[RID] Research Infrastructure Dataset for NER classification

Dataset used for NER classification for the task of Research Infrastructure (RI) Detection in Scientific articles. Sentences originated from scientific publication in various scientific fields.
Dataset contains 354 sentences (split 231 train_set, 123 test_set).
Structure ->
"sentence" : Sentence data sample (context is about usage of an RI)
"token_sen" : Sentence split into tokens using TreeBankWordTokenizer
"word_labels" : Token labels (produced from token_sen column) in BIO format.
"nltk_pos_tag" : POS tag of each of the tokens (produced from token_sen column)
"RIs" : List containing the substring of each RI in the sentence
"n_RIs" : number of RIs in the sentence [count(RIs)]

Using this dataset we benchmarked various NER solutions for optimal RI detection in text (in a few-shot environment).

For more information please read our paper published at LOD2023 :
Benchmarking Named Entity Recognition Approaches for Extracting Research Infrastructure Information from Text
[Link placeholder]