LDkp Dataset

LDkp (Long Document keyphrase) dataset is the first benchmark corpus of 1.3M documents for identifying keyphrases from long documents.The LDkp dataset is released in two versions : LDkp3k consists of 0.1M keyphrase tagged long documents, is created using keyphrases from KP20k (Meng et al., 2017) and their corresponding long document text from S2ORC (Lo et al., 2020). LDkp10k consists of 1.3M long documents along with target keyphrases is created using keyphrases from OAGKX (Çano, 2019) and their corresponding long document text from S2ORC (Lo et al., 2020)