This repo is the generalization of the lecture-summarizer repo. This tool utilizes the HuggingFace Pytorch transformers library to run extractive summarizations. This works by first embedding the sentences, then
全称: Fine-tune BERT for Extractive Summarization 时间:2019.09.05 团队: University of Edinburgh Code地址github.com/nlpyang/BertSum Paper地址arxiv.org/pdf/1903.10318.pdf 2.1.2 模型框架 背景:预训练模型BERT在多个NLP任务上的表现效果突出 主要贡献: 将BERT用于抽取式摘要中 ...
基于论文Fine-tune BERT for Extractive Summarization的方法论&源代码,进行调整,在中文数据集中进行实验。 参考论文作者主页(含论文pdf & 源代码链接):http://nlp-yang.github.io/ 数据集 中文数据集:LCSTS2.0(A Large Scale Chinese Short Text Summarization Dataset) ...
GitHub is where people build software. More than 150 million people use GitHub to discover, fork, and contribute to over 420 million projects.
疫情期间在家学习,期间学习到Fine-tune BERT for Extractive Summarization。将bert模型运用于抽取式文本摘要中,第一部分是数据处理篇。 代码复现需要的文件包,原论文都会提供的有,其[GitHub链接](nlpyang/BertSum) 数据集因为谷歌云盘下载起来比较麻烦,像bert_data、cnn/dailymail数据集在下方可寻。 一、环境要求 pyt...
代码链接:github.com/nlpyang/Bert Introduction: 这篇论文主要关注的是BERT在抽取式摘要(extractive summarization)上的应用,作者尝试使用了不同的Bert 不同的变种对于CNN/Dailymail还有NYT datasets 的影响,并且发现flat architecture with inter-sentence Transformer layers 效果最好,达到了当前最佳 Methodology: 令d表示...
Some codes are borrowed from ONMT(https://github.com/OpenNMT/OpenNMT-py) Data Preparation For CNN/Dailymail Option 1: download the processed data downloadhttps://drive.google.com/open?id=1x0d61LP9UAN389YN00z0Pv-7jQgirVg6 unzip the zipfile and put all.ptfiles intobert_data ...
论文笔记 _ Discourse-Aware Neural Extractive Text Summarization 作者:韩 单位:燕山大学 论文地址:https://www.aclweb.org/anthology/2020.acl-main.451/ 代码地址:https://github.com/jiacheng-xu/DiscoBERT 目录 一、文本摘要(Text Summarization )任务 1.1 任务概述 1.2 抽取式方法 1.3 生成式方法 1.3 ...
The results show the superiority of our model in terms of ROUGE scores. (The code is available at https://github.com/atharsefid/SciBERTSUM ).Sefid, AtharPennsylvania State UniversityGiles, C. LeePennsylvania State UniversitySpringer, ChamInternational Workshop on Document Analysis Systems...
GitHub is where people build software. More than 150 million people use GitHub to discover, fork, and contribute to over 420 million projects.