Search

Semantic parsing maps natural language questions into logical forms, which can be executed against a knowledge base for answers. In real-world applications, the performance of a parser is often limited by the lack of training data. To facilitate zero-shot learning, data synthesis has been widely studied to automatically generate paired questions and logical forms. However, data synthesis methods can hardly cover the diverse structures in natural languages, leading to a large gap in sentence structure between synthetic and natural questions. In this paper, we propose a decomposition-based method to unify the sentence structures of questions, which benefits the generalization to natural questions. Experiments demonstrate that our method significantly improves the semantic parser trained on synthetic data (+7.9% on KQA and +8.9% on ComplexWebQuestions in terms of exact match accuracy). Extensive analysis demonstrates that our method can better generalize to natural questions with novel text expressions compared with baselines. Besides semantic parsing, our idea potentially benefits other semantic understanding tasks by mitigating the distracting structure features. To illustrate this, we extend our method to the task of sentence embedding learning, and observe substantial improvements on sentence retrieval (+13.1% for Hit@1).

The genus Myotis is one of the most diverse and widely distributed mammals, providing a good model for studies of speciation and diversification across large geographic scales. However, the classification within this genus has long been chaotic. Taxonomic revisions based on multiple data sources are essential and urgent. In this study, morphometrics and genetic markers with different modes of inheritance were used to clarify the taxonomy of Myotis distributed in China. Based on 173 mitochondrial Cytb sequences and five morphological characteristics, 114 specimens collected nationwide over the past 20 years were assigned to 11 Myotis species. All Chinese samples classified into M. davidii and M. longipes were revised to M. alticraniatus and M. laniger. Then, two nuclear fragments (Rag2 and Chd1) and Cytb sequences from representative individuals of Chinese Myotis were used for multilocus phylogeny reconstruction and genetic divergence evaluation. The phylogenetic relationships were clearly demonstrated in the species tree: M. alticraniatus and M. laniger; M. fimbriatus, M. pilosus, M. macrodactylus, and M. petax; and M. pequinius, M. chinensis, and M. blythii formed three strongly supported monophyletic clades. Mitochondrial divergence was almost 10 times that of nuclear divergence, with interspecific K2P distances ranging from 8% to 20% for Cytb and 0.3% to 2.3% for concatenated nuclear genes. Low levels of genetic divergence were observed between M. alticraniatus and M. laniger, as well as M. fimbriatus and M. pilosus. These results provide new insights into the taxonomy and phylogeny of Myotis bats in China and are important for the future research and conservation of Chinese Myotis.

Spatially confined assembly of semimetallic oxyanions (AsO

Pre-trained Language Models (PLMs) have achieved great success on Machine Reading Comprehension (MRC) over the past few years. Although the general language representation learned from large-scale corpora does benefit MRC, the poor support in evidence extraction which requires reasoning across multiple sentences hinders PLMs from further advancing MRC. To bridge the gap between general PLMs and MRC, we present REPT, a REtrieval-based Pre-Training approach. In particular, we introduce two self-supervised tasks to strengthen evidence extraction during pre-training, which is further inherited by downstream MRC tasks through the consistent retrieval operation and model architecture. To evaluate our proposed method, we conduct extensive experiments on five MRC datasets that require collecting evidence from and reasoning across multiple sentences. Experimental results demonstrate the effectiveness of our pre-training approach. Moreover, further analysis shows that our approach is able to enhance the capacity of evidence extraction without explicit supervision., 14 pages, 3 figures, Findings of ACL 2021

Unsupervised commonsense question answering is appealing since it does not rely on any labeled task data. Among existing work, a popular solution is to use pre-trained language models to score candidate choices directly conditioned on the question or context. However, such scores from language models can be easily affected by irrelevant factors, such as word frequencies, sentence structures, etc. These distracting factors may not only mislead the model to choose a wrong answer but also make it oversensitive to lexical perturbations in candidate answers. In this paper, we present a novel SEmantic-based Question Answering method (SEQA) for unsupervised commonsense question answering. Instead of directly scoring each answer choice, our method first generates a set of plausible answers with generative models (e.g., GPT-2), and then uses these plausible answers to select the correct choice by considering the semantic similarity between each plausible answer and each choice. We devise a simple, yet sound formalism for this idea and verify its effectiveness and robustness with extensive experiments. We evaluate the proposed method on four benchmark datasets, and our method achieves the best results in unsupervised settings. Moreover, when attacked by TextFooler with synonym replacement, SEQA demonstrates much less performance drops than baselines, thereby indicating stronger robustness., Comment: Accepted by ACL 2021 (long paper)

