Looking at Words in Context

Natalie Gyenes and Rahul Bhargava.
April 18, 2017

We just released a new feature for the Topic Mapper tool that supports looking at how a word is used.  This kind of investigation into use of a word in context can be very helpful to start to gain more insights than you can from a simple check of word frequencies.  To support this, we've integrated the excellent Word Tree visualization from Google (as designed by Fernanda Viegas and Martin Wattenberg). Here are a few examples to illustrate why this can be useful.  Hopefully you find it as revealing as we have already!

"Child" in Reporting on Contraceptives in Nigeria

Our Contraceptives in Nigeria topic looks at reporting in Nigerian media outlets in 2015 and 2016.  We saw "children" was the most used word, leading us to believe that that contraceptives were being talked about as a way of preventing unintended pregnancies.  However, when you look at the word tree a different story emerges.

This word tree shows you the words that most often came before (on the left) and after (on the right) the word "child". The size of the word corresponds to how often it was used (in a sample of sentences). Here you can see that one of the heaviest weighted related terms that shows up after the word "child" is "marriage'".

With a closer look at that connection, we see that those articles largely refer to having a child outside of marriage, as well as the concept of child marriage itself, and FGM/C (female genital mutilation/cutting) - though legislation has been implemented to address the latter.

Without really diving into this content area with the word tree, it may have been difficult to understand this word frequency as the child marriage and FGM/C narratives for media coverage of contraceptives in Nigeria.

"Immigrant" in Reporting on Head Start in the US

A separate investigation led us to create a Topic looking at media coverage of Head Start and nutrition in the US.  It didn't surprise us to see that the third-most used word was "immigrants", because we suspected that a high percentage of participating families are immigrants (at first glance they don't appear to collect data about this).  However, we were surprised to see that this is yet another topic where the terms "legal" and "illegal" come to frame the narrative around immigrants in the US.

This word tree clearly shows that the majority of uses of "immigrants" in this coverage was preceded by either "legal" or "illegal".

Digging in deeper shows the phrase "legal immigrants from illegal immigrants" is a very common phrase.  This clearly sets a narrative for reporting about participants in Head Start.