arXiv subjects

Hong Yu

Publications and source records attributed to Hong Yu.

At least 127 records · Page 7Linked to original sources

arXiv2021

Improving Formality Style Transfer with Context-Aware Rule Injection

Models pre-trained on large-scale regular text corpora often do not work well for user-generated data where the language styles differ significantly from the mainstream text. Here we present Context-Aware Rule Injection (CARI), an innovative method for formality style transfer (FST). CARI injects multiple rules into an end-to-end BERT-based encoder and decoder model. It learns to select optimal rules based on context. The intrinsic evaluation showed that CARI achieved the new highest performance on the FST benchmark dataset. Our extrinsic evaluation showed that CARI can greatly improve the regular pre-trained models' performance on several tweet sentiment analysis tasks.

cs.CL↗

arXiv2021

CREAD: Combined Resolution of Ellipses and Anaphora in Dialogues

Anaphora and ellipses are two common phenomena in dialogues. Without resolving referring expressions and information omission, dialogue systems may fail to generate consistent and coherent responses. Traditionally, anaphora is resolved by coreference resolution and ellipses by query rewrite. In this work, we propose a novel joint learning framework of modeling coreference resolution and query rewriting for complex, multi-turn dialogue understanding. Given an ongoing dialogue between a user and a dialogue assistant, for the user query, our joint learning model first predicts coreference links between the query and the dialogue context, and then generates a self-contained rewritten user query. To evaluate our model, we annotate a dialogue based coreference resolution dataset, MuDoCo, with rewritten queries. Results show that the performance of query rewrite can be substantially boosted (+2.3% F1) with the aid of coreference modeling. Furthermore, our joint model outperforms the state-of-the-art coreference resolution model (+2% F1) on this dataset.

cs.AI↗

arXiv2021

Conversational Semantic Parsing for Dialog State Tracking

We consider a new perspective on dialog state tracking (DST), the task of estimating a user's goal through the course of a dialog. By formulating DST as a semantic parsing task over hierarchical representations, we can incorporate semantic compositionality, cross-domain knowledge sharing and co-reference. We present TreeDST, a dataset of 27k conversations annotated with tree-structured dialog states and system acts. We describe an encoder-decoder framework for DST with hierarchical representations, which leads to 20% improvement over state-of-the-art DST approaches that operate on a flat meaning space of slot-value pairs.

cs.CL↗

arXiv2021

Membership Inference Attack Susceptibility of Clinical Language Models

Deep Neural Network (DNN) models have been shown to have high empirical privacy leakages. Clinical language models (CLMs) trained on clinical data have been used to improve performance in biomedical natural language processing tasks. In this work, we investigate the risks of training-data leakage through white-box or black-box access to CLMs. We design and employ membership inference attacks to estimate the empirical privacy leaks for model architectures like BERT and GPT2. We show that membership inference attacks on CLMs lead to non-trivial privacy leakages of up to 7%. Our results show that smaller models have lower empirical privacy leakages than larger ones, and masked LMs have lower leakages than auto-regressive LMs. We further show that differentially private CLMs can have improved model utility on clinical domain while ensuring low empirical privacy leakage. Lastly, we also study the effects of group-level membership inference and disease rarity on CLM privacy leakages.

cs.CL↗

arXiv2021

A Dual-Questioning Attention Network for Emotion-Cause Pair Extraction with Context Awareness

Emotion-cause pair extraction (ECPE), an emerging task in sentiment analysis, aims at extracting pairs of emotions and their corresponding causes in documents. This is a more challenging problem than emotion cause extraction (ECE), since it requires no emotion signals which are demonstrated as an important role in the ECE task. Existing work follows a two-stage pipeline which identifies emotions and causes at the first step and pairs them at the second step. However, error propagation across steps and pair combining without contextual information limits the effectiveness. Therefore, we propose a Dual-Questioning Attention Network to alleviate these limitations. Specifically, we question candidate emotions and causes to the context independently through attention networks for a contextual and semantical answer. Also, we explore how weighted loss functions in controlling error propagation between steps. Empirical results show that our method performs better than baselines in terms of multiple evaluation metrics. The source code can be obtained at https://github.com/QixuanSun/DQAN.

