This work studies knowledge distillation (KD) and addresses its constraints for recurrent neural network transducer (RNN-T) models. In hard distillation, a teacher model transcribes large amounts of unlabelled speech ...
详细信息
The extraction of essential news elements through the 5W1H framework (What, When, Where, Why, Who, and How) is critical for event extraction and text summarization. The advent of Large language models (LLMs) such as C...
详细信息
ISBN:
(纸本)9798350359329;9798350359312
The extraction of essential news elements through the 5W1H framework (What, When, Where, Why, Who, and How) is critical for event extraction and text summarization. The advent of Large language models (LLMs) such as ChatGPT presents an opportunity to address language-related tasks through simple prompts without fine-tuning models with much time. While ChatGPT has encountered challenges in processing longer news texts and analyzing specific attributes in context, especially answering questions about What, Why, and How. The effectiveness of extraction tasks is notably dependent on highquality human-annotated datasets. However, the absence of such datasets for the 5W1H extraction increases the difficulty of finetuning strategies based on open-source LLMs. To address these limitations, first, we annotate a high-quality 5W1H dataset based on four typical news corpora (CNN/DailyMail, XSum, NYT, RAMDS);second, we design several strategies from zero-shot/fewshot prompting to efficient fine-tuning to conduct 5W1H aspects extraction from the original news documents. The experimental results demonstrate that the performance of the fine-tuned models on our labelled dataset is superior to the performance of ChatGPT. Furthermore, we also explore the domain adaptation capability by testing the source-domain (e.g. NYT) models on the target domain corpus (e.g. CNN/DailyMail) for the task of 5W1H extraction.
This paper presents a back propagation neural network (BPNN) patternrecognition algorithm based on distributed framework. In this paper, four typical partial discharge(PD) models of artificial oil-paper insulation de...
详细信息
Artificial neural networks use a lot of coefficients that take a great deal of computing power for their adjustment, especially if deep learning networks are employed. However, there exist coefficients-free extremely ...
详细信息
In order to recognize patterns in images, this study tests the performance of many 'machine learning algorithms' and feature extraction methods. Here, synthetic photographs of handwritten digits are used to co...
详细信息
Diabetic retinopathy is spreading dangerously worldwide among people with diabetes, leading to reduced vision and completely blindness. In this paper, a technique is proposed to identify diabetic retinopathy using the...
详细信息
Approaches to the development of the system of music synthesis and recognition are considered. In addition, such audio software as part of the smart house system can bring additional benefits and increased experience ...
详细信息
It is well known that photo response non-uniformity (PRNU) noise based source attribution helps to verify the camera used to take an image. Recent advances in content-aware image resizing method such as seam carving a...
详细信息
ISBN:
(纸本)9783031234798;9783031234804
It is well known that photo response non-uniformity (PRNU) noise based source attribution helps to verify the camera used to take an image. Recent advances in content-aware image resizing method such as seam carving allow an image to be resized while the critical content is retained. In this paper, we propose identifying the source camera from seam inserted images using blocks as small as 20 x 20. In particular, the correlation is computed between the noise residue of the seam inserted image and the camera PRNU constructed using different numbers of im-ages. We found that different correlation patterns with the camera PRNUs are ob-served, depending on whether the image is taken by that camera or not. Addition-ally, based on this observation, features are extracted from the correlation patterns which are then weighted and combined to form a decision metric for source cam-era identification. We demonstrate by our experimental results that our approach is effective in identifying the source camera in seam insertion images.
The study compares the local binary pattern histogram method to a PGC and stereo vision picture acquired utilising innovative violas jones algorithm. An automated face identification and recognition system was tested ...
The study compares the local binary pattern histogram method to a PGC and stereo vision picture acquired utilising innovative violas jones algorithm. An automated face identification and recognition system was tested utilising the innovative violas jones algorithm and Local Binary pattern Histogram (LBPH) Algorithm using 20 samples at various periods. The G power with 0.8 pre-testing power, 0.05 alpha, and 0.95 confidence interval is suggested for SPSS accuracy prediction. The model is tested and trained on 40 kaggle (***) samples. Automatic face identification and recognition system performance was assessed using the innovative violas jones method and Local Binary pattern Histogram method with 20 samples at various periods. After statistical analysis, the violas jones algorithm achieved 93% accuracy, while the local binary pattern histogram algorithm achieved 84% accuracy using the default setting. The independent sample t-test significance was =0.001 (<0.05).
Partial discharge (PD) has been proved to be a sensitive indicator to reveal the insulation condition of medium voltage (MV) switchgear. Since the influence of a PD on MV switchgear strongly depends on its type and lo...
详细信息
ISBN:
(纸本)9781665451451
Partial discharge (PD) has been proved to be a sensitive indicator to reveal the insulation condition of medium voltage (MV) switchgear. Since the influence of a PD on MV switchgear strongly depends on its type and location, accurate recognition of the PD is crucial to improve the efficiency of on-site measurement and optimize the maintenance strategy. Accordingly, this paper develops a PD recognition method based on gray-scale diagram and support vector machine. Labeled signals including many PD pulses are transformed into a gray-scale diagram via a two-dimensional maximum margin criterion, significantly reducing the dimensionality of the PD data thus improving computing efficiency. Models of different types of PD defects are then trained by applying the support vector machine to the labeled gray-scale diagrams. Eventually, the type of a PD is identified via the trained models. The effectiveness and feasibility of the proposed method are verified on four artificial PD models and three simulated PD defects on a real 10-kV switchgear.
暂无评论