arXiv subjects

Lu Wang

Publications and source records attributed to Lu Wang.

At least 307 records · Page 17Linked to original sources

arXiv2022

Towards Process-Oriented, Modular, and Versatile Question Generation that Meets Educational Needs

NLP-powered automatic question generation (QG) techniques carry great pedagogical potential of saving educators' time and benefiting student learning. Yet, QG systems have not been widely adopted in classrooms to date. In this work, we aim to pinpoint key impediments and investigate how to improve the usability of automatic QG techniques for educational purposes by understanding how instructors construct questions and identifying touch points to enhance the underlying NLP models. We perform an in-depth need finding study with 11 instructors across 7 different universities, and summarize their thought processes and needs when creating questions. While instructors show great interests in using NLP systems to support question design, none of them has used such tools in practice. They resort to multiple sources of information, ranging from domain knowledge to students' misconceptions, all of which missing from today's QG systems. We argue that building effective human-NLP collaborative QG systems that emphasize instructor control and explainability is imperative for real-world adoption. We call for QG systems to provide process-oriented support, use modular design, and handle diverse sources of input.

cs.HC↗

arXiv2022

Three-Module Modeling For End-to-End Spoken Language Understanding Using Pre-trained DNN-HMM-Based Acoustic-Phonetic Model

In spoken language understanding (SLU), what the user says is converted to his/her intent. Recent work on end-to-end SLU has shown that accuracy can be improved via pre-training approaches. We revisit ideas presented by Lugosch et al. using speech pre-training and three-module modeling; however, to ease construction of the end-to-end SLU model, we use as our phoneme module an open-source acoustic-phonetic model from a DNN-HMM hybrid automatic speech recognition (ASR) system instead of training one from scratch. Hence we fine-tune on speech only for the word module, and we apply multi-target learning (MTL) on the word and intent modules to jointly optimize SLU performance. MTL yields a relative reduction of 40% in intent-classification error rates (from 1.0% to 0.6%). Note that our three-module model is a streaming method. The final outcome of the proposed three-module modeling approach yields an intent accuracy of 99.4% on FluentSpeech, an intent error rate reduction of 50% compared to that of Lugosch et al. Although we focus on real-time streaming methods, we also list non-streaming methods for comparison.

cs.CL↗

arXiv2022

Efficient Argument Structure Extraction with Transfer Learning and Active Learning

The automation of extracting argument structures faces a pair of challenges on (1) encoding long-term contexts to facilitate comprehensive understanding, and (2) improving data efficiency since constructing high-quality argument structures is time-consuming. In this work, we propose a novel context-aware Transformer-based argument structure prediction model which, on five different domains, significantly outperforms models that rely on features or only encode limited contexts. To tackle the difficulty of data annotation, we examine two complementary methods: (i) transfer learning to leverage existing annotated data to boost model performance in a new target domain, and (ii) active learning to strategically identify a small amount of samples for annotation. We further propose model-independent sample acquisition strategies, which can be generalized to diverse domains. With extensive experiments, we show that our simple-yet-effective acquisition strategies yield competitive results against three strong comparisons. Combined with transfer learning, substantial F1 score boost (5-25) can be further achieved during the early iterations of active learning across domains.

cs.CL↗

arXiv2022

Improving Generalization of Deep Neural Networks by Leveraging Margin Distribution

Recent research has used margin theory to analyze the generalization performance for deep neural networks (DNNs). The existed results are almost based on the spectrally-normalized minimum margin. However, optimizing the minimum margin ignores a mass of information about the entire margin distribution, which is crucial to generalization performance. In this paper, we prove a generalization upper bound dominated by the statistics of the entire margin distribution. Compared with the minimum margin bounds, our bound highlights an important measure for controlling the complexity, which is the ratio of the margin standard deviation to the expected margin. We utilize a convex margin distribution loss function on the deep neural networks to validate our theoretical results by optimizing the margin ratio. Experiments and visualizations confirm the effectiveness of our approach and the correlation between generalization gap and margin ratio.

cs.LG↗

arXiv2022

HIBRIDS: Attention with Hierarchical Biases for Structure-aware Long Document Summarization

Document structure is critical for efficient information consumption. However, it is challenging to encode it efficiently into the modern Transformer architecture. In this work, we present HIBRIDS, which injects Hierarchical Biases foR Incorporating Document Structure into the calculation of attention scores. We further present a new task, hierarchical question-summary generation, for summarizing salient content in the source document into a hierarchy of questions and summaries, where each follow-up question inquires about the content of its parent question-summary pair. We also annotate a new dataset with 6,153 question-summary hierarchies labeled on long government reports. Experiment results show that our model produces better question-summary hierarchies than comparisons on both hierarchy quality and content coverage, a finding also echoed by human judges. Additionally, our model improves the generation of long-form summaries from lengthy government reports and Wikipedia articles, as measured by ROUGE scores.

