aravindsankar28 / Academic-Concept-Extractor

C++ Implementation of Academic Concept Extraction framework

Unsupervised Academic Concept Extraction

This is a C++ implementation of the Academic Concept Extraction framework as described in our paper,

Adit Krishnan^*, Aravind Sankar^*, Shi Zhi and Jiawei Han [Unsupervised Concept Categorization and Extraction from Scientific Document Titles](https://arxiv.org/pdf/1710.02271.pdf) (CIKM 2017)

Requirements

• g++

PhraseType demo

g++ phraseType.cpp -o phraseType.o
./phraseType.o sample.txt

The probability distributions inferred by the Gibbs sampler will be written to the PhraseType/ directory.

DomainPhraseType demo

g++ domainPhraseType.cpp -o domainPhraseType.o
./domainPhraseType.o sample.txt

The probability distributions inferred by the Gibbs sampler will be written to the DomainPhraseType/ directory.

Data

In order to use your own data, you have to provide a text file of phrases with the following details for each phrase.

• Left relation phrase - lrp
• List of words (comma-separated) words
• List of significant phrases (comma-separated) sig_phrases
• Right relation phrase rrp
• Venue venue

Each line in the text file contains a single input phrase of the format:

lrp$words$sig_phrases$rrp$venue

For example, in a title bootstrapped named entity recognition for product attribute extraction, the phrase bootstrapped named entity recognition may be given as:

empty$bootstrapped,named,entity,recognition$entity_recognition$for$emnlp

An efficient Pitman-Yor Adaptor Grammar implementation is available at http://web.science.mq.edu.au/~mjohnson/Software.htm.

Cite

Please cite our paper if you use this code in your own work:

@article{krishnan2017unsupervised,
  title={Unsupervised Concept Categorization and Extraction from Scientific Document Titles},
  author={Krishnan, Adit and Sankar, Aravind and Zhi, Shi and Han, Jiawei},
  journal={arXiv preprint arXiv:1710.02271},
  year={2017}
}