Questions regarding implicitness, ambiguity and underspecification are crucial for understanding the task validity and ethical concerns of multimodal image+text systems, yet have received little attention to date. This position paper maps out a conceptual framework to address this gap, focusing on systems which generate images depicting scenes from scene descriptions. In doing so, we account for how texts and images convey meaning differently. We outline a set of core challenges concerning textual and visual ambiguity, as well as risks that may be amplified by ambiguous and underspecified elements. We propose and discuss strategies for addressing these challenges, including generating visually ambiguous images, and generating a set of diverse images.
Code-switching is a common phenomenon among multilingual speakers, where alternation between two or more languages occurs within the context of a single conversation. While multilingual humans can seamlessly switch back and forth between languages, multilingual neural machine translation (NMT) models are not robust to such sudden changes in input. This work explores multilingual NMT models' ability to handle code-switched text. First, we propose checks to measure switching capability. Second, we investigate simple and effective data augmentation methods that can enhance an NMT model's ability to support code-switching. Finally, by using a glass-box analysis of attention modules, we demonstrate the effectiveness of these methods in improving robustness.
Pre-trained speech Transformers have facilitated great success across various speech processing tasks. However, fine-tuning these encoders for downstream tasks require sufficiently large training data to converge or to achieve state-of-the-art. In text domain this has been partly attributed to sub-optimality of the representation space in pre-trained Transformers. In this work, we take a sober look into pre-trained speech encoders and rewire their representation space without requiring any task-specific labels. Our method utilises neutrally synthesised version of audio inputs along with frame masking to construct positive pairs for contrastive self-supervised learning. When used for augmenting the wav2vec 2 encoder, we observe consistent improvement of isotropy in the representation space. Our experiments on 6 speech processing tasks, exhibit a significant convergence speedup during task fine-tuning as well as consistent task improvement, specially in low-resource settings.
Answering natural language questions using information from tables (TableQA) is of considerable recent interest. In many applications, tables occur not in isolation, but embedded in, or linked to unstructured text. Often, a question is best answered by matching its parts to either table cell contents or unstructured text spans, and extracting answers from either source. This leads to a new space of TextTableQA problems that was introduced by the HybridQA dataset. Existing adaptations of table representation to transformer-based reading comprehension (RC) architectures fail to tackle the diverse modalities of the two representations through a single system. Training such systems is further challenged by the need for distant supervision. To reduce cognitive burden, training instances usually include just the question and answer, the latter matching multiple table rows and text passages. This leads to a noisy multi-instance training regime involving not only rows of the table, but also spans of linked text. We respond to these challenges by proposing MITQA, a new TextTableQA system that explicitly models the different but closely-related probability spaces of table row selection and text span selection. Our experiments indicate the superiority of our approach compared to recent baselines. The proposed method is currently at the top of the HybridQA leaderboard with a held out test set, achieving 21 % absolute improvement on both EM and F1 scores over previous published results.
Historical documents present in the form of libraries needs to be digitised. The recognition of these unconstrained cursive handwritten documents is a challenging task. In the present work, neural network based classifier is used. The recognition of scanned document images which are easy to train on neural network based systems is usually done by a two step approach: segmentation followed by recognition. This approach has several shortcomings, which includes identification of text regions, layout diversity analysis present within pages and ground truth segmentation. These processes are prone to errors that create bottleneck in the recognition accuracies. Thus in this study, an end-to-end paragraph recognition system is presented with internal line segmentation and lexicon decoder as post processing step, which is free from those errors. We named our model as LexiconNet. In LexiconNet, given a paragraph image a combination of convolution and depth-wise separable convolutional modules generates the two dimension feature map of the image. The attention module is responsible for internal line segmentation that consequently processing a page in a line by line manner. At decoding step, we have added connectionist temporal classification based word beam search decoder as a post processing step. Our approach reports state-of-the-art results on standard datasets. The reported character error rate is 3.24% on IAM dataset with 27.19% improvement, 1.13% on RIMES with 40.83% improvement and 2.43% on READ-16 dataset with 32.31% improvement from existing literature and the word error rate is 8.29% on IAM dataset with 43.02% improvement, 2.94% on RIMES dataset with 56.25% improvement and 7.35% on READ-2016 dataset with 47.27% improvement from the existing results. The character error rate and word error rate reported in this work surpasses the results reported in literature.