cs.CL↗

arXiv2021

Continual Domain-Tuning for Pretrained Language Models

Pre-trained language models (LM) such as BERT, DistilBERT, and RoBERTa can be tuned for different domains (domain-tuning) by continuing the pre-training phase on a new target domain corpus. This simple domain tuning (SDT) technique has been widely used to create domain-tuned models such as BioBERT, SciBERT and ClinicalBERT. However, during the pretraining phase on the target domain, the LM models may catastrophically forget the patterns learned from their source domain. In this work, we study the effects of catastrophic forgetting on domain-tuned LM models and investigate methods that mitigate its negative effects. We propose continual learning (CL) based alternatives for SDT, that aim to reduce catastrophic forgetting. We show that these methods may increase the performance of LM models on downstream target domain tasks. Additionally, we also show that constraining the LM model from forgetting the source domain leads to downstream task models that are more robust to domain shifts. We analyze the computational cost of using our proposed CL methods and provide recommendations for computationally lightweight and effective CL domain-tuning procedures.

cs.CL↗

arXiv2021

A prognostic dynamic model applicable to infectious diseases providing easily visualized guides -- A case study of COVID-19 in the UK

A reasonable prediction of infectious diseases transmission process under different disease control strategies is an important reference point for policy makers. Here we established a dynamic transmission model via Python and realized comprehensive regulation of disease control measures. We classified government interventions into three categories and introduced three parameters as descriptions for the key points in disease control, these being intraregional growth rate, interregional communication rate, and detection rate of infectors. Our simulation predicts the infection by COVID-19 in the UK would be out of control in 73 days without any interventions; at the same time, herd immunity acquisition will begin from the epicentre. After we introduced government interventions, single intervention is effective in disease control but at huge expense while combined interventions would be more efficient, among which, enhancing detection number is crucial in control strategy of COVID-19. In addition, we calculated requirements for the most effective vaccination strategy based on infection number in real situation. Our model was programmed with iterative algorithms, and visualized via cellular automata, it can be applied to similar epidemics in other regions if the basic parameters are inputted, and is able to synthetically mimick the effect of multiple factors in infectious disease control.

q-bio.PE↗

arXiv2020

Conversational Machine Comprehension: a Literature Review

Conversational Machine Comprehension (CMC), a research track in conversational AI, expects the machine to understand an open-domain natural language text and thereafter engage in a multi-turn conversation to answer questions related to the text. While most of the research in Machine Reading Comprehension (MRC) revolves around single-turn question answering (QA), multi-turn CMC has recently gained prominence, thanks to the advancement in natural language understanding via neural language models such as BERT and the introduction of large-scale conversational datasets such as CoQA and QuAC. The rise in interest has, however, led to a flurry of concurrent publications, each with a different yet structurally similar modeling approach and an inconsistent view of the surrounding literature. With the volume of model submissions to conversational datasets increasing every year, there exists a need to consolidate the scattered knowledge in this domain to streamline future research. This literature review attempts at providing a holistic overview of CMC with an emphasis on the common trends across recently published models, specifically in their approach to tackling conversational history. The review synthesizes a generic framework for CMC models while highlighting the differences in recent approaches and intends to serve as a compendium of CMC for future researchers.

cs.CL↗

arXiv2020

Dynamic Data Selection for Curriculum Learning via Ability Estimation

Curriculum learning methods typically rely on heuristics to estimate the difficulty of training examples or the ability of the model. In this work, we propose replacing difficulty heuristics with learned difficulty parameters. We also propose Dynamic Data selection for Curriculum Learning via Ability Estimation (DDaCLAE), a strategy that probes model ability at each training epoch to select the best training examples at that point. We show that models using learned difficulty and/or ability outperform heuristic-based curriculum learning models on the GLUE classification tasks.

cs.CL↗

arXiv2020

Large-scale Gastric Cancer Screening and Localization Using Multi-task Deep Neural Network

Gastric cancer is one of the most common cancers, which ranks third among the leading causes of cancer death. Biopsy of gastric mucosa is a standard procedure in gastric cancer screening test. However, manual pathological inspection is labor-intensive and time-consuming. Besides, it is challenging for an automated algorithm to locate the small lesion regions in the gigapixel whole-slide image and make the decision correctly.To tackle these issues, we collected large-scale whole-slide image dataset with detailed lesion region annotation and designed a whole-slide image analyzing framework consisting of 3 networks which could not only determine the screening result but also present the suspicious areas to the pathologist for reference. Experiments demonstrated that our proposed framework achieves sensitivity of 97.05% and specificity of 92.72% in screening task and Dice coefficient of 0.8331 in segmentation task. Furthermore, we tested our best model in real-world scenario on 10,315 whole-slide images collected from 4 medical centers.

