Skip to content
Implementation of my paper "Reference Scope Identification for Citances Using Convolutional Neural Network"
Branch: master
Clone or download


This repository contains all files necessary to reproduce the results of our paper "Reference Scope Identification for Citances Using Convolutional Neural Network1".

Reference-Citance Pair Extraction: along with test*.py are meant for data set parsing.

Stopwords removal: writes the <RP, CP> pair into tab-separated csv cells along with assigning corresponding binary labels (1- true, 0 - false).

Feature Extraction Modules:

  1. Lexical Features: All the similarity measures require a pair of texts as input and work by averaging over all the words of the sentence.
  1. Knowledge-based Feature: measures the best semantic similarity score between words in the citance and the reference sentence out of all the sets of cognitive synonyms (synsets) present in the WordNet.

  2. Corpus-based Feature:, as the name denotes.

  3. Surface Features:

  • measures the overall positive and negative sentiment score of the reference sentence averaged over all the words, based on the SentiWordNet 3.0 lexical resource.
  • measures lexical richness of the reference sentence based on Yule’s K index.

Classification Algorithm: contains the 1-D CNN implementation along with GBC and ABC classifiers for training and testing the above generated feature vectors.

The original presentation can be found here.

In case of queries, feel free to reach out at


  • S. Jha, A. Chaurasia, A. Sudhakar, and A. K. Singh, “Reference scope identification for citances using convolutional neural networks,” in Proceedings of the 14th International Conference on Natural Language Processing (ICON-2017). Kolkata, India: NLP Association of India, December 2017, pp. 23–32. [Online]. Available:
You can’t perform that action at this time.