cs.CL↗

arXiv2022

2020 CATARACTS Semantic Segmentation Challenge

Surgical scene segmentation is essential for anatomy and instrument localization which can be further used to assess tissue-instrument interactions during a surgical procedure. In 2017, the Challenge on Automatic Tool Annotation for cataRACT Surgery (CATARACTS) released 50 cataract surgery videos accompanied by instrument usage annotations. These annotations included frame-level instrument presence information. In 2020, we released pixel-wise semantic annotations for anatomy and instruments for 4670 images sampled from 25 videos of the CATARACTS training set. The 2020 CATARACTS Semantic Segmentation Challenge, which was a sub-challenge of the 2020 MICCAI Endoscopic Vision (EndoVis) Challenge, presented three sub-tasks to assess participating solutions on anatomical structure and instrument segmentation. Their performance was assessed on a hidden test set of 531 images from 10 videos of the CATARACTS test set.

eess.IV↗

arXiv2022

Anchor Graph Structure Fusion Hashing for Cross-Modal Similarity Search

Cross-modal hashing still has some challenges needed to address: (1) most existing CMH methods take graphs as input to model data distribution. These methods omit to consider the correlation of graph structure among multiple modalities; (2) most existing CMH methods ignores considering the fusion affinity among multi-modalities data; (3) most existing CMH methods relax the discrete constraints to solve the optimization objective, significantly degrading the retrieval performance. To solve the above limitations, we propose a novel Anchor Graph Structure Fusion Hashing (AGSFH). AGSFH constructs the anchor graph structure fusion matrix from different anchor graphs of multiple modalities with the Hadamard product, which can fully exploit the geometric property of underlying data structure. Based on the anchor graph structure fusion matrix, AGSFH attempts to directly learn an intrinsic anchor graph, where the structure of the intrinsic anchor graph is adaptively tuned so that the number of components of the intrinsic graph is exactly equal to the number of clusters. Besides, AGSFH preserves the anchor fusion affinity into the common binary Hamming space. Furthermore, a discrete optimization framework is designed to learn the unified binary codes. Extensive experimental results on three public social datasets demonstrate the superiority of AGSFH.

cs.CV↗

arXiv2022

Spatio-Temporal Coupling Controlled Laser for Electron Acceleration

Limited by the difficulty in acceleration synchronization, it has been a long-term challenge for on-chip dielectric laser-based accelerators (DLA) to bridge the gap between non-relativistic and relativistic regimes. Here, we propose a DLA based on a spatio-temporal coupling (STC) controlled laser pulse, which enables the acceleration of a non-relativistic electron to a sub-MeV level in a single acceleration structure (chirped spatial grating). It provides high precision temporal and spatial tuning of the driving laser via the dispersion manipulation, leading to a synchronous acceleration of the velocity increasing electrons over a large energy range. Additionally, the STC scheme is a general method and can be extended to driving fields of other wavelengths such as terahertz pulses. Our results bring new possibilities to MeV-scale portable electron sources and table-top acceleration experiments.

physics.optics↗

arXiv2021

Auto robust relative radiometric normalization via latent change noise modelling

Relative radiometric normalization(RRN) of different satellite images of the same terrain is necessary for change detection, object classification/segmentation, and map-making tasks. However, traditional RRN models are not robust, disturbing by object change, and RRN models precisely considering object change can not robustly obtain the no-change set. This paper proposes auto robust relative radiometric normalization methods via latent change noise modeling. They utilize the prior knowledge that no change points possess small-scale noise under relative radiometric normalization and that change points possess large-scale radiometric noise after radiometric normalization, combining the stochastic expectation maximization method to quickly and robustly extract the no-change set to learn the relative radiometric normalization mapping functions. This makes our model theoretically grounded regarding the probabilistic theory and mathematics deduction. Specifically, when we select histogram matching as the relative radiometric normalization learning scheme integrating with the mixture of Gaussian noise(HM-RRN-MoG), the HM-RRN-MoG model achieves the best performance. Our model possesses the ability to robustly against clouds/fogs/changes. Our method naturally generates a robust evaluation indicator for RRN that is the no-change set root mean square error. We apply the HM-RRN-MoG model to the latter vegetation/water change detection task, which reduces the radiometric contrast and NDVI/NDWI differences on the no-change set, generates consistent and comparable results. We utilize the no-change set into the building change detection task, efficiently reducing the pseudo-change and boosting the precision.

