New Informative PTAB Decision on Speech Recognition Neural Network
The PTAB recently designated as informative, an ex parte appeal decision for one of Baidu’s patent applications related to transcription of speech into text. Claim 1 is set forth below (SN 14/735,002):
The examiner rejected the claims as too abstract. Specifically, the Examiner asserted the claims were a mental process since a human could listen to the audio file and transcribe the data into text. The PTAB cited the latest guidelines, as it typically does, and analyzed the claimed elements. As has been the PTAB’s pattern, the specification was specifically reviewed to identify the technical problem solved by the invention.
The Applicant specifically argued that the examiner was ignoring the aspects of the claim related to the trained neural network, since that was not a generic computer:
The Applicant further argued that the claims included specific implementation steps and deal with improvements to a technology field, improve the computer itself, etc.
The PTAB first confirmed that the claims did not organize human activity, nor did recite a mental process. One distinction the decision makes is that while the recited element of using predicted probabilities to decide a transcription of the audio input is perform based on a algorithm, because the claims do not recite that algorithm the claims therefore do not fall under the recent guidelines as to reciting a mathematical formula. Finally, even if the claims did recite a formula, the PTAB finds that they integrate the concept into a practical application because they claimed elements are described as achieving an improved technological result and provide improvements to a technical field.
So, if you are preparing arguments against a Section 101 rejection, the decision lays out a good framework. You just have to hope your specification provides an explanation of how the claimed elements achieve a technological result and improve a technological field.