一种面向文化产业数据的知识抽取方法数据集
收藏资源简介：
论文名称：CT-DA: A Knowledge Extraction Method for Cultural Industry Big Data，本文提出一种融合注意力机制和对抗迁移学习的文化产业知识抽取方法（CT-DA），旨在实现低资源、边界强模糊性的文化产业大数据的知识抽取。首先，设计一种针对文化产业大数据的标注策略，将知识抽取问题转化为知识学习问题；其次，设计多层神经网络模型完成知识学习工作，针对文化产业大数据的低资源特性，本文设计对抗迁移学习层实现资源迁移，同时考虑到文化产业大数据强模糊的特性，本文设计出动态注意力机制层实现文化领域实体和关系的重点关注。最后，搭建实验平台，构建文化领域数据集。
Paper Title: CT-DA: A Knowledge Extraction Method for Cultural Industry Big Data This paper proposes a knowledge extraction method (CT-DA) for cultural industry big data that integrates attention mechanism and adversarial transfer learning, aiming to achieve knowledge extraction from low-resource cultural industry big data with strong boundary ambiguity. First, we design an annotation strategy tailored for cultural industry big data, transforming the knowledge extraction task into a knowledge learning problem. Second, we construct a multi-layer neural network model to perform knowledge learning. Considering the low-resource characteristic of cultural industry big data, we design an adversarial transfer learning layer to realize resource migration. Meanwhile, to address the strong ambiguity of such data, we propose a dynamic attention mechanism layer to focus on entities and relationships in the cultural domain. Finally, we build an experimental platform and construct a cultural domain dataset.
