This repository provides three types of Neural Networks that can be used to solve the Context Independent Claim Detection. We employ our models on the IBM dataset and for each of them we consider the pretrained word embeddings built with Glove model. For the Tree-LSTM model we follow the code of the Stanford Tree-Structured Long Short-Term Memory Networks.
-
LSTM implementation of the LSTM: the model is defined in the file lstm.py, scores.py is used to evaluate the model. The considered topics are listed in considered_topic.txt.
-
RNN implementation of the RNN: the model is defined in the file rnn.py, scores_and_charts.py is used to evaluate the model.
-
Tree-LSTM contains the changes made to the Tree-Structured Long Short-Term Memory Networks to fit their implementation to our task.
- Tensorflow - Long Short Term Memory and Recursive Neural Network
- Torch - Tree Structured Long Short Term Memory
- Argumentation Mining (https://dl.acm.org/citation.cfm?id=2850417)
- Tree-structured Long Short-Term Memory networks (https://arxiv.org/abs/1503.00075)