Cutting the Boolean NOT

A classic problem in online searching is finding documents which aren’t about the usual suspects.

Suppose, for instance, you’re trying to find documents about types of sustainable energy other than wind, wave or solar power. The classic novice’s reaction is to use an advanced search phrasing along the lines of sustainable energy NOT wind wave solar

The problem with this phrasing, as every information retrieval specialist knows all too well, is that it will only find documents which mention sustainable energy and which do not mention the words wind, wave or solar. However, the vast majority of relevant records will mention those words as part of their introduction. Those records would all be discarded because of the NOT, even though they were relevant.

The other obvious solution, namely specifying the “not the usual suspects” energy sources that you want, also hits a major problem, since by definition you’ll probably be doing this search precisely because you don’t know what those other energy sources are, and you’re trying to find out.

One new solution to this problem involves making use of the human ability to make sense of patterns rapidly and efficiently. Here’s an example. It’s a search for wind wave solar on Searchvisualizer. It shows records returned from a standard Bing search. The Searchvisualizer has shown where each keyword occurs throughout each record.

There’s a distinct pattern in the keyword distribution within the Science Online article at the right of the screenshot. All three terms are mentioned at the start of the article, but then there’s a layer of mentions only of solar. That’s followed by a short gap, after which there are mentions of wind, followed immediately by mentions of wave.

If we look at the whole of the article in this format, we see that the closing section partially mirrors the opening section, with mentions of wind and solar but not wave.

What does this suggest?

The obvious, and correct, conclusion is that the article is structured in well-defined sections. There’s an introductory section which gives an overview of the topics covered below; then there are separate sections on each energy type. The gaps occur where the article describes energy types which are not wind, wave or solar. When we go to the article, we find the following header:

(screenshot from:

The first gap between layers of keywords is where there’s a section on geothermal energy; the second, longer, gap occurs where there are sections on tidal energy, hydroelectric energy and energy from biomass.

In other words, the gaps are the places where the “not the usual suspects” topics are located.

When you’ve found a relevant-looking record, Searchvisualizer lets you examine the visualization so you can decide whether or not you want to proceed to the article itself; if you hover over a keyword, Searchvisualizer shows you the text surrounding that keyword. If you want to proceed to the article itself, you simply click on one of the white squares.

One advantage of the Searchvisualizer’s format is that it can show very large documents on a single screen – for instance, it’s possible to show an entire Shakespeare play in one screen.

This format has the added advantage of letting the user find relevant records even in languages that they don’t speak; it’s simply a case of translating the keywords into the target language, and then identifying relevant records using the same process as with the wind wave solar example above; relevant records can then be translated using online translation software. This can be very useful for researchers who want to know about specialist research in countries whose language they don’t speak.

There’s a substantial literature about the human ability to identify patterns, and about the difficulties of getting software to identify those same patterns. What we’ve done with Searchvisualizer is to transform information into a format that plays to the strengths of the human being, particularly the strengths that the human hardly thinks about, such as knowing how a well-written document is structured.

There are other examples of how this can be used in the “Texts and Resources” section of the Searchvisualizer site:

The Searchvisualizer software on the site is free, and the site includes various hints, tips and resources.

We hope you’ll find it interesting and useful.

 By Gordon Rugg


About searchvisualizer

We welcome debate and disagreement, but not abuse, trolling or thread derailment. We reserve the time-honoured right of blog owners and moderators to be arbitrary, capricious and autocratic in our wielding of the ban hammer. Gordon Rugg is a former timberyard worker, archaeologist and English lecturer who ended up in computer science via psychology. He’s the same Gordon Rugg who did the Voynich Manuscript work, and the books with Marian Petre about research. He’s co-inventor of the Search Visualizer.
This entry was posted in About SV, background theory and tagged . Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s