***Update 20240307: Thanks to DavideZanutto (https://github.com/DavideZanutto) for organizing the requirements.txt, it can be found in "Issues #6" or download using this link https://github.com/hd10-iupui/AttentionRank/files/14361917/requirements.txt
***Update 20221025: Added a modified "modeling_tf2.py" to zip file "bert_update20221025" in case you are using tensorflow 2.
***Update 20221002: Fixed a computing mistake in step-011. Line 103 and line 131, move out the calculations of min and max from the loop.
This is an implement of paper AttentionRank: Unsupervised keyphrase Extraction using Self and Cross Attentions
Keyphrase Extractor can be run as below:
1, Download and extract all files.
2, Download "stanford-corenlp-full-2018-02-27" and pretrained bert-base from below link:
https://indiana-my.sharepoint.com/:f:/g/personal/hd10_iu_edu/Ep1hNQYehrlMkB734awOKhQBTv3qVVsW8iO8bMl4Vdg46Q?e=0oI0y4
3, Run "stanford-corenlp-full-2018-02-27" with terminal:
(1) cd stanford-corenlp-full-2018-02-27/
(2) java -mx4g -cp "*" edu.stanford.nlp.pipeline.StanfordCoreNLPServer -preload tokenize,ssplit,pos -status_port 9000 -port 9000 -timeout 15000 &
4, Run python files by their indices (001 - 011).