Figurative language generation is the task of reformulating a given text in the desired figure of speech while still being faithful to the original context. We take the first step towards multi-figurative language modelling by providing a benchmark for the automatic generation of five common figurative forms in English. We train mFLAG employing a scheme for multi-figurative language pre-training on top of BART, and a mechanism for injecting the target figurative information into the encoder; this enables the generation of text with the target figurative form from another figurative form without parallel figurative-figurative sentence pairs. Our approach outperforms all strong baselines. We also offer some qualitative analysis and reflections on the relationship between the different figures of speech.
We propose a goodness-of-fit measure for probability densities modelling observations with varying dimensionality, such as text documents of differing lengths or variable-length sequences. The proposed measure is an instance of the kernel Stein discrepancy (KSD), which has been used to construct goodness-of-fit tests for unnormalised densities. Existing KSDs require the model to be defined on a fixed-dimension space. As our major contributions, we extend the KSD to the variable dimension setting by identifying appropriate Stein operators, and propose a novel KSD goodness-of-fit test. As with the previous variants, the proposed KSD does not require the density to be normalised, allowing the evaluation of a large class of models. Our test is shown to perform well in practice on discrete sequential data benchmarks.
Diffusion models typically operate in the standard framework of generative modelling by producing continuously-valued datapoints. To this end, they rely on a progressive Gaussian smoothing of the original data distribution, which admits an SDE interpretation involving increments of a standard Brownian motion. However, some applications such as text generation or reinforcement learning might naturally be better served by diffusing categorical-valued data, i.e., lifting the diffusion to a space of probability distributions. To this end, this short theoretical note proposes Simplex Diffusion, a means to directly diffuse datapoints located on an n-dimensional probability simplex. We show how this relates to the Dirichlet distribution on the simplex and how the analogous SDE is realized thanks to a multi-dimensional Cox-Ingersoll-Ross process (abbreviated as CIR), previously used in economics and mathematical finance. Finally, we make remarks as to the numerical implementation of trajectories of the CIR process, and discuss some limitations of our approach.
Tensor decompositions have proven to be effective in analyzing the structure of multidimensional data. However, most of these methods require a key parameter: the number of desired components. In the case of the CANDECOMP/PARAFAC decomposition (CPD), this value is known as the canonical rank and greatly affects the quality of the results. Existing methods use heuristics or Bayesian methods to estimate this value by repeatedly calculating the CPD, making them extremely computationally expensive. In this work, we propose FRAPPE and Self-FRAPPE: a cheaply supervised and a self-supervised method to estimate the canonical rank of a tensor without ever having to compute the CPD. We call FRAPPE cheaply supervised because it uses a fully synthetic training set without requiring real-world examples. We evaluate these methods on synthetic tensors, real tensors of known rank, and the weight tensor of a convolutional neural network. We show that FRAPPE and Self-FRAPPE offer large improvements in both effectiveness and speed, with a respective $15\%$ and $10\%$ improvement in MAPE and an $4000\times$ and $13\times$ improvement in evaluation speed over the best-performing baseline.
We present a new framework AMOS that pretrains text encoders with an Adversarial learning curriculum via a Mixture Of Signals from multiple auxiliary generators. Following ELECTRA-style pretraining, the main encoder is trained as a discriminator to detect replaced tokens generated by auxiliary masked language models (MLMs). Different from ELECTRA which trains one MLM as the generator, we jointly train multiple MLMs of different sizes to provide training signals at various levels of difficulty. To push the discriminator to learn better with challenging replaced tokens, we learn mixture weights over the auxiliary MLMs' outputs to maximize the discriminator loss by backpropagating the gradient from the discriminator via Gumbel-Softmax. For better pretraining efficiency, we propose a way to assemble multiple MLMs into one unified auxiliary model. AMOS outperforms ELECTRA and recent state-of-the-art pretrained models by about 1 point on the GLUE benchmark for BERT base-sized models.