This repo is for code studying topic segmentation. We propose a pipeline based on contextual model to deal with the task of topic segmentation. The coupling of each component is very low, so it is very convenient to replace some parts with other models.
The overall framework of our pipeline is shown in the figure below:
git clone https://github.com/sebastianarnold/WikiSection.git Data
Python 3.7
Pytorch
Sklearn
Others requirements please see the requirements.txt. We strongly recommend to use Anaconda.
source src/run.sh