eess.IV↗

arXiv2020

Ontology-based annotation and analysis of COVID-19 phenotypes

The epidemic of COVID-19 has caused an unpredictable and devastated disaster to the public health in different territories around the world. Common phenotypes include fever, cough, shortness of breath, and chills. With more cases investigated, other clinical phenotypes are gradually recognized, for example, loss of smell, and loss of tastes. Compared with discharged or cured patients, severe or died patients often have one or more comorbidities, such as hypertension, diabetes, and cardiovascular disease. In this study, we systematically collected and analyzed COVID-19-related clinical phenotypes from 70 articles. The commonly occurring 17 phenotypes were classified into different groups based on the Human Phenotype Ontology (HPO). Based on the HP classification, we systematically analyze three nervous phenotypes (loss of smell, loss of taste, and headache) and four abdominal phenotypes (nausea, vomiting, abdominal pain, and diarrhea) identified in patients, and found that patients from Europe and USA turned to have higher nervous phenotypes and abdominal phenotypes than patients from Asia. A total of 23 comorbidities were found to commonly exist among COVID-19 patients. Patients with these comorbidities such as diabetes and kidney failure had worse outcomes compared with those without these comorbidities.

q-bio.OT↗

arXiv2020

Ontology-based systematic classification and analysis of coronaviruses, hosts, and host-coronavirus interactions towards deep understanding of COVID-19

Given the existing COVID-19 pandemic worldwide, it is critical to systematically study the interactions between hosts and coronaviruses including SARS-Cov, MERS-Cov, and SARS-CoV-2 (cause of COVID-19). We first created four host-pathogen interaction (HPI)-Outcome postulates, and generated a HPI-Outcome model as the basis for understanding host-coronavirus interactions (HCI) and their relations with the disease outcomes. We hypothesized that ontology can be used as an integrative platform to classify and analyze HCI and disease outcomes. Accordingly, we annotated and categorized different coronaviruses, hosts, and phenotypes using ontologies and identified their relations. Various COVID-19 phenotypes are hypothesized to be caused by the backend HCI mechanisms. To further identify the causal HCI-outcome relations, we collected 35 experimentally-verified HCI protein-protein interactions (PPIs), and applied literature mining to identify additional host PPIs in response to coronavirus infections. The results were formulated in a logical ontology representation for integrative HCI-outcome understanding. Using known PPIs as baits, we also developed and applied a domain-inferred prediction method to predict new PPIs and identified their pathological targets on multiple organs. Overall, our proposed ontology-based integrative framework combined with computational predictions can be used to support fundamental understanding of the intricate interactions between human patients and coronaviruses (including SARS-CoV-2) and their association with various disease outcomes.

q-bio.OT↗

arXiv2020

Birth-Burst in Evolving Networks

The evolution of complex networks is governed by both growing rules and internal properties. Most evolving network models (e.g. preferential attachment) emphasize on the growing strategy, while neglecting the characteristics of individual nodes. In this study, we analyzed a widely studied network: the evolving protein-protein interaction (PPI) network. We discovered the critical contribution of individual nodes, occurring particularly at their birth. Specifically, a node is born with a fitness value - a measurement of its intrinsic significance. Upon the introduction of a node with a large fitness into the network, a corresponding high birth-degree is determined accordingly, leading to an abrupt increase of connectivity in the network. The degree fraction of these large (hub) nodes does not decay away with the network evolution, while keeping a constant influence over the lifetime. Here we developed the birth-burst model, an adaptation of the fitness model, to simulate degree-burst and phase-transition in the network evolution.

