USPTO examiners will take a liberal view of written description requirements when it suits their rejection, particularly when they need to find support in a priority document because only the priority document qualifies as prior art.
Appeal 2024-001531, Application 17/626,453, to Bridgestone relates to an invention for a tire manufacturing system. Claim 1 recites:
1. A system to manufacture a tire, the system comprising:
one or more processors and a memory, the memory configured to store instructions that are executable by the one or more processors and cause the one or more processors to:
[A] receive, from one or more sensors of one or more pieces of tire manufacturing equipment in a tire manufacturing plant, one or more values corresponding to manufacture of a tire by the one or more pieces of tire manufacturing equipment;
[B] determine one or more metrics based on the one or more values;
[C] generate a matrix based on the one or more values and the one or more metrics;
[D] predict, via input of the matrix into a machine learning model constructed to output tire manufacturing performance data, [D1] a value for a splice tolerance metric for the tire;
[E] determine, based on the value for the splice tolerance metric, a parameter of at least one piece of equipment in the tire manufacturing plant to adjust; and
[F] provide a command to adjust the at least one piece of equipment in the tire manufacturing plant responsive to the value of the splice tolerance metric.
The examiner rejected the application based on a combination of references. However, one of the references only qualified as prior art as of its provisional filing date. As discussed in a previous post (HERE), examiners can sometimes take shortcuts by relying on the disclosure of the later application, while relying on the date of the earlier filed priority application. Whenever this situation arises, it can be important to confirm that the cited disclosure in the later application is also present in the priority document. If it is not, then the rejection falls. And that is exactly what transpired in this case with regard to limitation [D] of claim 1.
Specifically, the examiner relied on disclosure of a machine learning model in the later dated reference. However, the only way the reference qualified as a prior art was through its priority date, and when examining that priority document (a provisional application), the disclosure as to the machine learning model was different and lacked certain elements relied on by the examiner to formulate the rejection.
The PTAB easily dismissed of the case with a short opinion reversing the rejection, as reproduced below (internal citations omitted). However, interestingly, Bridgestone does not seem to have ever raised the priority issue in their briefs. While Bridgestone did argue that the teaching of the reference (Kucera) was deficient, and while the PTAB cited both the Appeal Brief and Reply Brief in its opinion, the cited pages do not seem to raise the issue of priority to the provisional application of Kucera. Nonetheless, the PTAB decided the case on this issue.
However, U.S. Provisional Application No. 62/755,297 fails to disclose support for language in Kucera that teaches or suggests limitation D recited in independent claim 1. … namely a machine learning model that receives a matrix as input, and then predicts a value for a manufacturing performance metric as an output (see e.g., claim 1, limitation D). The Federal Circuit noted that it is well settled that a description that merely renders the invention obvious does not satisfy the written description requirement. … Our careful review of the Drawings and Specification filed for U.S. Provisional Application No. 62/755,297 reveals that “model training” is performed using a “runtime algorithm” (see Drawings, page 5), such as “analytics algorithms” (possibly including “self-learning algorithms”) based on “training sets,” which in turn are based on “system control data” as described in the Specification at page 4.
In light of the above, U.S. Provisional Application No. 62/755,297 does not provide support for a manufacturing system or method in which a machine learning model receives a matrix as input, and then predicts a value for a manufacturing performance metric as an output, as required by claims 1 and 11 (see e.g., claim 1, limitation D). As a result, Kucera is not entitled to the benefit of the U.S. Provisional Application No. 62/755,297 filing date concerning at least disputed limitation D of claim 1 (and the similar limitation of claim 11). Therefore, we agree with Appellant …
In view of the foregoing, we are constrained by the record to reverse the Examiner’s rejections of claims 1–20, all over the same base combination of Bull and Kucera
So, look out for examiners playing fast and loose with the disclosure requirements of a priority document whose date is critical to a rejection. The standard is exacting – actual disclosure is required to support the features relied upon in the later filed application forming the basis of the rejection. And, be careful not to miss this issue as it can be determinative on appeal.