SIRIS-Lab/topic-modeling-maps
收藏资源简介:
该数据集名为主题建模参考地图,是一个预保存的主题建模地图集合,专为配合topic-modeling Python包设计。每个地图包含一个保存的BERTopic模型、其2-D UMAP投影器和一个精炼后的标签CSV文件。用户可以将新文档直接投影到这些地图上,以预测主题和2-D坐标,而无需重新训练模型。数据集中的地图基于SPECTER嵌入和KMeans聚类构建,例如eu_map_60_topics地图包含60个主题,并提供了精炼的标签和高级聚类信息。数据集适用于主题建模、科学计量学等任务,支持离线使用和自定义路径加载。
The dataset is named Topic-modelling reference maps and consists of pre-saved topic-modelling maps for use with the topic-modeling Python package. Each map includes a saved BERTopic model, its 2-D UMAP projector, and a refined-labels CSV file. It allows users to drop new documents onto a map to predict topics and 2-D coordinates without retraining. The maps are built using SPECTER embeddings and KMeans clustering, with examples like eu_map_60_topics featuring 60 topics, refined labels, and higher-level clusters. It is designed for applications in topic modeling and scientometrics, supporting offline usage and custom path loading.




