


Recently, OneKE, a large model knowledge extraction framework jointly developed by Ant Group and Zhejiang University, was announced as open source and donated to the OpenKG open knowledge graph community.
Knowledge graph is one of the key technologies to achieve trustworthiness and controllability of large models. Knowledge extraction can help build domain knowledge graphs. OneKE is committed to helping researchers and developers better handle issues such as information extraction, text data structuring, and knowledge graph construction.
Extracting risk events, person entities, institutional entities, etc. through OneKE can clearly present the event context, event development trends, and correlations between entities. The well-constructed graph can help large models realize complex reasoning across entities and documents. . OneKE is bilingual in Chinese and English, supports OpenSPG and DeepKE open source frameworks, and can be used out of the box.
Large language models have significantly improved the ability of artificial intelligence systems to process world knowledge. However, real-world information is highly fragmented and unstructured, so when large language models handle information extraction tasks, they will still have poor results due to the huge difference between the extracted content and natural language expressions; in addition, natural language text information There are many ambiguities, polysemy, metaphors, etc., which bring greater challenges to the knowledge extraction task. This also leads to the fact that generative artificial intelligence represented by large language models still has problems such as insufficient reasoning ability, lack of factual knowledge, and unstable generation results, which greatly hinders the industrialization of large language models.
The unified knowledge extraction framework can significantly reduce the cost of building domain knowledge graphs and has a wide range of application scenarios. This means that by extracting structured knowledge from massive data, building high-quality knowledge graphs and establishing logical connections between knowledge elements, explainable reasoning decisions can be achieved, and it can also be used to enhance large models to alleviate illusions and improve stability. , accelerating the application of large models in vertical fields.
In the medical field, knowledge management of doctors’ experience is realized through knowledge extraction, and controllable auxiliary diagnosis and treatment and medical Q&A are constructed. In the financial field, the knowledge extraction department is used for financial indicators, risk events, causal relationships, industrial chains, etc. to achieve automatic financial research report generation, risk prediction, industrial chain analysis, etc. In government affairs scenarios, the knowledge of government affairs regulations can be realized, improving the efficiency and accurate decision-making of government affairs services.
To accelerate the industrial implementation of production-based artificial intelligence, Ant Group and Zhejiang University have established a joint knowledge graph laboratory to focus on the construction of knowledge graphs enhanced by large models, trusted and controllable generation functions of knowledge enhancement, and domain knowledge. We have launched all-round cooperation on topics such as the World Graph, with a view to establishing a controllable generation functional paradigm with two-way enhancement of large language models and knowledge graphs through joint technical research.
Ant Group and Zhejiang University jointly established and upgraded the capabilities of the Ant Bailing large model in the field of knowledge extraction, and released the Chinese-English bilingual large model knowledge extraction framework OneKE. They also open sourced a version based on LLaMA2 full-parameter fine-tuning. Test indicators show that OneKE has achieved relatively good results on multiple fully supervised and zero-sample entity/relationship/event extraction tasks.
OneKE is an excellent bilingual generalizable knowledge extraction tool in Chinese and English. It performs well on Chinese NER named entity recognition tasks, RE relationship extraction tasks, and EE event extraction tasks. Relatively good results have been achieved.
Liang Lei, head of Ant Group’s knowledge graph, said that Ant will continue to optimize the performance of knowledge extraction to serve the controllable and trustworthy needs of large models in different scenarios. In the future, we will work with industry partners to apply relevant technical systems to various vertical fields such as finance, medical care, and government affairs, and promote the industrial implementation of controllable generation technology dual-driven by knowledge graphs and large language models.
OneKE official homepage: http://oneke.openkg.cn/
OpenSPG GitHub: https://github.com/OpenSPG/openspg
The above is the detailed content of Ant Group and Zhejiang University jointly release OneKE, an open source large model knowledge extraction framework. For more information, please follow other related articles on the PHP Chinese website!

PowerInfer提高了在消费级硬件上运行AI的效率上海交大团队最新推出了超强CPU/GPULLM高速推理引擎PowerInfer。PowerInfer和llama.cpp都在相同的硬件上运行,并充分利用了RTX4090上的VRAM。这个推理引擎速度有多快?在单个NVIDIARTX4090GPU上运行LLM,PowerInfer的平均token生成速率为13.20tokens/s,峰值为29.08tokens/s,仅比顶级服务器A100GPU低18%,可适用于各种LLM。PowerInfer与

要让大型语言模型(LLM)充分发挥其能力,有效的prompt设计方案是必不可少的,为此甚至出现了promptengineering(提示工程)这一新兴领域。在各种prompt设计方案中,思维链(CoT)凭借其强大的推理能力吸引了许多研究者和用户的眼球,基于其改进的CoT-SC以及更进一步的思维树(ToT)也收获了大量关注。近日,苏黎世联邦理工学院、Cledar和华沙理工大学的一个研究团队提出了更进一步的想法:思维图(GoT)。让思维从链到树到图,为LLM构建推理过程的能力不断得到提升,研究者也通

