Radar Station Supplementary Material 1/2
收藏资源简介:
# Radar Station data file. <br> <br> This repo contains the data necessary for the paper 'Radar Station: Using KG Embeddings for Semantic Table Interpretation and Entity Disambiguation.'<br> <br> The second part of the data is available at https://zenodo.org/record/6522921<br> <br> Its structure of it is as follows:<br> <br> '''<br> ├─readme.md<br> ├─DAGOBAHSL_Scoring<br> │ ├─Limaye_Result<br> │ ├─T2D_Result<br> │ ├─2T_Result<br> │ ├─ShortTable_Result<br> ├─Embeddings<br> │ ├─RotetE<br> │ ├─ComplEx<br> │ ├─TransE<br> │ ├─Dismult<br> ├─Datasets<br> │ ├─Key_Column_Index<br> │ │ ├─Limaye<br> │ │ ├─T2D<br> │ ├─ShortTable<br> ├─Wikidata_Ground_Truth<br> │ ├─Wikidata_GS_Limaye<br> │ ├─Wikidata_GS_T2D<br> │ ├─Wikidata_GS_2T<br> │ ├─Wikidata_GS_ShortTable<br> '''<br> <br> ## DAGOBAHSL_Scoring<br> It contains the result of the candidate scoring step after the four datasets were processed through a previous annotation system.<br> In this score step, we did not filter any candidates during the calculation and always kept their scores. <br> <br> <br> ## Embeddings<br> <br> This folder contains four embeddings used during our experiment.<br> We provide four embeddings for the experiments: TransE, RotatE, Dismult, and ComplEx. <br> In which the RotatE embeddings are from the pre-trained embeddings of GraphVite: https://graphvite.io/docs/latest/pretrained_model.html <br> TransE, Dismult, and ComplEx are trained using Pytorch-BigGraph with Wikidata dump version 2021 May.<br> <br> **NOTICE**: the Dismult and ComplEx Embeddings should be further added to this folder.<br> <br> <br> ## Datasets<br> ### Key_Column_Index<br> <br> Key_Column_Index folder contains the index of the key column position for T2D and Limaye that are manually annotated.<br> It tells Radar Station in which column we should run.<br> <br> ### ShortTable<br> <br> It contains the ShortTable dataset with tables of only two rows.<br> <br> ## Wikidata_Ground_Truth<br> It contains the ground truth of the four datasets with Wikidata entities.
# 雷达站(Radar Station)数据集文件 本仓库包含论文《雷达站:基于知识图谱嵌入的语义表格解析与实体消歧》所需的全部数据集。 数据集的第二部分可通过以下链接获取:https://zenodo.org/record/6522921 数据集的目录结构如下: ├─readme.md ├─DAGOBAHSL_Scoring │ ├─Limaye_Result │ ├─T2D_Result │ ├─2T_Result │ ├─ShortTable_Result ├─Embeddings │ ├─RotetE │ ├─ComplEx │ ├─TransE │ ├─Dismult ├─Datasets │ ├─Key_Column_Index │ │ ├─Limaye │ │ ├─T2D │ ├─ShortTable ├─Wikidata_Ground_Truth │ ├─Wikidata_GS_Limaye │ ├─Wikidata_GS_T2D │ ├─Wikidata_GS_2T │ ├─Wikidata_GS_ShortTable ## DAGOBAHSL_Scoring 该目录包含经前期标注系统处理后的四个数据集的候选集评分步骤结果。在本次评分计算过程中,我们未对任何候选集进行过滤,完整保留了所有候选集的评分结果。 ## 嵌入模型(Embeddings) 该目录包含本次实验中使用的四种嵌入模型文件。我们提供的嵌入模型分别为:TransE、RotatE、Dismult与ComplEx。其中,RotatE嵌入模型来源于GraphVite的预训练嵌入模型,获取链接为:https://graphvite.io/docs/latest/pretrained_model.html。TransE、Dismult及ComplEx嵌入模型则基于2021年5月版的维基数据(Wikidata)快照,通过Pytorch-BigGraph训练得到。 **注意事项**:本目录尚需补充Dismult与ComplEx嵌入模型文件。 ## 数据集(Datasets) ### 关键列索引(Key_Column_Index) 该目录包含经人工标注的T2D与Limaye数据集的关键列位置索引,用于告知雷达站(Radar Station)模型应在表格的哪一列执行任务。 ### 短表格数据集(ShortTable) 该目录包含仅含两行记录的ShortTable数据集。 ## 维基数据(Wikidata)基准真值集 该目录包含四个数据集对应维基数据实体的基准真值标注结果。



