A science map shows the structure of a research field: its main themes, how they connect and how they change. Building one used to require expensive database subscriptions. Today you can do it for free with OpenAlex, an open catalogue of more than 200 million scholarly works, and VOSviewer, a free desktop tool from Leiden University for building and visualising bibliometric networks.
This guide walks through a complete workflow, using misinformation research as an example.
Step 1: Define the field with a query
Everything depends on the set of papers you start with. Write down:
- Concepts and synonyms: misinformation, disinformation, “fake news”, rumour/rumor, infodemic, fact-checking.
- Time window: for example 2010 to the present.
- Document types: articles and reviews, usually excluding editorials.
- Languages: English only, or more? Remember that coverage differs by language.
Test the query by checking whether well-known key papers are included, and by reading a random sample of 20 results. Are they relevant? Adjust until both checks look right.
Step 2: Retrieve the records from OpenAlex
OpenAlex offers a web interface and a free API.
- Web interface: search, apply filters (year, type, language) and download results as a CSV file. This is good for exploring and for sets up to the download limit.
- API: for larger sets, query the
worksendpoint with a search filter and page through the results. The documentation explains the filters, and adding your email to requests gets you into the faster “polite pool”.
Useful fields to keep: title, abstract, publication year, authors and institutions, concepts/topics, referenced works and citation count.
Step 3: Clean the data
- Remove duplicates and records with missing titles.
- Check institution and country names for variants.
- Build a thesaurus file for VOSviewer to merge synonyms (“rumor” = “rumour”, “covid-19” = “covid”) and remove meaningless terms (“study”, “paper”, “results”).
This step takes longer than expected and has more influence on the final map than any software setting.
Step 4: Build networks in VOSviewer
VOSviewer can read bibliographic data and build several kinds of maps:
| Map type | Nodes | Links | Shows |
|---|---|---|---|
| Co-occurrence of terms | Keywords or title/abstract terms | Terms used together | Themes in the field |
| Co-citation | Cited references or sources | Cited together | Intellectual foundations |
| Bibliographic coupling | Documents, sources, authors | Share references | Current research fronts |
| Co-authorship | Authors, institutions, countries | Joint papers | Collaboration structure |
For a first overview, start with a term co-occurrence map built from titles and abstracts:
- Choose “Create a map based on text data” and load your file.
- Choose binary counting, so a term counts once per document.
- Set a minimum number of occurrences, so that a few hundred terms remain.
- Review the term list and deselect generic terms, or add them to your thesaurus.
- Generate the map. VOSviewer clusters related terms and colours each cluster.
Step 5: Interpret, do not just display
A map is a starting point for analysis, not a finding in itself.
- Name each cluster by reading its top terms and, crucially, several of its most cited papers.
- Use the overlay view to colour terms by average publication year. This reveals emerging and declining themes.
- Look at the bridges: terms or papers that connect clusters often mark interdisciplinary work.
- Check robustness. Does the structure survive if you change the minimum occurrence threshold or the time window?
In a misinformation map, you might find clusters around psychology of belief and correction, social-media diffusion and network analysis, health misinformation and the COVID-19 infodemic, political disinformation and elections, and automated detection with machine learning. The overlay would likely show the detection and AI-related themes as the most recent.
Step 6: Report transparently
State the query, database, date of retrieval, cleaning steps, thresholds and software settings. Without these, nobody can reproduce or judge your map.
Limits to keep in mind
- Coverage is never complete. OpenAlex is broad, but metadata quality varies, especially for older and non-English records.
- Terms are not concepts. The same word can mean different things in different fields.
- Maps reflect choices. Different queries and thresholds give different pictures. Present the map as one view of the field, not the only one.
Used carefully, these free tools let a single analyst produce in a day what once took a team weeks: a clear picture of who is researching what, how topics connect and where a field is heading.