Clinical Guides: Data Collection and Storage: Text Mining

The text mining process uses Natural Language Processing (NLP) to examine large collections of documents, books, articles, websites, emails, survey results and reports. It aids in discovery of new information and helps the research literature search process.

Tools are used to analyse documents and identify facts, relationships, word usage and frequency that would otherwise remain buried in the mass of textual big data.
It can be used during the literature searching process to help build a list of keyword and search terms related to a particular research topic.

Text mining is neatly explained in this 1m 50s video by Elsevier:

The below text mining tools can inform, improve and refine your search terms and strategy by mining medical databases for keywords and MeSH subject headings. Note: While most of these tools rely on PubMed data, be aware that none of the below tools are as up to date as PubMed itself because PubMed doesn’t allow a third party tool access to all of their records in real time.

Yale MeSH Analyzer
Upload PMIDs to create a grid of MeSH terms and author keywords from individual papers. This can help identify gaps in your search strategy. Useful when using a 'gold set' of articles in a systematic search. How to video
PubMed Reminer
A free tool that queries the PubMed database for further search terms based on your results. It will provide you with analysis in a tabular format showing the frequency of terms and MeSH headings. It will also generate rankings that show the most active journals and authors for the terms you have entered.
Ovid Citation Analyzer
Analyse citations exported from an Ovid database to build a search strategy based on those results (eg. gold standard articles or a gold set). This tool allows you to nominate specific fields for analysis (eg abstract). This analyzer will then suggest a search string which you can use to run a new search.
PubVenn
Free tool that generates Venn diagrams based on text search terms. It provides a visual representation of a citation list and example search strategies based on your terms. It also links to articles within the PubMed databases.
MeSH on Demand
Identifies MeSH terms in your submitted text (abstract or manuscript). MeSH on Demand also lists PubMed similar articles relevant to your submitted text.
MeSHy
Ranks MeSH term pairs. Search on a term for a categorised listing of positive and negative MeSH terms, which can then be considered for inclusion in your search strategy.
Carrot2
Carrot2 organizes your search results into topics. With an instant visual overview of what's available, you will quickly find search terms and articles related to your search.

When used as an analysis tool, text mining cuts down the time researchers need to spend reading textual information. This means key information can be identified quickly, an increasing benefit considering the continuous growth in the volume of published information.

The following tools for analysing and exploring language in texts are freely available.

JSTOR Text Analyzer
This tool allows you to drag and drop a key full text article (from a gold set for example) to produce additional terms for inclusion in a search strategy. Text Analyser checks the article and identifies its key topics. You can then adjust the priority of the terms returned for additional analysis or enter your own suggestions.
Voyant Tools
Voyant Tools is a web-based text analysis tool. Search results from databases can be uploaded to the tool to show word frequency, words in phrases and word trending.
VosViewer
A visualisation tool. It can be used to explore biblometric networks (journals, research institutions, authorship) and concept development through the co-occurrence of important terms in selected literature. Note: Requires download
AusNC - Australian National Corpus
A linguistic resource supported by the Australian National Data Service which provides examples of text analysis and language samples.
Vard2
A tool for dealing with and identifying historical changes in spelling and language over time.
WriteWords
A simple phrase frequency counter that allows you to identify the number of times a phrase appears in a particular text.

Using EndNote for subject heading frequency counts

Source: CADTH Text Mining Opportunities: White Paper, Appendix A, pp.34-35

EndNote allows you to calculate subject heading frequencies across a collection of records at an aggregated level. PubReminer does this too, but only for MEDLINE, so EndNote is a good option for other databases such as Embase.

Procedure

Export relevant records from PubMed, MEDLINE on Ovid, Embase on Ovid, or any other database that uses a controlled vocabulary (and includes that information in the data that is imported into EndNote), then examine the syntax of the subject headings in the keyword field (for MeSH terms from PubMed or MEDLINE on Ovid, note the asterisk for major subject headings and the /; note also that there are no semi-colons)

Before exporting records to EndNote, you can set up some preprocessing for the subject headings that will be imported (into the keyword field, in this case) for better results, depending on how you want the headings to be tabulated:
- Begin with an empty EndNote library
- In EndNote 20, go to Library > define term lists > keywords
- Select the correct delimiters — for example, if you want to treat the subheadings separately from the subject headings in PubMed or MEDLINE on Ovid, tick the box next to the “/” symbol to make sure that subheadings will be split apart from the subject headings and counted separately
- Click on “update list”
- Click “OK.”
Import the records for analysis into the new EndNote library
If you want to ignore the asterisk (i.e., treat major subject headings and minor subject headings in the same way), you can perform a Find and Replace to eliminate the asterisks from the keyword field:
- Go to Edit > Find and Replace
- Select the Keyword field (if that is the field into which the subject headings were in fact imported)
- Find the * (uncheck Match Words) and Replace with blank
- Click Change

To perform the frequency analysis:

Go to Tools > Subject Bibliography
Select Keywords from the list and click OK
Choose Select All and click OK
Click on Layout > Terms > Subject Terms Only
Change the number of lines between entries by removing ^p^p next to Suffix (this will reduce the length of the saved or printed document)
Change the display order to frequency by selecting By Term Count - Descending and click OK
Print or Save

Monash Health Library has a wide range of online journal subscriptions for staff and students to access for research and study purposes. Click here to browse through our large online journal collection. Here is a small sample of relevant Text Mining journals:

International Journal of Information Management Data Insights
Open Access journal for information management using emerging data science methods
Systematic Reviews
International Journal of Medical Informatics

Our journal subscriptions are fully integrated with the Read App to bring you easy, mobile access to medical journals and the latest articles. The app uses your preferences to provide a personalised digital newsfeed based on the specialities, favourite journals and keyword searches you set up. By connecting to Monash Health subscriptions under Settings > Institutional settings, you will be automatically connected to articles in full text.