Thursday, May 28, 2009

Conflict Checking Improvements

OK, here is my post about conflict checks. Things have been very busy here with Pika installs so Iʼm getting this posting out later than I wanted to.

Weʼre going to add Alien ID number to the conflict check. This will be pretty straightforward, itʼll work like the Social Security Number. Also, weʼre writing a new name searching algorithm to better handle individuals whose last name contains both a paternal and maternal family name.

Take the following (contrived) example. In 2007, your program successfully represented a client, Jane Williams, in her landlord-tenant case with her landlord, Brazilian oil magnate José Sergio Gabrielli de Azevedo. Now letʼs say Mr. Gabrielli calls up your program in 2009, looking for legal assistance. If provides his full name, José Sergio Gabrielli de Azevedo, to your intake screen, the software will view this as a potential conflict of interest. However, if he simply identifies himself as “José Gabrielli”, the phonetic name search will not catch him because, to the computer, “Gabrielli” is not the same as “Gabrielli de Azevedo”. If you use the “Browse Names” feature on the intake screen, he will still show up, but ideally he really needs to show up using either method.

So weʼve narrowed our options with the change to two. The first option would be to give the staff person the option to enter an alias during data entry. So with the previous example, José would be entered into the system with a last name of “Gabrielli de Azevedo”, and would have one alias with a last name of “Gabrielli”. That way heʼd show up no matter which form of his last name was used as the search term. We would find a way to streamline the process somehow, maybe a box labeled maternal surname or maiden name that can be filled out, and would trigger the creation of an alias record in the background.

The other option weʼre looking at would be to give you a single Name Search box where you can enter any number of first and last names. If you search for “José Gabrielli”, or for “José Sergio Gabrielli de Azevedo”, it will bring up any José Gabriellis and any José Sergio Gabrielli de Azevedos on one list. This method might bring up more false positives that the current method (“Arthur Scott” would come up as a possible match for “Scott Arthur”,) but Iʼm not sure how much this would be an issue in day-to-day use.

Please let me know if you have any comments or ideas on this topic.

1 comment:

Carol Garner said...

This option ["The other option weʼre looking at would be to give you a single Name Search box where you can enter any number of first and last names."] sounds like it might provide the broadest possible matching - which I would support.

We run into maddening issues with names (double-surnames may be hyphenated or merely separated by spaces; we sometimes have surnames with and apostrophe and then sometimes a space and sometimes no space between it and the next letters). So the broadest possible way to capture names is very attractive