> I did gloss over the tags but I seem to recall some in the past that were not very identifiable. At any rate, I‰??m now informed.
>
I'm routinely surprised at how hard it can be to identify the location of an event in a random news story that Google highlights. Rob's logic is able to identify the location to the state level correctly about 80% of the time, but in the 20% where no match is found, it can take several minutes of searching on the cited web site to identify the location of the event. Another oddity in this regard is that often an event in say, California, will be reported by a midwest newspaper and that is the only report on that event that google displays. I suspect that there are biases built into the google search engine that are rather subtle. I think that this illustrates the importance of including information literacy education as part of safety education for lab workers.

This experience over the last 8 years has been quite enlightening about the opportunities and limits of machine learning in reports written for other purposes. Rob's search logic has not changed very much since the first few months of use and it has been consistently aboout 80% accurate in terms of location, type of event, chemicals involved, etc.. But the process still requires human intervention to address the 20% where logic comes up empty or goes wrong. I think I am able to catch most of the erroneous results returned by the routine, but I'm sure some erroneous tags still slip through. I guess that's why safety people have jobs ;).

Let me know if you have any questions about this.

- Ralph

Ralph Stuart, CIH, CCHO
Environmental Safety Manager
Keene State College
603 358-2859

ralph.stuart**At_Symbol_Here**keene.edu

---
For more information about the DCHAS-L e-mail list, contact the Divisional membership chair at membership**At_Symbol_Here**dchas.org
Follow us on Twitter **At_Symbol_Here**acsdchas

Previous post   |  Top of Page   |   Next post