Neural models have achieved great success on machine reading comprehension (MRC), many of which typically consist of two components: an evidence extractor and an answer predictor. The former seeks the most relevant information from a reference text, while the latter is to locate or generate answers from the extracted evidence. Despite the importance of evidence labels for training the evidence extractor, they are not cheaply accessible, particularly in many non-extractive MRC tasks such as YES/NO question answering and multi-choice MRC. To address this problem, we present a Self-Training method (STM), which supervises the evidence extractor with auto-generated evidence labels in an iterative process. At each iteration, a base MRC model is trained with golden answers and noisy evidence labels. The trained model will predict pseudo evidence labels as extra supervision in the next iteration. We evaluate STM on seven datasets over three MRC tasks. Experimental results demonstrate the improvement on existing MRC models, and we also analyze how and why such a self-training method works in MRC. The source code can be obtained from https://github.com/SparkJiao/Self-Training-MRC, 12 pages, accepted by ACL 2020

Performance of models highly depend not only on the used algorithm but also the data set it was applied to. This makes the comparison of newly developed tools to previously published approaches difficult. Either researchers need to implement others' algorithms first, to establish an adequate benchmark on their data, or a direct comparison of new and old techniques is infeasible. The Ischemic Stroke Lesion Segmentation (ISLES) challenge, which has ran now consecutively for 3 years, aims to address this problem of comparability. ISLES 2016 and 2017 focused on lesion outcome prediction after ischemic stroke: By providing a uniformly pre-processed data set, researchers from all over the world could apply their algorithm directly. A total of nine teams participated in ISLES 2015, and 15 teams participated in ISLES 2016. Their performance was evaluated in a fair and transparent way to identify the state-of-the-art among all submissions. Top ranked teams almost always employed deep learning tools, which were predominately convolutional neural networks (CNNs). Despite the great efforts, lesion outcome prediction persists challenging. The annotated data set remains publicly available and new approaches can be compared directly via the online evaluation system, serving as a continuing benchmark (www.isles-challenge.org).

Introduction: Predicting the final infarct for acute stroke from acute imaging is valuable for triage and prognosis (1). We developed and tested a Deep Learning model, based on a 3D Convolutional Neural Network (CNN) architecture, to predict the lesion based on a 30-90 day scan from acute images. Method: The model has 3 advantages compared with traditional CNN (Figure): 1. 3D CNN more efficiently utilizes spatial information across multiple slices. 2. Patches instead of entire images are used as inputs. With a patch-based approach, we avoid the impact of minimally-relevant information from distant voxels. Training on patches also prevents over-fitting by augmenting each image into thousands of samples. 3. Multi-scale structure is used by processing patches with different resolutions. Segmentation based on a single scale image cannot fully capture varying local information and may miss contextual information. We use two scales of patches to learn both local and global context. The model was trained and tested using the MICCAI ISLES 2017 challenge dataset (2), which consists of 43 cases with acute ADC and PWI maps paired with annotated final infarct segmentations at day 30-90. Dice score coefficient (DSC) and AUC for segmentation were used as quality metrics. Results: Using 70% of dataset for training and rest for testing, we achieved a DSC of 0.43±0.18 and 0.90 for AUC. As a comparison, the winning entry for the ISLES 2016 challenge achieved a DSC of 0.31, while several previous research works (threshold based, cluster based, Generalized Linear Model, etc.) achieve up to 0.84 AUC (1). Conclusion: We demonstrate that a deep learning approach can predict the final stroke lesion based on acute diffusion and perfusion neuroimaging data. Given its inherent speed, high performance, and capacity for further training, deep learning is a promising method for stroke lesion outcome prediction. References 1. Rekik et al., Neuroimage Clin 2012; 2. http://www.isles-challenge.org/

This paper presents a MOOC-ready online FPGA laboratory platform which targets computer system experiments. Goal of design is to provide user with highly approximate experience and results as offline experiments. Rich functions are implemented by utilizing SoC FPGA as the controller of lab board. The design details and effects are discussed in this paper.

Sememes are minimum semantic units of word meanings, and the meaning of each word sense is typically composed by several sememes. Since sememes are not explicit for each word, people manually annotate word sememes and form linguistic common-sense knowledge bases. In this paper, we present that, word sememe information can improve word representation learning (WRL), which maps words into a low-dimensional semantic space and serves as a fundamental step for many NLP tasks. The key idea is to utilize word sememes to capture exact meanings of a word within specific contexts accurately. More specifically, we follow the framework of Skip-gram and present three sememe-encoded models to learn representations of sememes, senses and words, where we apply the attention scheme to detect word senses in various contexts. We conduct experiments on two tasks including word similarity and word analogy, and our models significantly outperform baselines. The results indicate that WRL can benefit from sememes via the attention scheme, and also confirm our models being capable of correctly modeling sememe information.