We present a Chinese BERT model dubbed MarkBERT that uses word information. Existing word-based BERT models regard words as basic units, however, due to the vocabulary limit of BERT, they only cover high-frequency words and fall back to character level when encountering out-of-vocabulary (OOV) words. Different from existing works, MarkBERT keeps the vocabulary being Chinese characters and inserts boundary markers between contiguous words. Such design enables the model to handle any words in the same way, no matter they are OOV words or not. Besides, our model has two additional benefits: first, it is convenient to add word-level learning objectives over markers, which is complementary to traditional character and sentence-level pre-training tasks; second, it can easily incorporate richer semantics such as POS tags of words by replacing generic markers with POS tag-specific markers. MarkBERT pushes the state-of-the-art of Chinese named entity recognition from 95.4\% to 96.5\% on the MSRA dataset and from 82.8\% to 84.2\% on the OntoNotes dataset, respectively. Compared to previous word-based BERT models, MarkBERT achieves better accuracy on text classification, keyword recognition, and semantic similarity tasks.
Recently, segmentation-based methods are quite popular in scene text detection, as the segmentation results can more accurately describe scene text of various shapes such as curve text. However, the post-processing of binarization is essential for segmentation-based detection, which converts probability maps produced by a segmentation method into bounding boxes/regions of text. In this paper, we propose a module named Differentiable Binarization (DB), which can perform the binarization process in a segmentation network. Optimized along with a DB module, a segmentation network can adaptively set the thresholds for binarization, which not only simplifies the post-processing but also enhances the performance of text detection. Based on a simple segmentation network, we validate the performance improvements of DB on five benchmark datasets, which consistently achieves state-of-the-art results, in terms of both detection accuracy and speed. In particular, with a light-weight backbone, the performance improvements by DB are significant so that we can look for an ideal tradeoff between detection accuracy and efficiency. Specifically, with a backbone of ResNet-18, our detector achieves an F-measure of 82.8, running at 62 FPS, on the MSRA-TD500 dataset. Code is available at: https://github.com/MhLiao/DB
Neural network NLP models are vulnerable to small modifications of the input that maintain the original meaning but result in a different prediction. In this paper, we focus on robustness of text classification against word substitutions, aiming to provide guarantees that the model prediction does not change if a word is replaced with a plausible alternative, such as a synonym. As a measure of robustness, we adopt the notion of the maximal safe radius for a given input text, which is the minimum distance in the embedding space to the decision boundary. Since computing the exact maximal safe radius is not feasible in practice, we instead approximate it by computing a lower and upper bound. For the upper bound computation, we employ Monte Carlo Tree Search in conjunction with syntactic filtering to analyse the effect of single and multiple word substitutions. The lower bound computation is achieved through an adaptation of the linear bounding techniques implemented in tools CNN-Cert and POPQORN, respectively for convolutional and recurrent network models. We evaluate the methods on sentiment analysis and news classification models for four datasets (IMDB, SST, AG News and NEWS) and a range of embeddings, and provide an analysis of robustness trends. We also apply our framework to interpretability analysis and compare it with LIME.
Large language models (LMs) are able to in-context learn -- perform a new task via inference alone by conditioning on a few input-label pairs (demonstrations) and making predictions for new inputs. However, there has been little understanding of how the model learns and which aspects of the demonstrations contribute to end task performance. In this paper, we show that ground truth demonstrations are in fact not required -- randomly replacing labels in the demonstrations barely hurts performance, consistently over 12 different models including GPT-3. Instead, we find that other aspects of the demonstrations are the key drivers of end task performance, including the fact that they provide a few examples of (1) the label space, (2) the distribution of the input text, and (3) the overall format of the sequence. Together, our analysis provides a new way of understanding how and why in-context learning works, while opening up new questions about how much can be learned from large language models through inference alone.
Microblogging sites, like Twitter, have emerged as ubiquitous sources of information. Two important tasks related to the automatic extraction and analysis of information in Microblogs are Entity Mention Detection (EMD) and Entity Detection (ED). The state-of-the-art EMD systems aim to model the non-literary nature of microblog text by training upon offline static datasets. They extract a combination of surface-level features -- orthographic, lexical, and semantic -- from individual messages for noisy text modeling and entity extraction. But given the constantly evolving nature of microblog streams, detecting all entity mentions from such varying yet limited context of short messages remains a difficult problem. To this end, we propose a framework named EMD Globalizer, better suited for the execution of EMD learners on microblog streams. It deviates from the processing of isolated microblog messages by existing EMD systems, where learned knowledge from the immediate context of a message is used to suggest entities. After an initial extraction of entity candidates by an EMD system, the proposed framework leverages occurrence mining to find additional candidate mentions that are missed during this first detection. Aggregating the local contextual representations of these mentions, a global embedding is drawn from the collective context of an entity candidate within a stream. The global embeddings are then utilized to separate entities within the candidates from false positives. All mentions of said entities from the stream are produced in the framework's final outputs. Our experiments show that EMD Globalizer can enhance the effectiveness of all existing EMD systems that we tested (on average by 25.61%) with a small additional computational overhead.
