Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Election/Restrictions
Applicant’s election without traverse of Species I corresponding to claims 1 - 11 in the reply filed on 6/02/2026 is acknowledged.
DETAILED ACTION
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 1 and 8 is/are rejected under 35 U.S.C. 103 as being unpatentable over REZAGHOLIZADEH et al (US 2020/0097554) in view of Sallee (US 2021/0295105).
As to claim 1, REZAGHOLIZADEH et al figure 7 shows/teaches an operation method of a first communication node (paragraph [0006]…a computing system for concurrently generating parallel sentences in at least two languages, the system comprising a neural machine translator for learning a shared latent space of the at least two languages, the neural machine translator), comprising:
determining a latent space correction operation including an operation for correcting latent data output from a first encoder of a first artificial neural network corresponding to the first communication node, based on information of a reference data set provided from a second communication node (paragraph [0055]…In at least one embodiment where neural text generator 110 comprises a GAN, neural text generator 110 may be considered to have learned the manifold of shared latent space 120 when the generator (see e.g., 410 of FIG. 4) of neural text generator 110 can satisfactorily generate “fake” coded representations that fool the discriminator (see e.g., 420 of FIG. 4) into identifying them as coded representations originating from shared latent space 120 (i.e., a “real” coded representation) ; paragraph [0068]…The autoencoder may be considered to be “shared” in the sense that the same autoencoder is used for encoding (and decoding) text in multiple languages, the first and second languages L1, L2 in this example. The shared autoencoder of the neural machine translator 310 comprises an encoder 430 that learns lower-level, coded representations from training data input to the encoder 430 (e.g., sentences in languages L1 and/or L2), and a decoder 440 that attempts to reconstruct the original training data from the lower-level coded representations. Each of encoder 430 and decoder 440 is implemented using a neural network (also referred to herein as an “an encoder neural network” and a “decoder neural network” respectively). For example, encoder 430 and/or decoder 440 may be implemented as a recurrent neural network)(Examiner’s Note: “where neural text generator 110 comprises a GAN, neural text generator 110 may be considered to have learned the manifold of shared latent space 120” reads on ‘’determining a latent space correction operation including a transformation operation for correcting latent data output” ; “The shared autoencoder of the neural machine translator 310 comprises an encoder 430 that learns lower-level, coded representations from training data input to the encoder 430 (e.g., sentences in languages L1 and/or L2));
encoding first input data including first feedback information through the first encoder (paragraph [0076]… Another type of loss function may be used to compute an “adversarial loss”. Starting with two parallel sentences in the first language L1 and the second language L2, each sentence is encoded by encoder 430 into a vector representing a mapping into the shared latent space 120. Since the sentences have the same meaning, their vector representations in shared latent space 120 should be “close” to each other according to some distance measure; ideally, they define the same point on a manifold representing the shared latent space 120. A discriminator (not explicitly shown in FIG. 4) samples one of the vectors encoded from the parallel sentences and guesses from which of the first language L1 and the second language L2 the vector was encoded. By iteratively penalizing the discriminator and the encoder 430 based on feedback from the discriminator, the distance between the vectors encoded from parallel sentences in the first and second languages L1, L2 should converge to one another during the training process. The adversarial loss, which is to be minimized through training, refers to the distance between the two vectors encoded from the parallel sentences)(Examiner’s Note: “each sentence is encoded by encoder 430 into a vector representing a mapping into the shared latent space 120…” the encoder 430 based on feedback from the discriminator” reads on “encoding first input data including first feedback information through the first encoder”);
correcting first latent data output from the first encoder based on the determined latent space correction operation (paragraph [0054]… when the discriminator “correctly” identifies a fake coded representation as being fake (i.e., as being generated by the generator module), then the generator is penalized, requiring it to improve on its generation of fake coded representations)(Examiner’s Note: “when the discriminator “correctly” identifies a fake coded representation as being fake (i.e., as being generated by the generator module), then the generator is penalized, requiring it to improve on its generation of fake coded representations” reads on “correcting first latent data output from the first encoder based on the determined latent space correction operation”); and
