• 

  Sample political word analysis output

• 

  Text-paste UI

Iplo Analysis

Analyzing differences in word use between people groups or prominent figures is a common task in modern data-driven journalism.

Many sources, however, use naive methods that produce spurious results.

Iplo, the Informative Prior Log Odds Ratio method, is attested in research as being highly useful for determining which words are representative of the differences between the language use of two groups.

This precise, complex method produces high-quality distinctions through probabilistic methods and large text corpora.

We believe we are the first to bring Iplo from research to the click of a button.

Uses

By any of three methods:

1. Copying and pasting text into two boxes, (say, the Republican text and the Democratic text), or
2. inputting the twitter handles of 2 people to compare, or
3. uploading 2 .txt files to compare,

one can harness the illuminating power of Iplo. We automatically determine salient 2-word phrases to be included in the analysis, and report:

1. A graph of the most heavily biased words for each population

Built With

• d3.js
• flask
• javascript
• natural-language-processing
• nltk
• python