A command line information extraction tool which works with Swedish text. The tool performs various natural language processing tasks, such as parsing, named entity recognition and information extraction. Swedish text can either be input using custom files in the input_data directory, or sample data generated from the SUC 3.0 corpus in training_data.
This tool was written alongside my undergraduate level dissertation which explored Natural Language Processing with Swedish text.
In order to use the tool, various packages need to be installed. It is recommended that the package installer, pip, is used for this process. Installation can be achieved using the following command:
$ pip3 install -r requirements.txt
Once the packages are installed, the tool can be ran using the command:
$ python3 sv_information_extraction mode source
mode can be either: parse for the parser module, ner for the ner module, or ie for the information extraction module.
source can be either a filename, such as sample_swedish.txt, which is provided out of the box, or --sample which uses excerpts from SUC 3.0. In order to use the SUC 3.0 corpus, it must first be downloaded from here and placed in the training_data folder with the filename "suc3.xml".
Python 3 is the recommended version to be used with this tool, following the official deprecation of Python 2.
The tool requires a model to operate. Models can be created through custom spaCy pipelines or downloaded from the internet. The model which was used during development was the sv_model_xpos model available here. There are both UPOS and XPOS-tagged models available, with the XPOS model using Swedish-specific tags, while UPOS uses universal tags. These models have very small differences in performance between them and are both sufficient.
Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.
Please make sure to update tests as appropriate.
It is worth noting that this is a legacy-style archive, and likely won't be updated.