Sentiment Piecewise Convolutional Neural Network

This project represents an implementation of PCNN [zeng2015distant], dubbed as Piecewise Convolutional Neural Network, written in Tensorflow. Considered as an application for sentiment attitudes extraction task.

Architectures implementation:

• [cnn]
• [pcnn]

Model configuration parameters:

• [base]
• [cnn/pcnn]

Changeset

The master branch represents the latest implementation. Experiment details presented in other branches. The historical changeset is as follows:

• 2018 -- Original implementation, classification is strongly by contexts:
  • DAMDID-2018 Conference [paper] [code]
  • RUSSIR-2018 Conference (Not published) [poster] [paper] [code]
  • CLLS-2018 Conference [paper] [code]
• 2019 -- Implementation has been significantly refactored. Application of aggregation function of related contexts with different functions, especially with RNN application:
  • CCIS (vol. 1003) Journal [code]

Dataset

We use RuSentRel 1.0 corpus consisted of analytical articles from Internet-portal inosmi.ru.

Related works

• Daojian Zeng, Kang Liu, Yubo Chen, and Jun Zhao, Distant supervision for relation extraction via piecewise convolutional neural networks, Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, 2015, pp. 1753–1762 [paper]