AI Concepts for Business Applications by Nelson E. Brestoff

Author:Nelson E. Brestoff [Nelson E. Brestoff]
Language: eng
Format: epub
Publisher: Business Expert Press
Published: 2019-08-07T16:00:00+00:00

Figure 6.3 This ROC-AUC graph above is FIG. 5 in Patent No. 9,552,548

“Indico also used about 75 held-out documents in order to provide a graph of a curve showing a ROC, illustrated in FIG. 5, which included a related indicator called the AUC. These statistics put ‘true positives’ on the y-axis and ‘false positives’ on the x-axis. On both axes, the maximum score is 1.0. A strong algorithm will score high and to the left, which is what we saw.

“Also since the maximum score for each axis is 1.0, the area under the ROC curve is also 1.0. According to the second provider, Indico, the AUC score was 0.967 (see the AUC score in the lower right hand corner), which means that the algorithm, as trained, is strong.

“In these experiments, the early warning system, having been trained (as an example) to detect an employment discrimination risk, found one (1) e-mail out of 7,665 e-mails which signaled just such a risk. Thus, the inventive concept of the system has been tested using a specific litigation category, the same training data, two different algorithms, and Enron test data, and has functioned in accordance with the inventive concept. The system, as trained, had found a needle in a haystack.

“Furthermore, the early warning system would function for an authorized enterprise employee in an unobtrusive way. Enterprise personnel need not be involved in data-mining to train the Deep Learning system, and the algorithm itself would scan internal e-mails and would run in the background. The system would not require anyone’s attention until the system had a reportable result, at a threshold the legal department could set.

“At that point, the system would output an alert to a privileged list of in-house personnel, likely in-house counsel or employees under their direction and control, and they would be able to see a spreadsheet with a score in column A, the related text in column B, along with a bar chart for context, and then would be enabled to call forward the E-mails of interest.

“The experiments discussed above employed deep learning algorithms which are described in the academic literature as Recursive Neural Tensor Networks or as RNNs with either Long Short-Term Memory (LSTM) or Gated Recurrent Units (GRUs). Those of skill in the arts will appreciate that use of other algorithms, including those which are now open-sourced, are contemplated.

“While the invention herein disclosed has been described by means of specific embodiments and applications thereof, numerous modifications and variations could be made thereto by those skilled in the art without departing from the scope of the ­invention set forth in the claims.”

An Example of Claims

Note that the last paragraph indicated that the “scope of the invention” was “set forth in the claims.” I think I’d be remiss if I didn’t recite some of the claim language for the “parent” as an example. Before I do that, there were 20 claims. Some of them are stand-alone, like Claim 1, but many of them refer to a previous claim. And, while the

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.