Various deep neural networks (DNNs) are developed and reported for their tremendous success in multiple domains. Given a specific task, developers can collect massive DNNs from public sources for efficient reusing and avoid redundant work from scratch. However, testing the performance (e.g., accuracy and robustness) of multiple DNNs and giving a reasonable recommendation that which model should be used is challenging regarding the scarcity of labeled data and demand of domain expertise. Existing testing approaches are mainly selection-based where after sampling, a few of the test data are labeled to discriminate DNNs. Therefore, due to the randomness of sampling, the performance ranking is not deterministic. In this paper, we propose a labeling-free comparison testing approach to overcome the limitations of labeling effort and sampling randomness. The main idea is to learn a Bayesian model to infer the models' specialty only based on predicted labels. To evaluate the effectiveness of our approach, we undertook exhaustive experiments on 9 benchmark datasets spanning in the domains of image, text, and source code, and 165 DNNs. In addition to accuracy, we consider the robustness against synthetic and natural distribution shifts. The experimental results demonstrate that the performance of existing approaches degrades under distribution shifts. Our approach outperforms the baseline methods by up to 0.74 and 0.53 on Spearman's correlation and Kendall's $\tau$, respectively, regardless of the dataset and distribution shift. Additionally, we investigated the impact of model quality (accuracy and robustness) and diversity (standard deviation of the quality) on the testing effectiveness and observe that there is a higher chance of a good result when the quality is over 50\% and the diversity is larger than 18\%.
Speech recognition is very challenging in student learning environments that are characterized by significant cross-talk and background noise. To address this problem, we present a bilingual speech recognition system that uses an interactive video analysis system to estimate the 3D speaker geometry for realistic audio simulations. We demonstrate the use of our system in generating a complex audio dataset that contains significant cross-talk and background noise that approximate real-life classroom recordings. We then test our proposed system with real-life recordings. In terms of the distance of the speakers from the microphone, our interactive video analysis system obtained a better average error rate of 10.83% compared to 33.12% for a baseline approach. Our proposed system gave an accuracy of 27.92% that is 1.5% better than Google Speech-to-text on the same dataset. In terms of 9 important keywords, our approach gave an average sensitivity of 38% compared to 24% for Google Speech-to-text, while both methods maintained high average specificity of 90% and 92%. On average, sensitivity improved from 24% to 38% for our proposed approach. On the other hand, specificity remained high for both methods (90% to 92%).
Capitalization and punctuation are important cues for comprehending written texts and conversational transcripts. Yet, many ASR systems do not produce punctuated and case-formatted speech transcripts. We propose to use a multi-task system that can exploit the relations between casing and punctuation to improve their prediction performance. Whereas text data for predicting punctuation and truecasing is seemingly abundant, we argue that written text resources are inadequate as training data for conversational models. We quantify the mismatch between written and conversational text domains by comparing the joint distributions of punctuation and word cases, and by testing our model cross-domain. Further, we show that by training the model in the written text domain and then transfer learning to conversations, we can achieve reasonable performance with less data.
Music performance synthesis aims to synthesize a musical score into a natural performance. In this paper, we borrow recent advances in text-to-speech synthesis and present the Deep Performer -- a novel system for score-to-audio music performance synthesis. Unlike speech, music often contains polyphony and long notes. Hence, we propose two new techniques for handling polyphonic inputs and providing a fine-grained conditioning in a transformer encoder-decoder model. To train our proposed system, we present a new violin dataset consisting of paired recordings and scores along with estimated alignments between them. We show that our proposed model can synthesize music with clear polyphony and harmonic structures. In a listening test, we achieve competitive quality against the baseline model, a conditional generative audio model, in terms of pitch accuracy, timbre and noise level. Moreover, our proposed model significantly outperforms the baseline on an existing piano dataset in overall quality.
Automatic evaluation metrics are essential for the rapid development of open-domain dialogue systems as they facilitate hyper-parameter tuning and comparison between models. Although recently proposed trainable conversation-level metrics have shown encouraging results, the quality of the metrics is strongly dependent on the quality of training data. Prior works mainly resort to heuristic text-level manipulations (e.g. utterances shuffling) to bootstrap incoherent conversations (negative examples) from coherent dialogues (positive examples). Such approaches are insufficient to appropriately reflect the incoherence that occurs in interactions between advanced dialogue models and humans. To tackle this problem, we propose DEAM, a Dialogue coherence Evaluation metric that relies on Abstract Meaning Representation (AMR) to apply semantic-level Manipulations for incoherent (negative) data generation. AMRs naturally facilitate the injection of various types of incoherence sources, such as coreference inconsistency, irrelevancy, contradictions, and decrease engagement, at the semantic level, thus resulting in more natural incoherent samples. Our experiments show that DEAM achieves higher correlations with human judgments compared to baseline methods on several dialog datasets by significant margins. We also show that DEAM can distinguish between coherent and incoherent dialogues generated by baseline manipulations, whereas those baseline models cannot detect incoherent examples generated by DEAM. Our results demonstrate the potential of AMR-based semantic manipulations for natural negative example generation.