Text visualization tools for qualitive analysis of Library survey comments


Say you are conducting a survey and like most surveys you include a final open ended question. Again assume you have a large number of such responses (over 200), how do you analyze such unstructured data?

This is a problem faced by many librarians, whether they are running standardised Libqual+ surveys or their own home-brew survey.

One way of course would be to do content analysis. At the risk of over-simplifying you read through the comments, try to code them with some system that best allows you to combine similar comments together.

For instance, if you find many comments complaining about the lack of seats in the library, you might just code all such comments as “Too crowded”

For an example of this method see this paper (subscription required) on analysing Libqual+ comments.

Another interesting approach would be to subject all the comments with  Text Visualization methods first to quickly spot patterns (possible codes). I found this amazing service – Many eyes by IBM that allows you to upload data and create visualizations of many different forms. Many eyes can do different types of visualizations but here I focus only on Text visualizations.

Tag Cloud

By now, every knows of tag clouds

Click on images to see them in their full glory.

You know the deal, larger the font it appears in, the more frequently it appears. You can choose to create tag clouds using one word, or two word phrases. Mouse over the tags and you can see the comments that contains those text in context. More details.

Gives you a quick overview of the codes that you might use for content analysis. For instance you can quickly see that “discussion rooms” are mentioned quite a bit.

Next up is Wordle


Click on images to see them in their full glory.

Word Trees

Basically another kind of tag cloud. Moving on to some even more interesting, word-trees.

Click on images to see them in their full glory.

Here, I typed in the word “good” and it will show, all other possible continuations, with the most common ones in larger font. For instance, the word tree above shows that “good” is followed most often by “Service” and “work”. Click on say service, and you get the image below

Click on images to see them in their full glory.

Instead of the word good service appearing in front, you can also set your chosen term (“good service” in image below) to appear at the end as the image below shows.

Click on images to see them in their full glory.

Phrase Nets

And lastly you can create phrase-nets out of the comments you get.

Click on images to see them in their full glory.

Words linked together in the phrase-nets can be connected by “and”, “a” , “at”, “the” etc (see selected option in orange on the left side of the image). In the example, above, they are simply connected by a space (they are next to each other).

The size of a word is proportional to the number of times it occurred in a match; the thickness of an arrow between words tells you how many times those two words occurred in the same phrase. The color of a word indicates whether it was more likely to be found in the first of second slot of a pattern. The darker the word, the more often it appeared in the first position.” — more.

Click on images to see them in their full glory.

The image above, shows the same phrase-net , except they are connected by “is”.  So the above shows that the phrase “library is excellent” and “service is excellent” appears a lot.

You can even use regular expressions to find and connect phrases!


To be frank, the images alone don’t do them justice.

Want to play with a real data set? Try using this Libqual comments dataset uploaded by Cook University Library? Here are the  Word clouds , Word tree and Phrase net visualizations of this data.