Prior studies in privacy policies frame the question answering (QA) tasks as identifying the most relevant text segment or a list of sentences from the policy document for a user query. However, annotating such a dataset is challenging as it requires specific domain expertise (e.g., law academics). Even if we manage a small-scale one, a bottleneck that remains is that the labeled data are heavily imbalanced (only a few segments are relevant) --limiting the gain in this domain. Therefore, in this paper, we develop a novel data augmentation framework based on ensembling retriever models that captures the relevant text segments from unlabeled policy documents and expand the positive examples in the training set. In addition, to improve the diversity and quality of the augmented data, we leverage multiple pre-trained language models (LMs) and cascaded them with noise reduction oracles. Using our augmented data on the PrivacyQA benchmark, we elevate the existing baseline by a large margin (10\% F1) and achieve a new state-of-the-art F1 score of 50\%. Our ablation studies provide further insights into the effectiveness of our approach.
The stylistic properties of text have intrigued computational linguistics researchers in recent years. Specifically, researchers have investigated the Text Style Transfer (TST) task, which aims to change the stylistic properties of the text while retaining its style independent content. Over the last few years, many novel TST algorithms have been developed, while the industry has leveraged these algorithms to enable exciting TST applications. The field of TST research has burgeoned because of this symbiosis. This article aims to provide a comprehensive review of recent research efforts on text style transfer. More concretely, we create a taxonomy to organize the TST models and provide a comprehensive summary of the state of the art. We review the existing evaluation methodologies for TST tasks and conduct a large-scale reproducibility study where we experimentally benchmark 19 state-of-the-art TST algorithms on two publicly available datasets. Finally, we expand on current trends and provide new perspectives on the new and exciting developments in the TST field.
We seek to improve text classification by leveraging naturally annotated data. In particular, we construct a general purpose text categorization dataset (NatCat) from three online resources: Wikipedia, Reddit, and Stack Exchange. These datasets consist of document-category pairs derived from manual curation that occurs naturally by their communities. We build general purpose text classifiers by training on NatCat and evaluate them on a suite of 11 text classification tasks (CatEval). We benchmark different modeling choices and dataset combinations, and show how each task benefits from different NatCat training resources.
Compared to the linear MIMO detectors, the Belief Propagation (BP) detector has shown greater capabilities in achieving near optimal performance and better nature to iteratively cooperate with channel decoders. Aiming at real applications, recent works mainly fall into the category of reducing the complexity by simplified calculations, at the expense of performance sacrifice. However, the complexity is still unsatisfactory with exponentially increasing complexity or required exponentiation operations. Furthermore, due to the inherent loopy structure, the existing BP detectors persistently encounter error floor in high signal-to-noise ratio (SNR) region, which becomes even worse with calculation approximation. This work aims at a revised BP detector, named {Belief-selective Propagation (BsP)} detector by selectively utilizing the \emph{trusted} incoming messages with sufficiently large \textit{a priori} probabilities for updates. Two proposed strategies: symbol-based truncation (ST) and edge-based simplification (ES) squeeze the complexity (orders lower than the Original-BP), while greatly relieving the error floor issue over a wide range of antenna and modulation combinations. For the $16$-QAM $8 \times 4$ MIMO system, the $\mathcal{B}(1,1)$ {BsP} detector achieves more than $4$\,dB performance gain (@$\text{BER}=10^{-4}$) with roughly $4$ orders lower complexity than the Original-BP detector. Trade-off between performance and complexity towards different application requirement can be conveniently obtained by configuring the ST and ES parameters.
Despite the success of deep neural networks (DNNs) for real-world applications over time-series data such as mobile health, little is known about how to train robust DNNs for time-series domain due to its unique characteristics compared to images and text data. In this paper, we propose a novel algorithmic framework referred as RObust Training for Time-Series (RO-TS) to create robust DNNs for time-series classification tasks. Specifically, we formulate a min-max optimization problem over the model parameters by explicitly reasoning about the robustness criteria in terms of additive perturbations to time-series inputs measured by the global alignment kernel (GAK) based distance. We also show the generality and advantages of our formulation using the summation structure over time-series alignments by relating both GAK and dynamic time warping (DTW). This problem is an instance of a family of compositional min-max optimization problems, which are challenging and open with unclear theoretical guarantee. We propose a principled stochastic compositional alternating gradient descent ascent (SCAGDA) algorithm for this family of optimization problems. Unlike traditional methods for time-series that require approximate computation of distance measures, SCAGDA approximates the GAK based distance on-the-fly using a moving average approach. We theoretically analyze the convergence rate of SCAGDA and provide strong theoretical support for the estimation of GAK based distance. Our experiments on real-world benchmarks demonstrate that RO-TS creates more robust DNNs when compared to adversarial training using prior methods that rely on data augmentation or new definitions of loss functions. We also demonstrate the importance of GAK for time-series data over the Euclidean distance. The source code of RO-TS algorithms is available at https://github.com/tahabelkhouja/Robust-Training-for-Time-Series
Document classification is the detection specific content of interest in text documents. In contrast to the data-driven machine learning classifiers, knowledge-based classifiers can be constructed based on domain specific knowledge, which usually takes the form of a collection of subject related keywords. While typical knowledge-based classifiers compute a prediction score based on the keyword abundance, it generally suffers from noisy detections due to the lack of guiding principle in gauging the keyword matches. In this paper, we propose a novel knowledge-based model equipped with Shannon Entropy, which measures the richness of information and favors uniform and diverse keyword matches. Without invoking any positive sample, such method provides a simple and explainable solution for document classification. We show that the Shannon Entropy significantly improves the recall at fixed level of false positive rate. Also, we show that the model is more robust against change of data distribution at inference while compared with traditional machine learning, particularly when the positive training samples are very limited.
