A modification of PageRank to find the most prestigious authors in a scientific collaboration network.
A key question in the analysis of collaborative networks is: "Who are the most prestigious authors?" Answering this question can be useful in identifying subject matter experts or in ranking search results. This library implements AuthorRank - an approach by Liu et. al. to determine the most prestigious authors across a research network utilizing a body of research papers.
- Python 3.5 - 3.11
- networkx == 2.6.3
- python-utils == 3.5.2
- scipy == 1.7.3
First install the package locally:
pip install author-rank # or pip3 install ... if you're using both Python 3 and 2.Then, assuming you have input data in a format like the following:
documents = [
{
"title": "Always lurking: understanding and mitigating bias in online human trafficking detection",
"authors": [
{
"first_name": "Kyle",
"last_name": "Hundman",
"affiliation": {
"name": "NASA Jet Propulsion Laboratory",
"department": "Office of the Chief Information Officer"
}
},
{
"first_name": "Thamme",
"last_name": "Gowda",
"affiliation": {
"name": "University of Southern California"
}
},
{
"first_name": "Mayank",
"last_name": "Kejriwal",
"affiliation": {
"name": "University of Southern California"
}
},
{
"first_name": "Benedikt",
"last_name": "Boecking",
"affiliation": {
"name": "Carnegie Mellon University"
}
}
]
},
{
"title": "Measurement Context Extraction from Text: Discovering Opportunities and Gaps in Earth Science",
"authors": [
{
"first_name": "Kyle",
"last_name": "Hundman",
"affiliation": {
"name": "NASA Jet Propulsion Laboratory",
"department": "Office of the Chief Information Officer"
}
},
{
"first_name": "Chris",
"last_name": "Mattmann",
"affiliation": {
"name": "University of Southern California",
"department": "Computer Science Department"
}
}
]
},
{
"title": "An approach for automatic and large scale image forensics",
"authors": [
{
"first_name": "Thamme",
"last_name": "Gowda",
"affiliation": {
"name": "University of Southern California"
}
},
{
"first_name": "Kyle",
"last_name": "Hundman",
"affiliation": {
"name": "NASA Jet Propulsion Laboratory",
"department": "Office of the Chief Information Officer"
}
},
{
"first_name": "Chris",
"last_name": "Mattmann",
"affiliation": {
"name": "University of Southern California",
"department": "Computer Science Department"
}
}
]
}
]
One can compute retrieve a ranked list of authors and their scores according to the AuthorRank algorithm:
# import AuthorRank
import author_rank as ar
# create an AuthorRank object
ar_graph = ar.Graph()
# fit to the data
ar_graph.fit(
documents=documents
)
# get the top authors for a set of documents
ar_graph.top_authors(normalize_scores=True, n=10)Setting normalized_scores to True normalizes the AuthorRank scores
on a scale of 0 to 1 (inclusive), which may be helpful for interpretation.
By default, AuthorRank looks for a list of authors - with each author
represented as a dictionary of keys and values - from each document
in the list of documents passed into fit using
the key authors, with the keys first_name and last_name as the
keys used to uniquely identify each author. However, if desired other keys
could be specified and utilized, as in the example below:
ar_graph.fit(
documents=documents,
authorship_key="creators",
keys=set(["given", "family"])
)
ar_graph.top_authors(normalize_scores=True, n=10)It is also possible to export the directed graph from the provided input data, with weights, into a JSON format for use in visualization or additional analysis:
export = ar_graph.as_json()
print(json.dumps(export, indent=4))When creating the AuthorRank graph, the progress_bar
parameter can be used to indicate the progress of applying AuthorRank to
a set of documents. This can be helpful when processing larger corpora
of documents as it provides a rough indication of the remaining time
needed to complete execution.
# fit to the data
ar_graph.fit(
documents=documents,
progress_bar=True
)Make sure to download the CORD-19 dataset and adjust the file paths in the examples appropriately for your dataset version (indicated by a date). The dataset is too large to be included in the repository and thus needs to be downloaded separately.
To get started with the examples in the repository (using the same version of the CORD-19 dataset), run the following:
cd data
mkdir CORD-19
cd CORD-19
wget https://ai2-semanticscholar-cord-19.s3-us-west-2.amazonaws.com/historical_releases/cord-19_2020-07-16.tar.gz
tar -xf cord-19_2020-07-16.tar.gz
rm cord-19_2020-07-16.tar.gz
cd ../..
This library (of the same name) implements AuthorRank [1]. AuthorRank is a modification of PageRank, Google's original algorithm for ranking webpage search results. PageRank works on the idea of transferred status. The rank of a page is the sum of the ranks of its backlinks - if a webpage has many backlinks or a few highly ranked backlinks, its rank is also high. The algorithm works over a directed graph in which nodes are webpages and a directed edge represents a link from one page to another. It is assumed that each node transfers its rank evenly to all of the other nodes it connects to. Instead of webpages, AuthorRank creates a co-authorship network that represents the structure of scientific collaborations and the status of individual researchers. In the network, each node represents an author and each edge represents a collaboration. Edges are bi-directional to represent the symmetric nature of collaboration. Unlike PageRank in which each node is assumed to transfer status equally, when considering status in a collaboration, greater status should be given to authors who frequently coauthor together, and status should be diminished as the number of authors in a paper increases. Thus, edges are weighted according to frequency of co-authorship and total number of co-authors on articles according to the diagram shown below.
The applicability of this approach is not confined to research collaborations and this module could be extended into other useful applications utilizing similar directed graphs (for example, it may be interesting to examine the Enron email corpus).
Please use the issue tracker to report any erroneous behavior or desired feature requests.
If you would like to contribute to development, please fork the repository and make
any changes to a branch which corresponds to an open issue. Hot fixes
and bug fixes can be represented by branches with the prefix fix/ versus
feature/ for new capabilities or code improvements. Pull requests will
then be made from these branches into the repository's dev branch
prior to being pulled into main. Pull requests which are works in
progress or ready for merging should be indicated by their respective
prefixes ([WIP] and [MRG]). Pull requests with the [MRG] label will be
reviewed prior to being pulled into the main branch.
Tests are contained within the tests directory. To run tests for
AuthorRank, call pytest and pytest-cov via the command line:
python3 -m pytest --cov=author_rank -vvThe tests included within the repository are automatically run on commit to repository branches and any external pull requests using Travis CI.
Semantic versioning is used for this project. If contributing, please conform to semantic versioning guidelines when submitting a pull request.
This project is licensed under the MIT license.
- Xiaoming Liu, Johan Bollen, Michael L. Nelson, Herbert Van de Sompel, Co-authorship networks in the digital library research community, Information Processing & Management, Volume 41, Issue 6, 2005, Pages 1462-1480, ISSN 0306-4573, http://dx.doi.org/10.1016/j.ipm.2005.03.012. Pre-print PDF.