evaluating the sufficiency of LLM explanations without predefined biases.
method Generalizing sufficiency to arbitrary explanations, using LLM's input beliefs, and introducing SCSuff metric.
result Explanation sufficiency can vary with input distribution and is weakly correlated with model size, accuracy, or output entropy.
Improves text-to-image generation with bidirectional capabilities.
problem Generating realistic images from text descriptions.
method Integrates text and image modalities using MMVR architecture with n-gram cost function and multiple sentences.
result Significant improvement in image quality over existing methods (over 20%).
EBMs improve text discrimination by generating negatives from auto-regressive models.
problem Discriminating machine-generated text from human-generated text.
method Use energy-based models to discriminate text, generating negatives using pre-trained auto-regressive language models.
result EBMs can generalize well to changes in generator architectures but are sensitive to training set.
This work uses image generation models to find vision model bugs.
problem Automatically discovering failures in vision models.
method Conditional text-to-image generation and captioning models.
result Demonstrated utility of large-scale generative models to find vision model bugs.
GENs model generates sparse graphs from text-based inputs, achieving high validity.
problem Efficiently modeling and generating sparse graphs with unique and valid structures.
method RNN-based GENs model trained with an examination mechanism to predict graph characters.
result Moderate to high validity achieved in LGI strings for sparse graph generation.
Model improves classification with scarce data by focusing on relevant parts of inputs.
problem Classification tasks with limited training data.
method Approximates exact match with probability computation, learns relevant parts, boosts errors.
result Remarkable results on text classification, benefits in balanced and unbalanced cases.
Model tags biomedical texts with tree-structured tags.
problem Tagging unstructured texts with an arbitrary number of terms from an ontology.
method Neural tree decoding approach starting from the root node of an ontology.
result Outperforms state-of-the-art approaches on MeSH term assignment.
This paper tackles text infilling, a task of filling missing text portions, and presents a self-attention model that outperforms other methods.
problem The task of filling missing text portions, especially when the number and length of missing portions are unknown.
method A self-attention model with segment-aware position encoding and bidirectional context modeling, trained on extensive supervised data.
result The self-attention model significantly outperforms other approaches, setting a strong baseline for future research.
Proposes a method to train neural networks directly on compressed text data.
problem Training neural networks on compressed text data without decompression.
method Introduces composer modules to encode symbols from grammar compression rules into vector representations.
result Demonstrates that the proposed method can achieve both memory and computational efficiency while maintaining moderate performance.
Speech emotion recognition system using features and text.
problem Improving accuracy in emotion recognition from speech.
method Used speech features (Spectrogram, MFCC) and text, trained Deep Neural Networks.
result Combined MFCC-Text CNN model achieved highest accuracy.
SAM adds semantic attributes to language models for better interpretation and style variation.
problem Improving text interpretation and style variation in language models.
method SAM includes document attributes, scores them, and embeds them into the model's input space.
result SAM generates interpretable texts and shows superior performance on various datasets.
New method improves interpretability of text adversarial examples.
problem Generating interpretable adversarial examples in text is challenging.
method Sparse Projected Gradient Descent (SPGD) with directional and sparsity constraints.
result Improves interpretability and likelihood of adversarial examples compared to state-of-the-art methods.
Ground-A-Video edits videos without training, preserving intended changes.
problem Complex multi-attribute video editing with omitted or wrong changes.
method Grounding-guided video-to-video translation with Cross-Frame Gated Attention.
result Zero-shot multi-attribute video editing with improved accuracy and frame consistency.
Auxiliary Tuning adapts pre-trained models for novel tasks efficiently.
problem Adapting pre-trained models for new tasks efficiently.
method Supplementing pre-trained model with an auxiliary model that shifts output distribution.
result Achieved similar results to training from scratch with fewer resources.
A new text representation model combines CNN and VAE for better semantic extraction.
problem Difficult to effectively extract semantic features and distinguish polysemy in text data.
method Integrates CNN for feature extraction and VAE for consistent Gaussian distribution.
result The model outperforms traditional classification algorithms in text classification tasks.
DMTE integrates global connectivity for better text embeddings.
problem Lack of capturing complete connectivity between texts in graphs.
method Integrates global structural information through diffusion-convolution on text inputs, preserving high-order proximity.
result DMTE outperforms state-of-the-art methods on vertex-classification and link-prediction tasks.
Convolutional neural network (CNN) is a neural network that can make use of the internal structure of data such as the 2D structure of image data. This paper studies CNN on text categorization to exploit the 1D structure (namely, word order) of text data for accurate prediction. Instead of using low-dimensional word ve…
Paper introduces context-sensitive filters for better text processing.
problem Static filters limit NLP performance; need dynamic context.
method Meta network learns context-aware filters for sentences.
result Context-sensitive filters improve NLP tasks.
Recurrent neural networks (RNNs) process input text sequentially and model the conditional transition between word tokens. In contrast, the advantages of recursive networks include that they explicitly model the compositionality and the recursive structure of natural language. However, the current recursive architectur…
Paper surveys and introduces Acoustic Dialect Decoder for voice translation.
problem Machine understanding of natural language in speech translation.
method Recognition, Translation, and Synthesis units using HMMs, RNNs, and HTS.
result Initial successful translation of English to Tamil.
New method explains high-dimensional text classifiers.
problem Limited explainability tools for high-dimensional inputs and neural networks.
method Theoretical high-dimensional properties in neural networks.
result Improved explainability for neural network classifiers.
New method for explaining dialogue response generation models.
problem Interpreting sequence generation models, especially dialogue response generation.
method Local Explanation of Response Generation (LERG) method.
result LERG improves dialogue response generation explanations compared to existing methods.