Distant-Supervised Slot-Filling for E-Commerce Queries

Manchanda, Saurav; Sharma, Mohit; Karypis, George

doi:10.1109/bigdata52589.2021.9671825

Cited by 5 publications

(3 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The usefulness of these agents was at peak during pandemic when the life was restricted without human assistance for any problem. For designing these agents two main components are required namely slot filling (Firdaus et al, 2021;Manchanda et al, 2021) and intent detection (Popov et al, 2019;Reza et al, 2020) and their performance have a direct effect on human language understanding and efficiently performing various natural language related tasks (Yang et al, 2021;Abro et al,2022). Most of the research considers existing datasets and apply their slot filling and intent detection models on them either separately or in an integrated way.…”

Section: Introductionmentioning

confidence: 99%

RETRACTED: Self Learning and Context based approach of Slot Label Creation and Slot Filling for Interactive Systems

Magoo

Singh

2023

Preprint

View full text Add to dashboard Cite

Slot-filling is considered one of the most important components of task-based interactive systems. The main challenge in slot filling is to assign a semantic label to each text in the utterance which is further useful for generating the response by the interactive system. Another challenge is the non-availability of extensive labeled data. Thus in this research, we primarily focus on developing two approaches namely (i) automatic slot identification through IOB labeling for slot labels creation and preparing a labeled dataset (ii) a self-learning based model where the slot labels are created in step (i) are predicted for an untrained query from the same or different domains. We use BI-LSTM with Elmo Embedding for our proposed frameworks with POS Tags to generate slot-independent tags and label descriptions for slot labeling. We applied and compared our model on the SNIPS and Mobile service provider dataset through the F1 score and analyzed that the efficiency of self-learning-based models highly depends upon the quality and correlation among different domains of the same dataset.

show abstract

Section: Introductionmentioning

confidence: 99%

RETRACTED: Self Learning and Context based approach of Slot Label Creation and Slot Filling for Interactive Systems

Magoo

Singh

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…These methods pretrain large neural networks on self-supervised tasks in order to encode common contextual knowledge in the structured input. Another approach imparts domain-specific knowledge by pretraining models on tasks related to the target task but for which abundant labeled data is available [21,32]. The pretrained model is then fine-tuned on the limited training data of the target task.…”

Section: Transfer Learningmentioning

confidence: 99%

Injecting Domain Knowledge from Empirical Interatomic Potentials to Neural Networks for Predicting Material Properties

Shui¹,

Karls²,

Wen³

et al. 2022

Preprint

View full text Add to dashboard Cite

For decades, atomistic modeling has played a crucial role in predicting the behavior of materials in numerous fields ranging from nanotechnology to drug discovery. The most accurate methods in this domain are rooted in first-principles quantum mechanical calculations such as density functional theory (DFT). Because these methods have remained computationally prohibitive, practitioners have traditionally focused on defining physically motivated closed-form expressions known as empirical interatomic potentials (EIPs) that approximately model the interactions between atoms in materials. In recent years, neural network (NN)-based potentials trained on quantum mechanical (DFT-labeled) data have emerged as a more accurate alternative to conventional EIPs. However, the generalizability of these models relies heavily on the amount of labeled training data, which is often still insufficient to generate models suitable for general-purpose applications. In this paper, we propose two generic strategies that take advantage of unlabeled training instances to inject domain knowledge from conventional EIPs to NNs in order to increase their generalizability. The first strategy, based on weakly supervised learning, trains an auxiliary classifier on EIPs and selects the best-performing EIP to generate energies to supplement the ground-truth DFT energies in training the NN. The second strategy, based on transfer learning, first pretrains the NN on a large set of easily obtainable EIP energies, and then fine-tunes it on ground-truth DFT energies. Experimental results on three benchmark datasets demonstrate that the first strategy improves baseline NN performance by 5% to 51% while the second improves baseline performance by up to 55%. Combining them further boosts performance.

show abstract

“…In the case of small data sets, transfer learning has been very successful, as exemplified by Bert in natural language processing. Unlike the method of pretraining large neural networks on self-supervised tasks in these models, we pretrain the model on another relevant task with abundant labeled data, then finetune the pretrained model on limited data from the target task. , …”

mentioning

confidence: 99%

A Multiple-Fidelity Method for Accurate Simulation of MoS₂ Properties Using JAX-ReaxFF and Neural Network Potentials

Wang,

Xu,

Shao

et al. 2024

J. Phys. Chem. Lett.

View full text Add to dashboard Cite

Reactive force field (ReaxFF) is a commonly used force field for modeling chemical reactions at the atomic level. Recently, JAX-ReaxFF, combined with automatic differentiation, has been used to efficiently parametrize ReaxFF. However, its analytical formula may lead to inaccurate predictions. While neural network-based potentials (NNPs) trained on density functional theory-labeled data offer a more accurate method, it requires a large amount of training data to be trained from scratch. To overcome these issues, we present a multiple-fidelity method that combines JAX-ReaxFF and NNP and apply the method on MoS 2 , a promising two-dimensional semiconductor for flexible electronics. By incorporating implicit prior physical information, ReaxFF can serve as a cost-effective way to generate pretraining data, facilitating more accurate simulations of MoS 2 . Moreover, in the Mo−S−H system, the pretraining strategy can reduce root-mean-square errors of energy by 20%. This approach can be extended to a wide variety of material systems, accelerating their computational research.

show abstract

Distant-Supervised Slot-Filling for E-Commerce Queries

Cited by 5 publications

References 32 publications

RETRACTED: Self Learning and Context based approach of Slot Label Creation and Slot Filling for Interactive Systems

RETRACTED: Self Learning and Context based approach of Slot Label Creation and Slot Filling for Interactive Systems

Injecting Domain Knowledge from Empirical Interatomic Potentials to Neural Networks for Predicting Material Properties

A Multiple-Fidelity Method for Accurate Simulation of MoS₂ Properties Using JAX-ReaxFF and Neural Network Potentials

Contact Info

Product

Resources

About

Distant-Supervised Slot-Filling for E-Commerce Queries

Cited by 5 publications

References 32 publications

RETRACTED: Self Learning and Context based approach of Slot Label Creation and Slot Filling for Interactive Systems

RETRACTED: Self Learning and Context based approach of Slot Label Creation and Slot Filling for Interactive Systems

Injecting Domain Knowledge from Empirical Interatomic Potentials to Neural Networks for Predicting Material Properties

A Multiple-Fidelity Method for Accurate Simulation of MoS2 Properties Using JAX-ReaxFF and Neural Network Potentials

Contact Info

Product

Resources

About

A Multiple-Fidelity Method for Accurate Simulation of MoS₂ Properties Using JAX-ReaxFF and Neural Network Potentials