cs.SI↗

arXiv2020

Calibrating Structured Output Predictors for Natural Language Processing

We address the problem of calibrating prediction confidence for output entities of interest in natural language processing (NLP) applications. It is important that NLP applications such as named entity recognition and question answering produce calibrated confidence scores for their predictions, especially if the system is to be deployed in a safety-critical domain such as healthcare. However, the output space of such structured prediction models is often too large to adapt binary or multi-class calibration methods directly. In this study, we propose a general calibration scheme for output entities of interest in neural-network based structured prediction models. Our proposed method can be used with any binary class calibration scheme and a neural network model. Additionally, we show that our calibration method can also be used as an uncertainty-aware, entity-specific decoding step to improve the performance of the underlying model at no additional training cost or data requirements. We show that our method outperforms current calibration techniques for named-entity-recognition, part-of-speech and question answering. We also improve our model's performance from our decoding step across several tasks and benchmark datasets. Our method improves the calibration and model performance on out-of-domain test scenarios as well.

cs.CL↗

arXiv2020

Neural Data-to-Text Generation with Dynamic Content Planning

Neural data-to-text generation models have achieved significant advancement in recent years. However, these models have two shortcomings: the generated texts tend to miss some vital information, and they often generate descriptions that are not consistent with the structured input data. To alleviate these problems, we propose a Neural data-to-text generation model with Dynamic content Planning, named NDP for abbreviation. The NDP can utilize the previously generated text to dynamically select the appropriate entry from the given structured data. We further design a reconstruction mechanism with a novel objective function that can reconstruct the whole entry of the used data sequentially from the hidden states of the decoder, which aids the accuracy of the generated text. Empirical results show that the NDP achieves superior performance over the state-of-the-art on ROTOWIRE dataset, in terms of relation generation (RG), content selection (CS), content ordering (CO) and BLEU metrics. The human evaluation result shows that the texts generated by the proposed NDP are better than the corresponding ones generated by NCP in most of time. And using the proposed reconstruction mechanism, the fidelity of the generated text can be further improved significantly.

cs.CL↗

arXiv2020

Ghost imaging based on Y-net: a dynamic coding and conjugate-decoding approach

Ghost imaging incorporating deep learning technology has recently attracted much attention in the optical imaging field. However, deterministic illumination and multiple exposure are still essential in most scenarios. Here we propose a ghost imaging scheme based on a novel conjugate-decoding deep learning framework (Y-net), which works well under both deterministic and indeterministic illumination. Benefited from the end-to-end characteristic of our network, the image of a sample can be achieved directly from a pair of correlated speckles collected by the detectors, and the sample is illuminated only once in the experiment. The spatial distribution of the speckles encoding the sample in the experiment can be completely different from that of the simulation speckles for training, as long as the statistical characteristics of the speckles remain unchanged. This approach is particularly important to high-resolution x-ray ghost imaging applications due to its potential for improving image quality and reducing radiation damage. And the idea of conjugate-decoding network may also be applied to other learning-based imaging

eess.IV↗

arXiv2020

Histogram Transform-based Speaker Identification

A novel text-independent speaker identification (SI) method is proposed. This method uses the Mel-frequency Cepstral coefficients (MFCCs) and the dynamic information among adjacent frames as feature sets to capture speaker's characteristics. In order to utilize dynamic information, we design super-MFCCs features by cascading three neighboring MFCCs frames together. The probability density function (PDF) of these super-MFCCs features is estimated by the recently proposed histogram transform~(HT) method, which generates more training data by random transforms to realize the histogram PDF estimation and recedes the commonly occurred discontinuity problem in multivariate histograms computing. Compared to the conventional PDF estimation methods, such as Gaussian mixture models, the HT model shows promising improvement in the SI performance.

cs.SD↗

arXiv2020

Language Identification with Deep Bottleneck Features

In this paper we proposed an end-to-end short utterances speech language identification(SLD) approach based on a Long Short Term Memory (LSTM) neural network which is special suitable for SLD application in intelligent vehicles. Features used for LSTM learning are generated by a transfer learning method. Bottle-neck features of a deep neural network (DNN) which are trained for mandarin acoustic-phonetic classification are used for LSTM training. In order to improve the SLD accuracy of short utterances a phase vocoder based time-scale modification(TSM) method is used to reduce and increase speech rated of the test utterance. By splicing the normal, speech rate reduced and increased utterances, we can extend length of test utterances so as to improved improved the performance of the SLD system. The experimental results on AP17-OLR database shows that the proposed methods can improve the performance of SLD, especially on short utterance with 1s and 3s durations.

cs.CL↗