CCL23 古籍命名实体识别评测——封闭赛道

本次古籍文献的命名实体识别评测，通过发布全新的基于“二十四史”的训练和测试数据集，提供统一的评测提交平台，以此推动技术的突破和发展，助力古籍资源的智能开发与利用。
<br />
官方网址：<a href="https://guner2023.pkudh.org/" target="_blank">https://guner2023.pkudh.org/</a>
<br /><br />
请注意，此页面是封闭赛道的评测提交入口，要求参赛队伍禁止使用 ChatGPT、文心一言、ChatGLM 等大模型。仅允许使用拥有开源License（如 GPL、BSD、MIT、Apache 等）且参数量小于 10B 的预训练模型。
开放赛道的评测提交入口为：<a href="https://tianchi.aliyun.com/dataset/151111" target="_blank">https://tianchi.aliyun.com/dataset/151111</a>。要求参赛队伍使用ChatGPT、文心一言、ChatGLM 等大模型。
<br /><br />
请注意：评测榜单提交时，“组织”一栏请务必填写报名时的队伍名。若之前提交时填写有误，可联系我们撤回。

This named entity recognition (NER) evaluation for ancient Chinese literature releases a brand-new training and test dataset based on the Twenty-Four Histories, and provides a unified evaluation submission platform, so as to promote technological breakthroughs and development, and facilitate the intelligent development and utilization of ancient Chinese literature resources.
<br />
Official website: <a href="https://guner2023.pkudh.org/" target="_blank">https://guner2023.pkudh.org/</a>
<br /><br />
Please note that this page is the submission portal for the Closed Track. Participating teams are prohibited from using Large Language Models (LLMs) such as ChatGPT, Wenxin Yiyan, and ChatGLM. Only pretrained models with open-source licenses (e.g., GPL, BSD, MIT, Apache, etc.) and parameter sizes less than 10 billion are permitted.
<br /><br />
The submission portal for the Open Track is: <a href="https://tianchi.aliyun.com/dataset/151111" target="_blank">https://tianchi.aliyun.com/dataset/151111</a>. Participating teams are required to use Large Language Models (LLMs) such as ChatGPT, Wenxin Yiyan, and ChatGLM.
<br /><br />
Please note: When submitting to the evaluation leaderboard, please be sure to fill in the team name used during registration in the "Organization" field. If you filled in the information incorrectly in previous submissions, you may contact us to withdraw your submissions.

搜集汇总

背景与挑战

背景概述

该数据集是CCL23古籍命名实体识别评测的封闭赛道部分，基于二十四史构建，标注了人名、书名和官职名等实体，用于训练和测试模型。评测禁止使用大模型，仅允许开源小规模预训练模型，旨在通过统一平台推动古籍资源智能开发，评价指标包括准确率、召回率和F1值。