Making sense of numbers and words: Statistical methods

Peter Grimbeek

Contact details

Leximancer Vsn 3

In a previous comments about the brave new world of Leximancer 3, I waxed nostalgic about the lost joys of version 2. But I’ve realised that the thing to do is to press on. So, since I regard Leximancer as the software of choice for text analysis I will comment instead on the various types of output available in version 3, and how to export this output into word-processor documents.

First, Leximancer 3’s transformation from stand-alone application to server-based software has brought particular advantages, more especially the aptly named Insight Dashboard. A minimalist way to set up the Dashboard is to select folder or file name tags or any concept of particular interest as categories (outcome variables), and to select all available concepts as attributes (predictor variables). The Insight Dashboard produces HTML or PDF files that can include a quadrant map and ranked concept lists (attributes) relative to specific categories (e.g., tags). This can be especially useful when you want to view content from differing perspectives (e.g., Interviewer vs. interviewee). The resulting concept lists (per category) rank concepts not only in terms of frequency and strength (of connection) but also in terms of prominence (here 3+ is more highly regarded). The automated report includes links between each concept and the relevant set of extracts. Insight Dashboard has particular application when generating automated reports based on open-ended responses to customer surveys on websites but is a generally useful addition to the Leximancer toolkit.

However, the new look Leximancer still includes an ability to compile reports via the concept map. Concept map output now includes ranked concept and thematic list graphs, and rotatable concept maps without or with the trademark resizable thematic circles. The concept map can also include sentiment filters (utilising automatically generated compound concepts for favourable, unfavourable, & negation terms), and an ability to view and export text extracts based on concepts that are highly connected to themes as well as other concepts.

The concept map can be copied and pasted into Word after a transitional step of generating images in JPG, PNG or other formats. The ranked concept and thematic summary graphs can be pasted into word-processor documents after copying all of the screen (Control+Alt+Print-Screen key, followed by use of crop tool to hide superfluous parts of image: Windows) or taking snap-shots of relevant parts of the screen (Command+Option+Shift+4: Apple Mac).

Text extracts based on the connectivity of concepts to themes are visible directly below the ranked thematic summary graph. Setting the level of thematic detail to low, medium or high, can vary the number of these. In a like manner, extracts based on the connectivity between concepts can be viewed by selecting a concept of interest, and then clicking on the relevant document icon. These extracts can be exported (via Export or Export all) to summary web pages. Extracts based on concepts can also be transferred to a logbook, which is advantageous given the logbook's ability to collate the results of a series of examinations of the connections between concepts.

Once in summary web pages, thematic and concept extracts  can be copied and pasted directly into word processor documents. The resulting material is likely to include formatting that reflects its web based origin, in particular manual line breaks, which ideally one would replace with spaces. A work-around that involves saving the web page as an HTML page, then opening that page with a text editor, and then saving the opened document as an RTF file removes the manual line breaks but replaces these with words that run together (e.g., runtogether: Where run precedes the line break and together follows it) as well as producing oddly resized tables. For this reason, it’s probably as easy to copy and paste directly from web page to word-processor, and then use find and replace options to replace manual line breaks with spaces, etc. Where extracts based on concepts have been collated in the logbook, the resulting summary web page still includes manual line breaks but is otherwise somewhat better formatted for the word-processor environment.

Finally, Leximancer 3 also offers the option of exporting a spreadsheet containing the co-occurrences between concepts as a third type of output. This option would be of definite interest to those who would like to analyse the co-occurrences between concepts in other ways.