This is a repository implementing Latent Semantic Summarization from this paper and some form of abstractive summarization using this short guide. You can find a copy of my report summarizing understanding of the implementation and my comparison between the two methods in the repo, and you can acces my video demoing my implementation here
I've included a YAML file which holds all the packages that I've installed to implement these text summarization systems. The file also includes other packages that were installed for convenience and simplicity and may be uninstalled/edited out of the YAML file.
To construct a conda environment from these packages, simply install Anaconda or miniconda by following the instructions here
After installing Anaconda or miniconda run
conda env create --file=ai.yaml
To run the python script with your own text segment or document simply edit the sample strings provided to hold your own document and run the script with
python3 cross-extraction/lsa-cross.py
Please note that the LSA cross method implementation requires no newlines to be present in the string to be able to parse into sentences.
To run the abstractive summarization code again edit the python file to include your own text and run
python3 transformer/transformer.py