Update documents

Author: cuihaoleo

README.md

@@ -10,9 +10,11 @@ This repository hosts the source code for PoliGraph, including:
 - Dataset preparation scripts under `datasets/`.
 - Model training scripts under `models/`.
 
+PoliGraph is part of [the Policy-Technology project](https://athinagroup.eng.uci.edu/projects/auditing-and-policy-analysis/) of the UCI Networking Group.
+
 ## Citation
 
-If you create a publication based on PoliGraph, please cite the paper as follows:
+If you create a publication based on PoliGraph and/or its dataset, please cite the paper as follows:
 
 ```bibtex
 @inproceedings{cui2023poligraph,
@@ -31,11 +33,11 @@ We have tested all the code in this repository on a server with the following co
 - GPU: NVIDIA RTX A5000 (24 GiB of video memory)
 - OS: Debian GNU/Linux 11 (bullseye)
 
-A Linux server with 32 GiB of memory, 20 GiB of free disk space (after installing conda), and a similar NVIDIA GPU should suffice to run everything. Note that PoliGraph-er can run without a GPU, but the performance would be significantly lower.
+A Linux server with 32 GiB of memory, 20 GiB of free disk space (after installing conda), and a similar NVIDIA GPU should suffice to run everything. A GPU is required to enable hardware acceleration for [transformer-based NLP models](https://spacy.io/usage/embeddings-transformers). Note that PoliGraph-er can run without a GPU, but the performance would be significantly lower.
 
 ## PoliGraph-er
 
-PoliGraph-er is the software used to generate PoliGraphs from the text of a privacy policy.
+PoliGraph-er is the NLP software used to generate PoliGraphs from the text of a privacy policy.
 
 ### Installation
 
@@ -102,7 +104,15 @@ Lastly, execute the `build_graph` script to generate the PoliGraph:
 $ python -m poligrapher.scripts.build_graph example/
 ```
 
-The generated graph is stored at `example/graph-original.yml`.
+The generated graph is stored at `example/graph-original.yml`. You may use a text editor to view it. The format is human-readable and fairly straightforward.
+
+Alternatively, if you run `build_graph` with the `--pretty` parameter, it will generate a PoliGraph in the GraphML format (`example/graph-original.graphml`), which can be imported to some graph editor software:
+
+```sh
+$ python -m poligrapher.scripts.build_graph --pretty example/
+```
+
+For more instructions on how to view the graphs, please refer to the document [Viewing a PoliGraph](./docs/view-poligraph.md).
 
 ### Batch Processing
 
@@ -116,19 +126,11 @@ $ python -m poligrapher.scripts.build_graph dataset/policy1 dataset/policy2 data
 
 If all the subdirectories under `dataset` contain valid crawled webpages, you may simply supply `dataset/*` to let the shell expand the arguments.
 
-### View PoliGraph
-
-You may use a text editor to view `graph-original.yml`. The format is human-readable and fairly straightforward.
-
-Alternatively, you can run `build_graph` with the `--pretty` parameter:
-
-```sh
-$ python -m poligrapher.scripts.build_graph --pretty example/
-```
+## PoliGraph Dataset
 
-This generates a PoliGraph in the GraphML format (`example/graph-original.graphml`), which can be imported to [yEd](https://www.yworks.com/products/yed), a GUI graph editor. For more instructions, refer to the document [Viewing a PoliGraph](./docs/view-poligraph.md).
+We released the privacy policy dataset used in our paper for reproducibility and other research usages. Please visit [the dataset page](https://athinagroup.eng.uci.edu/projects/auditing-and-policy-analysis/poligraph-dataset/) to request access to the dataset.
 
-## Documentation
+## Artifact Evaluation
 
 We will update the documentation under the `docs/` directory to explain the usage of other scripts.
 

docs/usenix-artifact-evaluation-additional.md

@@ -66,6 +66,12 @@ $ xsv search -s labels '\bCONFLICT\b' contradictions/policylint-reclassify.csv |
 
 In this experiment, we compare each app's data flows (i.e., the actual data collection practices observed in the network traffic) to the data collection statements inferred from its PoliGraph.
 
+Change to the `~/dataset` directory before proceeding:
+
+```
+$ cd ~/dataset
+```
+
 Step 1. Run PoliGraph-er with the historical (2019) version of privacy policies from the PoliCheck dataset to create PoliGraphs for them:
 
 ```

docs/usenix-artifact-evaluation.md

@@ -2,11 +2,11 @@
 
 This document provides instructions on how to reproduce the main results in our paper in support of the USENIX Security 2023 Artifact Evaluation.
 
-Please note that these instructions are intended for artifact reviewers, who have been given access to our test server.
+Please note the exact commands shown in this document are intended for artifact reviewers, who have been given access to a preconfigured environment. For other users, please check [README.md](../README.md) at the top level of this repository for general information about the use of our code.
 
 ## Overall Instruction
 
-For artifact reviewers, we provide access to our test server which has the recommended hardware configuration. Please check HotCRP comments for instructions on how to access the server via SSH.
+For artifact reviewers, we provide access to our test server which has the recommended hardware configuration. Please check the "Artifact access" information on HotCRP for instructions on how to access the server via SSH.
 
 We have prepared the following directories in the home directory:
 
@@ -334,4 +334,4 @@ The row indicates that the data type "technical information" is defined in 126 a
 
 ## Additional Experiments
 
-This document only covers the four experiments mentioned in our USENIX'23 Artifact Appendix. We provide another document [Artifact Evaluation (Additional Experiments)](./usenix-artifact-evaluation-additional.md) to show how to reproduce results in Sections 5.3 and 5.4 of our main paper.
+Corresponding to our USENIX '23 Artifact Appendix, this document covers experiments in Sections 4.2 (Evaluation), 5.1 (Policies Summarization), and 5.2 (Correct Definitions of Terms), which we believe showcase the main merits of PoliGraph. For those interested in Sections 5.3 (Contradiction Analysis) and 5.4 (Data Flow-to-Policy Consistency Analysis), we provide another document [Artifact Evaluation (Additional Experiments)](./usenix-artifact-evaluation-additional.md) to show how to reproduce the main results there.