Stance detection aims to identify whether the author of a text is in favor of, against, or neutral to a given target. The main challenge of this task comes two-fold: few-shot learning resulting from the varying targets and the lack of contextual information of the targets. Existing works mainly focus on solving the second issue by designing attention-based models or introducing noisy external knowledge, while the first issue remains under-explored. In this paper, inspired by the potential capability of pre-trained language models (PLMs) serving as knowledge bases and few-shot learners, we propose to introduce prompt-based fine-tuning for stance detection. PLMs can provide essential contextual information for the targets and enable few-shot learning via prompts. Considering the crucial role of the target in stance detection task, we design target-aware prompts and propose a novel verbalizer. Instead of mapping each label to a concrete word, our verbalizer maps each label to a vector and picks the label that best captures the correlation between the stance and the target. Moreover, to alleviate the possible defect of dealing with varying targets with a single hand-crafted prompt, we propose to distill the information learned from multiple prompts. Experimental results show the superior performance of our proposed model in both full-data and few-shot scenarios.
Graph neural networks have triggered a resurgence of graph-based text classification. We show that already a simple MLP baseline achieves comparable performance on benchmark datasets, questioning the importance of synthetic graph structures. When considering an inductive scenario, i. e., when adding new documents to a corpus, a simple MLP even outperforms the recent graph-based models TextGCN and HeteGCN and is comparable with HyperGAT. We further fine-tune DistilBERT and find that it outperforms all state-of-the-art models. We suggest that future studies use at least an MLP baseline to contextualize the results. We provide recommendations for the design and training of such a baseline.
We introduce an audiovisual method for long-range text-to-video retrieval. Unlike previous approaches designed for short video retrieval (e.g., 5-15 seconds in duration), our approach aims to retrieve minute-long videos that capture complex human actions. One challenge of standard video-only approaches is the large computational cost associated with processing hundreds of densely extracted frames from such long videos. To address this issue, we propose to replace parts of the video with compact audio cues that succinctly summarize dynamic audio events and are cheap to process. Our method, named ECLIPSE (Efficient CLIP with Sound Encoding), adapts the popular CLIP model to an audiovisual video setting, by adding a unified audiovisual transformer block that captures complementary cues from the video and audio streams. In addition to being 2.92x faster and 2.34x memory-efficient than long-range video-only approaches, our method also achieves better text-to-video retrieval accuracy on several diverse long-range video datasets such as ActivityNet, QVHighlights, YouCook2, DiDeMo and Charades.
This paper is our attempt at answering a twofold question covering the areas of ethics and authorship analysis. Firstly, since the methods used for performing authorship analysis imply that an author can be recognized by the content he or she creates, we were interested in finding out whether it would be possible for an author identification system to correctly attribute works to authors if in the course of years they have undergone a major psychological transition. Secondly, and from the point of view of the evolution of an author's ethical values, we checked what it would mean if the authorship attribution system encounters difficulties in detecting single authorship. We set out to answer those questions through performing a binary authorship analysis task using a text classifier based on a pre-trained transformer model and a baseline method relying on conventional similarity metrics. For the test set, we chose works of Arata Osada, a Japanese educator and specialist in the history of education, with half of them being books written before the World War II and another half in the 1950s, in between which he underwent a transformation in terms of political opinions. As a result, we were able to confirm that in the case of texts authored by Arata Osada in a time span of more than 10 years, while the classification accuracy drops by a large margin and is substantially lower than for texts by other non-fiction writers, confidence scores of the predictions remain at a similar level as in the case of a shorter time span, indicating that the classifier was in many instances tricked into deciding that texts written over a time span of multiple years were actually written by two different people, which in turn leads us to believe that such a change can affect authorship analysis, and that historical events have great impact on a person's ethical outlook as expressed in their writings.