Quotes over Time tracks the top-quoted people from Reuters Alertnet News on a range of topics, and presents their quotes on a timeline. This demonstration automatically collected pilot data over the last few months of 2006. Please click on the graphs and quotes to see full stories (requires that popups be turned on in your browser; the site has been optimized for the Firefox browser).

Quotes over Time was begun as an internal resource for employees at the Parsons Institute for Information Mapping (PIIM) as a purely academic endeavor (all sources are cited). As it has garnered interest, we felt that it could serve as a generally useful resource, particularly for news organizations, policy analysts, and media analysts.

For more information, please contact arno[at]binarybottle.com.



Behind the scenes:

News stories are collected from Reuters Alertnet with custom code written in Python

The text is parsed to extract headlines, timestamps, sources, quotes, "coined expressions," names, and such (with more custom Python code) and after further processing these are stored in a MySQL database.

Pronominal coreferencing (e.g., attributing "she said" to the appropriate individual) is performed with opennlp. a Java suite of programs for natural language processing.

The front-end application is written in PHP, style sheets, and some javascript (popups and fades).
The larger graphs are created in JPGraph, based on PHP's GD library, and the smaller graphs are created with a PHP implementation of Tufte's Sparklines .

Programming: @rno klein (arno[at] binarybottle.com ).
Thanks to Jamie Smith for partnering in the conceptual development,
creating an alternative front-end in the Processing language ,
and for SQL programming.