近期,复旦大学自然语言处理团队(FudanNLP)推出LLM-basedAgents综述论文,全文长达86页,共有600余篇参考文献!作者们从AIAgent的历史出发,全面梳理了基于大型语言模型的智能代理现状,包括:LLM-basedAgent的背景、构成、应用场景、以及备受关注的代理社会。同时,作者们探讨了Agent相关的前瞻开放问题,对于相关领域的未来发展趋势具有重要价值。论文链接:https://arxiv.org/pdf/2309.07864.pdfLLM-basedAgent论文列表:

将不同的基模型象征为不同品种的狗,其中相同的「狗形指纹」表明它们源自同一个基模型。大模型的预训练需要耗费大量的计算资源和数据,因此预训练模型的参数成为各大机构重点保护的核心竞争力和资产。然而,与传统软件知识产权保护不同,对预训练模型参数盗用的判断存在以下两个新问题:1)预训练模型的参数,尤其是千亿级别模型的参数,通常不会开源。预训练模型的输出和参数会受到后续处理步骤(如SFT、RLHF、continuepretraining等)的影响,这使得判断一个模型是否基于另一个现有模型微调得来变得困难。无

FATE2.0全面升级,推动隐私计算联邦学习规模化应用FATE开源平台宣布发布FATE2.0版本,作为全球领先的联邦学习工业级开源框架。此次更新实现了联邦异构系统之间的互联互通,持续增强了隐私计算平台的互联互通能力。这一进展进一步推动了联邦学习与隐私计算规模化应用的发展。FATE2.0以全面互通为设计理念,采用开源方式对应用层、调度、通信、异构计算(算法)四个层面进行改造,实现了系统与系统、系统与算法、算法与算法之间异构互通的能力。FATE2.0的设计兼容了北京金融科技产业联盟的《金融业隐私计算

大型语言模型(LLM)被广泛应用于需要多个链式生成调用、高级提示技术、控制流以及与外部环境交互的复杂任务。尽管如此,目前用于编程和执行这些应用程序的高效系统却存在明显的不足之处。研究人员最近提出了一种新的结构化生成语言(StructuredGenerationLanguage),称为SGLang,旨在改进与LLM的交互性。通过整合后端运行时系统和前端语言的设计,SGLang使得LLM的性能更高、更易控制。这项研究也获得了机器学习领域的知名学者、CMU助理教授陈天奇的转发。总的来说,SGLang的

IBM再度发力。随着AI系统的飞速发展,其能源需求也在不断增加。训练新系统需要大量的数据集和处理器时间,因此能耗极高。在某些情况下,执行一些训练好的系统,智能手机就能轻松胜任。但是,执行的次数太多,能耗也会增加。幸运的是,有很多方法可以降低后者的能耗。IBM和英特尔已经试验过模仿实际神经元行为设计的处理器。IBM还测试了在相变存储器中执行神经网络计算,以避免重复访问RAM。现在,IBM又推出了另一种方法。该公司的新型NorthPole处理器综合了上述方法的一些理念,并将其与一种非常精简的计算运行

去噪扩散模型(DDM)是目前广泛应用于图像生成的一种方法。最近,XinleiChen、ZhuangLiu、谢赛宁和何恺明四人团队对DDM进行了解构研究。通过逐步剥离其组件,他们发现DDM的生成能力逐渐下降,但表征学习能力仍然保持一定水平。这说明DDM中的某些组件对于表征学习的作用可能并不重要。针对当前计算机视觉等领域的生成模型,去噪被认为是一种核心方法。这类方法通常被称为去噪扩散模型(DDM),通过学习一个去噪自动编码器(DAE),能够通过扩散过程有效地消除多个层级的噪声。这些方法实现了出色的图


Hot AI Tools

Undresser.AI Undress
AI-powered app for creating realistic nude photos

AI Clothes Remover
Online AI tool for removing clothes from photos.

Undress AI Tool
Undress images for free

Clothoff.io
AI clothes remover

AI Hentai Generator
Generate AI Hentai for free.

Hot Article

Hot Tools

SublimeText3 English version
Recommended: Win version, supports code prompts!

DVWA
Damn Vulnerable Web App (DVWA) is a PHP/MySQL web application that is very vulnerable. Its main goals are to be an aid for security professionals to test their skills and tools in a legal environment, to help web developers better understand the process of securing web applications, and to help teachers/students teach/learn in a classroom environment Web application security. The goal of DVWA is to practice some of the most common web vulnerabilities through a simple and straightforward interface, with varying degrees of difficulty. Please note that this software

mPDF
mPDF is a PHP library that can generate PDF files from UTF-8 encoded HTML. The original author, Ian Back, wrote mPDF to output PDF files "on the fly" from his website and handle different languages. It is slower than original scripts like HTML2FPDF and produces larger files when using Unicode fonts, but supports CSS styles etc. and has a lot of enhancements. Supports almost all languages, including RTL (Arabic and Hebrew) and CJK (Chinese, Japanese and Korean). Supports nested block-level elements (such as P, DIV),

Notepad++7.3.1
Easy-to-use and free code editor

PhpStorm Mac version
The latest (2018.2.1) professional PHP integrated development tool
