The Small World of Words: Mapping the Mental Lexicon
Graphs provide a common language to represent networks that at first glance are nothing alike, from friendships on social media to interstate traffic patterns. One ambitious project is attempting to model the network of all word associations in the collective mental lexicon of native speakers. Called the “Small World of Words,” it began at a Belgian university in 2003 and has since become a large-scale international effort. By gathering data with simple online word association tasks, it has collected over one million responses for English alone and claims to be the largest project of its kind. From this, the researchers have been able to analyze the mental lexicon on both a very broad and very detailed level, showing the power of graphs to synthesize information that would otherwise be incalculable.
While this image looks like something you might see under a microscope, it is in fact the complete network of English word associations collected thus far by the Small World of Words project. It has several features that are worth analyzing. First, it appears to be a connected network, with no small components or nodes separated from the giant component. The reason for this seems fairly straightforward: the online tasks require users to generate associated words; if they cannot think of any, they abandon the task and their data is not collected. Another factor is the research method: the researchers began with a set of common words (7,000 cue words for the English version) and branched out from there. Because the words all come from a common source, there can’t be a lone word with no connections.
Another interesting feature is the network’s set of hubs, which are the large labeled circles in the picture above. They seem to represent most of the core human needs: food, water, money, work, time, and love are the most prominent hubs. While it’s not clear if the proximity of the nodes to one another means anything in this graph, it is notable that money, work, and time are near each other, as are food and water. The money-time-work association could be due to their conceptual relatedness and to the many idioms that contain some combination of those words.
Last year, one of the project’s main researchers gave a presentation detailing some of the findings about the network’s structure. One topic the presentation examined was semantic richness of different words. The image above shows a comparison between the words “igloo” and “house,” color-coded for the different categories of words connected to the main word. Because “house” is a less specific word than “igloo,” it has more semantic associations and thus more semantic richness. Igloo, on the other hand, is only connected to two main semantic categories: houses and coldness. Many words on the “coldness” side have to pass through it to get to “house” and vice versa.
While we have only briefly discussed the mathematics of the small-world phenomenon in class, this study illustrates in a more general sense the ways network structure underlies our everyday lives. In a way, the Small World of Words project owes its success to such structures, which have allowed it to propagate through the internet and reach over a million English speakers (not to mention the countless others who have participated in other languages). Even so, its propagation would not have been possible without the inherent appeal of the concept: a graph that makes visible the contents of our minds.
Links
Interactive visualizations: http://www.smallworldofwords.com/new/visualize/
Researcher presentation: http://www.smallworldofwords.com/new/data/presentations/netsci2013.html#/
Participate in the project: http://www.smallworldofwords.com/en/