cs.CV↗

arXiv2021

LoRA: Low-Rank Adaptation of Large Language Models

An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA.

cs.CL↗

arXiv2021

Rapid Assessments of Light-Duty Gasoline Vehicle Emissions Using On-Road Remote Sensing and Machine Learning

In-time and accurate assessments of on-road vehicle emissions play a central role in urban air quality and health policymaking. However, official insight is hampered by the Inspection/Maintenance (I/M) procedure conducted in the laboratory annually. It not only has a large gap to real-world situations (e.g., meteorological conditions) but also is incapable of regular supervision. Here we build a unique dataset including 103831 light-duty gasoline vehicles, in which on-road remote sensing (ORRS) measurements are linked to the I/M records based on the vehicle identification numbers and license plates. On this basis, we develop an ensemble model framework that integrates three machining learning algorithms, including neural network (NN), extreme gradient boosting (XGBoost), and random forest (RF). We demonstrate that this ensemble model could rapidly assess the vehicle-specific emissions (i.e., CO, HC, and NO). In particular, the model performs quite well for the passing vehicles under normal conditions (i.e., lower VSP (< 18 kw/t), temperature (6 ~ 32 °C), relative humidity (< 80%), and wind speed (< 5m/s)). Together with the current emission standard, we identify a large number of the dirty (2.33%) or clean (74.92%) vehicles in the real world. Our results show that the ORRS measurements, assisted by the machine-learning-based ensemble model developed here, can realize day-to-day supervision of on-road vehicle-specific emissions. This approach framework provides a valuable opportunity to reform the I/M procedures globally and mitigate urban air pollution deeply.

cs.LG↗

arXiv2021

CLIFF: Contrastive Learning for Improving Faithfulness and Factuality in Abstractive Summarization

We study generating abstractive summaries that are faithful and factually consistent with the given articles. A novel contrastive learning formulation is presented, which leverages both reference summaries, as positive training data, and automatically generated erroneous summaries, as negative training data, to train summarization systems that are better at distinguishing between them. We further design four types of strategies for creating negative samples, to resemble errors made commonly by two state-of-the-art models, BART and PEGASUS, found in our new human annotations of summary errors. Experiments on XSum and CNN/Daily Mail show that our contrastive learning framework is robust across datasets and models. It consistently produces more factual summaries than strong comparisons with post error correction, entailment-based reranking, and unlikelihood training, according to QA-based factuality evaluation. Human judges echo the observation and find that our model summaries correct more errors.

cs.CL↗

arXiv2021

Hierarchical Electricity and Carbon Trading in Transmission and Distribution Networks Based on Virtual Federated Prosumer

Facing the dilemma of growing energy demand and mitigating carbon emissions, this paper proposes an energy sharing mechanism based on virtual federated prosumers (VFPs) with budget allocation for joint electricity and carbon market to incentivize distributed energy resources to participate in the hierarchical market and reduce carbon emissions. At the transmission level, the regional transmission operator coordinates transactions between two markets, the inter-VFP energy sharing market and the wholesale market, intending to minimize the overall cost of VFPs. The energy sharing market clearing problem is formulated as a generalized Nash game, for which we develop a first-order response algorithm to obtain the equilibrium. At the distribution level, the VFPs play the role of selfless auctioneer that leverage discriminatory weights and benchmark prices to allocate the electricity-carbon budget among entities in the VFP to maximize social welfare. The Nash game is exploited to characterize the budget allocation problem, for which a distributed feedback allocation algorithm is proposed. The entire hierarchical electricity and carbon trading is modeled as an equilibrium problem and is solved iteratively. Case studies based on a practical regional grid verify the effectiveness of the proposed algorithm and show that the mechanism is effective in improving energy efficiency and reducing carbon emissions.

math.OC↗

arXiv2021

Mechanism of keyhole pore formation in metal additive manufacturing

Metal additive manufacturing has gained extensive attention from research institutes and companies to fabricate intricate parts and functionally graded materials. However, the porosity of the as-built part deteriorates the mechanical property and even hinders the further application of metal additive manufacturing. Particularly, the mechanisms of keyhole pores associated with the keyhole fluctuation are not fully understood. To reveal the mechanisms of the keyhole pores formation, we adopt a multiphysics thermal-fluid flow model incorporating heat transfer, liquid flow, metal evaporation, Marangoni effect, and Darcy's law to simulate the keyhole pore formation process, and the results are validated with the in-situ X-ray images. The simulation results present the instant bubble formation due to the keyhole instability and motion of the instant bubble when it pins on the solidification front. Moreover, the unevenly distributed recoil pressure on the keyhole surface is an important factor for keyhole collapse and penetration. Furthermore, comparing the keyhole pore formation under different laser scanning speeds shows that the keyhole pore is sensitive to the manufacturing parameters. The keyhole fluctuation features and energy absorptivity variation on the rear keyhole wall could be metrics to evaluate the likelihood of the keyhole pore formation. Additionally, the simulation under a low ambient pressure shows the feasibility of improving the keyhole stability to reduce and even avoid the formation of keyhole pores.

