wosfile is a Python package designed to read and handle data exported from Clarivate Analytics Web of Science™. It supports both tab-delimited files and so-called ‘plain text’ files.
The point of wosfile is to read export files from WoS and give you a simple data structure—essentially a dict—that can be further analyzed with tools available in standard Python or with third-party packages. If you're looking for a ‘one-size-fits-all’ solution, this is probably not it.
Pros:
- It has no requirements beyond Python 3.6+ and the standard library.
- Completely iterator-based, so useful for working with large datasets. At no point should we ever have more than one single record in memory.
- Simple API: usually one needs just one function
wosfile.records_from()
.
Cons:
- Pure Python, so might be slow.
- At the moment, wosfile does little more than reading WoS files and generating Record objects for each record. While it does some niceties like parsing address fields, it does not have any analysis functionality.
These examples use a dataset exported from Web of Science in multiple separate files. The maximum number of exported records per file is 1000 (or 5000, depending on how much metadata you need).
import glob
import wosfile
from collections import Counter
subject_cats = Counter()
# Create a list of all relevant files. Our folder may contain multiple export files.
files = glob.glob("data/savedrecs*.txt")
# wosfile will read each file in the list in turn and yield each record
# for further handling
for rec in wosfile.records_from(files):
# Records are very thin wrappers around a standard Python dict,
# whose keys are the WoS field tags.
# Here we look at the SC field (subject categories) and update our counter
# with the categories in each record.
subject_cats.update(rec.get("SC"))
# Show the five most common subject categories in the data and their number.
print(subject_cats.most_common(5))
For this example you will need the NetworkX package. The data must be exported as ‘Full Record and Cited References’.
import networkx as nx
import wosfile
# Create a directed network (empty at this point).
G = nx.DiGraph()
nodes_in_data = set()
for rec in wosfile.records_from(files):
# Each record has a record_id, a standard string uniquely identifying the reference.
nodes_in_data.add(rec.record_id)
# The CR field is a list of cited references. Each reference is formatted the same
# as a record_id. This means that we can add citation links by connecting the record_id
# to the reference.
for reference in rec.get("CR", []):
G.add_edge(rec.record_id, reference)
# At this point, our network also contains all references that were not in the original data.
# The line below ensures that we only retain publications from the original data set.
G.remove_nodes_from(set(G) - nodes_in_data)
# Show some basic statistics and save as Pajek file for visualization and/or further analysis.
print(nx.info(G))
nx.write_pajek(G, 'network.net')
The following packages also read WoS files (+ sometimes much more):
- WOS+
- metaknowledge
- python-bibx
- wostools (replaced by python-bibx)
Other packages query WoS directly through the API and/or by scraping the web interface:
- pywos (elsewhere called wos-statistics)
- wos
- wosclient