document reference context inference

CONTRIBUTING.md

@@ -49,7 +49,7 @@ All commits shall pass the lint pipeline of the following tools:
 - Mypy (see [pyproject.toml](https://github.com/crocs-muni/sec-certs/blob/main/pyproject.toml) for settings)
 - Ruff (see [pyproject.toml](https://github.com/crocs-muni/sec-certs/blob/main/pyproject.toml) for settings)
 
-These tools can be installed via [dev_requirements.txt](https://github.com/crocs-muni/sec-certs/blob/main/dev_requirements.txt) You can use [pre-commit](https://pre-commit.com/) tool to register git hook that will evalute these checks prior to any commit and abort the commit for you. Note that the pre-commit is not meant to automatically fix the issues, just warn you.
+These tools can be installed via [dev_requirements.txt](https://github.com/crocs-muni/sec-certs/blob/main/requirements/dev_requirements.txt) You can use [pre-commit](https://pre-commit.com/) tool to register git hook that will evalute these checks prior to any commit and abort the commit for you. Note that the pre-commit is not meant to automatically fix the issues, just warn you.
 
 It should thus suffice to:
 

docs/Makefile

@@ -14,6 +14,9 @@ help:
 
 .PHONY: help Makefile
 
+linkcheck:
+	@$(SPHINXBUILD) -b linkcheck . $(BUILDDIR)/linkcheck
+
 # Catch-all target: route all unknown targets to Sphinx using the new
 # "make mode" option.  $(O) is meant as a shortcut for $(SPHINXOPTS).
 %: Makefile

docs/user_guide.md

@@ -21,3 +21,14 @@ nvd_api_key: null # or the actual key value
 ```
 
 If you aim to fetch the sources from NVD, we advise you to get an [NVD API key](https://nvd.nist.gov/developers/request-an-api-key) and set the `nvd_api_key` setting accordingly. The download from NVD will work even without API key, it will just be slow. No API key is needed when `preferred_source_nvd_datasets: "sec-certs"`
+
+
+## Infering inter-certificate reference context
+
+```{important}
+This is an experimental feature.
+```
+
+We provide a model that can predict the context of inter-certificate references based on the text embedded in the artifacts. The model output is not incorporated into the `CCCertificate` instances, but can be dumped into a `.csv` file from where it can be correlated with a DataFrame of certificate features.
+
+To train and deploy the model, it should be sufficient to change some paths and run the [prediction notebook](https://github.com/crocs-muni/sec-certs/blob/main/notebooks/cc/reference_annotations/prediction.ipynb). The output of this notebook is a `prediction.csv` file that can be loaded into the [references notebook](https://github.com/crocs-muni/sec-certs/blob/main/notebooks/cc/references.ipynb). This notebook documents the full analysis of references conducted on the Common Criteria certificates. Among others, the notebook generates some further `.csv` files that can subsequently be plotted via [plotting notebook](https://github.com/crocs-muni/sec-certs/blob/main/notebooks/cc/paper2_plots.ipynb).