transmitting a first feedback signal including the corrected first latent data to the second communication node (paragraph [0092]… The “fake” coded-representation ĉ is then transmitted to the discriminator 420. The discriminator 420 receives the coded representation, and determines, based at least in part on the coded representation sampled at 612, if the fake coded-representation c was sampled from the shared latent space)(Examiner’s Note: “The “fake” coded-representation ĉ is then transmitted to the discriminator 420. The discriminator 420 receives the coded representation” reads on “transmitting a first feedback signal including the corrected first latent data to the second communication node”),
wherein the corrected first latent data is decoded into first output data corresponding to the first input data in a second decoder of a second artificial neural network corresponding to the second communication node (paragraph [0076]…during training, encoder 430a receives input sentences x of the first language L1, the input sentences x being drawn from a distribution P.sub.x, and encodes the input sentences x to intermediate codes c.sub.x, which represents the contribution of codes corresponding to language L1 to shared latent space 120; similarly, with regard to the functionality of shared autoencoder 310 with respect to language L2 (310b), encoder 430b receives input sentences y of the second language L2, the input sentences y being drawn from a distribution P.sub.y, and encodes the input sentences y to intermediate codes c.sub.y, which represents the contribution of codes corresponding to language L2 to shared latent space 120. Moreover, decoder 440a is operable to decode codes c.sub.x encoded by encoder 430a, to generate sentence reconstructions in the first language L1 (x-tilde), whereas decoder 440b is operable to decode codes c.sub.y encoded by encoder 430b, to generate sentence reconstructions in the second language L2 (y-tilde))(Examiner’s Note: “encoder 430b receives input sentences y of the second language L2, the input sentences y being drawn from a distribution P.sub.y, and encodes the input sentences y to intermediate codes c.sub.y, which represents the contribution of codes corresponding to language L2 to shared latent space 120” reads on “wherein the corrected first latent data is decoded into first output data corresponding to the first input data in a second decoder of a second artificial neural network corresponding to the second communication node”).
REZAGHOLIZADEH et al fails to explicitly show/teach that the operation is a transformation operation.
However, Sallee teaches the operation is a transformation operation (paragraph [0016]…Some embodiments regard training an NN architecture to disentangle static features from dynamic features. Such disentangling provides improved encoding and decoding. Embodiments can retain both the static and dynamic features. Embodiments can be used for image matching or object recognition that is invariant to transformations to the object. The transformations can include rotation, scale, perspective (viewpoint), color, or other transformation that can be learned by an NN.).
Therefore, it would have been obvious for having ordinary skill in the art, at the time the invention was made, for REZAGHOLIZADEH et al’s operation to be transformation operation, as in Sallee, for the purpose of minimizing the reconstruction error of the reconstructions.
As to claim 8, Sallee teaches the transformation operation included in the latent space correction operation is determined to include at least one of a transition transformation operation, a rotation transformation operation, or a scaling transformation operation for the latent data output from the first encoder within a first latent space corresponding to an output end of the first encoder (paragraph [0016]…Some embodiments regard training an NN architecture to disentangle static features from dynamic features. Such disentangling provides improved encoding and decoding. Embodiments can retain both the static and dynamic features. Embodiments can be used for image matching or object recognition that is invariant to transformations to the object. The transformations can include rotation, scale, perspective (viewpoint), color, or other transformation that can be learned by an NN.).
It would have been obvious for the transformation operation included in the latent space correction operation is determined to include at least one of a transition transformation operation, a rotation transformation operation, or a scaling transformation operation for the latent data output from the first encoder within a first latent space corresponding to an output end of the first encoder, for the same reasons as above.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BRANDON S COLE whose telephone number is (571)270-5075. The examiner can normally be reached Mon - Fri 7:30pm - 5pm EST (Alternate Friday's Off).
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Omar Fernandez can be reached at 571-272-2589. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/BRANDON S COLE/ Primary Examiner, Art Unit 2128
Prosecution Insights