cond-mat.mtrl-sci↗

arXiv2021

Controllable Summarization with Constrained Markov Decision Process

We study controllable text summarization which allows users to gain control on a particular attribute (e.g., length limit) of the generated summaries. In this work, we propose a novel training framework based on Constrained Markov Decision Process (CMDP), which conveniently includes a reward function along with a set of constraints, to facilitate better summarization control. The reward function encourages the generation to resemble the human-written reference, while the constraints are used to explicitly prevent the generated summaries from violating user-imposed requirements. Our framework can be applied to control important attributes of summarization, including length, covered entities, and abstractiveness, as we devise specific constraints for each of these aspects. Extensive experiments on popular benchmarks show that our CMDP framework helps generate informative summaries while complying with a given attribute's requirement.

cs.CL↗

arXiv2021

An Efficient Group-based Search Engine Marketing System for E-Commerce

With the increasing scale of search engine marketing, designing an efficient bidding system is becoming paramount for the success of e-commerce companies. The critical challenges faced by a modern industrial-level bidding system include: 1. the catalog is enormous, and the relevant bidding features are of high sparsity; 2. the large volume of bidding requests induces significant computation burden to both the offline and online serving. Leveraging extraneous user-item information proves essential to mitigate the sparsity issue, for which we exploit the natural language signals from the users' query and the contextual knowledge from the products. In particular, we extract the vector representations of ads via the Transformer model and leverage their geometric relation to building collaborative bidding predictions via clustering. The two-step procedure also significantly reduces the computation stress of bid evaluation and optimization. In this paper, we introduce the end-to-end structure of the bidding system for search engine marketing for Walmart e-commerce, which successfully handles tens of millions of bids each day. We analyze the online and offline performances of our approach and discuss how we find it as a production-efficient solution.

cs.CL↗

arXiv2021

An Evaluation of Generative Pre-Training Model-based Therapy Chatbot for Caregivers

With the advent of off-the-shelf intelligent home products and broader internet adoption, researchers increasingly explore smart computing applications that provide easier access to health and wellness resources. AI-based systems like chatbots have the potential to provide services that could provide mental health support. However, existing therapy chatbots are often retrieval-based, requiring users to respond with a constrained set of answers, which may not be appropriate given that such pre-determined inquiries may not reflect each patient's unique circumstances. Generative-based approaches, such as the OpenAI GPT models, could allow for more dynamic conversations in therapy chatbot contexts than previous approaches. To investigate the generative-based model's potential in therapy chatbot contexts, we built a chatbot using the GPT-2 model. We fine-tuned it with 306 therapy session transcripts between family caregivers of individuals with dementia and therapists conducting Problem Solving Therapy. We then evaluated the model's pre-trained and the fine-tuned model in terms of basic qualities using three meta-information measurements: the proportion of non-word outputs, the length of response, and sentiment components. Results showed that: (1) the fine-tuned model created more non-word outputs than the pre-trained model; (2) the fine-tuned model generated outputs whose length was more similar to that of the therapists compared to the pre-trained model; (3) both the pre-trained model and fine-tuned model were likely to generate more negative and fewer positive outputs than the therapists. We discuss potential reasons for the problem, the implications, and solutions for developing therapy chatbots and call for investigations of the AI-based system application.

cs.HC↗

arXiv2021

Controllable Open-ended Question Generation with A New Question Type Ontology

We investigate the less-explored task of generating open-ended questions that are typically answered by multiple sentences. We first define a new question type ontology which differentiates the nuanced nature of questions better than widely used question words. A new dataset with 4,959 questions is labeled based on the new ontology. We then propose a novel question type-aware question generation framework, augmented by a semantic graph representation, to jointly predict question focuses and produce the question. Based on this framework, we further use both exemplars and automatically generated templates to improve controllability and diversity. Experiments on two newly collected large-scale datasets show that our model improves question quality over competitive comparisons based on automatic metrics. Human judges also rate our model outputs highly in answerability, coverage of scope, and overall quality. Finally, our model variants with templates can produce questions with enhanced controllability and diversity.

cs.CL↗