Comments on: Free Pacer? - The implications (guest post) http://blog.justiceserved.com/?p=487 A blog from Justice Served. Thu, 09 Sep 2010 20:05:09 +0000 http://wordpress.org/?v=2.6 By: G. Thomas Sandbach http://blog.justiceserved.com/?p=487#comment-3629 G. Thomas Sandbach Mon, 19 Oct 2009 16:01:13 +0000 http://blog.justiceserved.com/?p=487#comment-3629 @Rasmussen - Thanks for your comments. I apologize for understating the ability of automated redaction tools to identify and redact specific types of data. I am not an expert in redaction, but common sense suggests some potential factors bear on redaction success rates: - Data to be redacted. Generalized data types are easier to recognize than specific data types. Recognizing proper names is different than finding all proper names. Not all dates are birth dates. As a result, in order to redact home addresses, but not those of businesses, the tool must be able to understand these data items in context. - Document types. Context is easier to recognize in some documents than in others. In forms, for instance, a field may be labeled "DOB", enabling software to identify a date in proximity as a birth date. I'm sure that the identification task becomes more difficult as contents become more complex. Registration forms present less of a challenge than depositions, for instance. - Accuracy. The white papers on your website suggest that redaction of a given document can provide different (and predictable) levels of confidence in the results of data to be redacted. Those confidence levels would apparently reflect a combination of the success of the automated tool and the amount of verification necessary to produce a finished product. My initial statement obviously failed to reflect that level of complexity. @Rasmussen - Thanks for your comments. I apologize for understating the ability of automated redaction tools to identify and redact specific types of data. I am not an expert in redaction, but common sense suggests some potential factors bear on redaction success rates:

- Data to be redacted. Generalized data types are easier to recognize than specific data types. Recognizing proper names is different than finding all proper names. Not all dates are birth dates. As a result, in order to redact home addresses, but not those of businesses, the tool must be able to understand these data items in context.

- Document types. Context is easier to recognize in some documents than in others. In forms, for instance, a field may be labeled “DOB”, enabling software to identify a date in proximity as a birth date. I’m sure that the identification task becomes more difficult as contents become more complex. Registration forms present less of a challenge than depositions, for instance.

- Accuracy. The white papers on your website suggest that redaction of a given document can provide different (and predictable) levels of confidence in the results of data to be redacted. Those confidence levels would apparently reflect a combination of the success of the automated tool and the amount of verification necessary to produce a finished product.

My initial statement obviously failed to reflect that level of complexity.

]]> By: david rasmussen http://blog.justiceserved.com/?p=487#comment-3547 david rasmussen Tue, 06 Oct 2009 02:53:44 +0000 http://blog.justiceserved.com/?p=487#comment-3547 Extract Systems has been providing redaction software to courts and other public and private entities for a long time. The author incorrectly states that tools "don’t currently exist to identify and redact children’s names, account numbers, birth dates and home addresses". These and more, are exactly the fields we DO redact (automatically). For example, we handled 22 distinct field types for the New York Secretary of State. Full disclosure... I am employed by Extract Systems. For more information see www.extractsystems.com. Extract Systems has been providing redaction software to courts and other public and private entities for a long time. The author incorrectly states that tools “don’t currently exist to identify and redact children’s names, account numbers, birth dates and home addresses”. These and more, are exactly the fields we DO redact (automatically). For example, we handled 22 distinct field types for the New York Secretary of State. Full disclosure… I am employed by Extract Systems. For more information see http://www.extractsystems.com.

]]>