Lawyerbot Will Do The Donkey Work, Facilitates Database Search
A recent judicial ruling in US has cleared the roads for "predictive coding", a technique implemented in order to sieve through a large database of information to produce relevant results. This alleviates database search, a feature of particular importance in the field of law, where a frequent task of "legal discovery" requires the lawyers to study documents related to the case they are working on.
#-Link-Snipped-#
Thomas Gricks, who is currently working on a certain defense case for Landow Aviation, was required to examine about 2 million emails and attachments. This task demands 20,000 person hours and hence cost his clients $2 million. Predictive Coding, on the other hand, asks for a sample set of just a few thousand documents, marked as either relevant or non-relevant, to be used as training data. As a result, the system learns to classify documents into "scam" or "genuine" based on the keywords it learnt from training data.
Predictive Coding has been proven to match, and in some cases, even beat human accuracy at finding relevant documents. The biggest advantage of this technique is that it saves time and hence the clients' money as well.
Source: #-Link-Snipped-#Â Image Credit: #-Link-Snipped-#
#-Link-Snipped-#
Thomas Gricks, who is currently working on a certain defense case for Landow Aviation, was required to examine about 2 million emails and attachments. This task demands 20,000 person hours and hence cost his clients $2 million. Predictive Coding, on the other hand, asks for a sample set of just a few thousand documents, marked as either relevant or non-relevant, to be used as training data. As a result, the system learns to classify documents into "scam" or "genuine" based on the keywords it learnt from training data.
Predictive Coding has been proven to match, and in some cases, even beat human accuracy at finding relevant documents. The biggest advantage of this technique is that it saves time and hence the clients' money as well.
Source: #-Link-Snipped-#Â Image Credit: #-Link-Snipped-#
0