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 .
Claims 1-20 are presented for examination.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-20 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor, or for pre-AIA the applicant regards as the invention.
Claim(s) 1, 14 and 20, each recite(s) “generating, ... a final latent representation of the sequence of discrete tokens that includes a determined value for each of a plurality of latent variables”.
It is unclear whether “that” refers to “a final latent representation” or “sequence of discrete tokens”, rendering the claim(s) indefinite.
For examination purposes the examiner has interpreted “that” to be “a final latent representation”.
Claim(s) 1, 14 and 20, each recite(s) “a distribution over a continuous space of possible values for each of the plurality of latent variables”. It is unclear whether
a single distribution ...corresponds to all of the plurality of latent variables, or
a respective distribution ... corresponds to each respective variable of the plurality of latent variables, rendering the claim(s) indefinite.
For examination purposes the examiner has interpreted “a distribution over a continuous space of possible values for each of the plurality of latent variables” to be “a respective distribution over a continuous space of possible values for each respective variable of the plurality of latent variables””.
Claim(s) 1, 14 and 20, each recite(s) “the distributions”. There is lack of antecedent basis for this limitation in these claim(s), since there is possibly a single distribution recited previously in the claims.
Claim(s) 2-13 and 15-19, do not contain claim limitations that cure the indefiniteness of claim(s) 1 and 14 respectively, and therefore are also indefinite under 35 U.S.C. 112(b).
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claim 20 is rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter.
Regarding 20, this claim recites a “computer storage medium” encoded with instructions that perform various functions. There is no structural component associated with the computer storage medium and therefore the computer product can include transitory media. Therefore claim 20 is directed to non-statutory subject matter.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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.
Claims 1-11, 13-20, are rejected under 35 U.S.C. 103 as being unpatentable over Li et al
“Diffusion-LM Improves Controllable Text Generation” dated 27 May 2022 and retrieved from arXiv:2205.14217v1, in view of Yu (US 20240112088 A1).
Regarding claim 1, Li teaches a ...method for generating an output sequence of discrete tokens using a diffusion model, the method comprising (Li Abstract algorithm to generate output token sequence using a diffusion model):
generating, by using the diffusion model, a final latent representation of the sequence of discrete tokens that includes a determined value for each of a plurality of latent variables, wherein generating the final latent representation comprises, at each of multiple reverse diffusion time steps(Li Abstract, Sec 3.3, 4.2, multiple iterations are performed to reverse sequence of discrete tokens to a final latent representation of the sequence of discrete tokens using multiple iterations):
processing a diffusion model input comprising an intermediate latent representation of the sequence of discrete tokens for the reverse diffusion time step to generate an estimate of the sequence of discrete tokens as of the reverse diffusion time step (Li Secs 3.3 and 4.1, each iteration (time-step) may use previous iteration’s intermediate variables to generate approximate (estimate) discrete token sequence);
using the estimate to define a distribution over a continuous space of possible values for each of the plurality of latent variables; and generating an updated intermediate latent representation of the sequence of discrete tokens for the reverse diffusion time step through sampling from the distributions (Li Intro and Secs 3.3, 4.1, 4.2, 5.1, latent variable continuous distribution is based on approximate sequence, distribution is sample to generate latent representation for current iteration) ;
applying a de-embedding ...operation... having learned values to the final latent representation of the output sequence of discrete tokens to generate a de-embedded final latent representation that includes, for each of the plurality of latent variables, a respective numeric score for each discrete token in a vocabulary of multiple discrete tokens; selecting, for each of the plurality of latent variables, a discrete token from among the multiple discrete tokens in the vocabulary that has a highest numeric score (Li Secs 4.1, 4.2 and 5.1, decoding reverse process (de-embedding) with learned parameters is used on final latent representation, to generate output token sequence based on probability of each token for each latent variable); and
generating the output sequence of discrete tokens that includes the selected discrete tokens (Li 4.2 and 5.2, tokens are formed into an output sequence of discrete tokens).
Li does not specifically teach computer-implemented method, a de-embedding matrix
However Yu teaches computer-implemented method (Yu [127]), and
applying a de-embedding matrix having learned values to the final latent representation of the output sequence of discrete tokens to generate a de-embedded final latent representation (Yu [12, 28, 52, 61, 89, 90, 108] transformer based decoder may be applied to generate final latent representation, transformers may use matrices with learned values to perform operations, models may be diffusion models, yields better efficiency and reconstruction fidelity).
It would have been obvious to one of an ordinary skill in the art before the effective filing date of the claimed invention, to have incorporated the concept taught by Yu of computer-implemented method, and applying a de-embedding matrix having learned values to the final latent representation of the output sequence of discrete tokens to generate a de-embedded final latent representation, into the invention suggested by Li; since both inventions are directed towards using continuous distributions of intermediate latent values to process and reconstruct discrete token sequences, and incorporating the teaching of Yu into the invention suggested by Li would provide the added advantage of yielding better efficiency and reconstruction fidelity,
and the combination would perform with a reasonable expectation of success (Yu [127, 12, 28, 52, 61, 89, 90, 108]).
Regarding claim 2, Li and Yu teach the invention as claimed in claim 1 above.
Li further teaches wherein the discrete tokens comprise text (Li Sec 4.1, tokens may be words).
Regarding claim 3, Li and Yu teach the invention as claimed in claim 1 above.
Li further teaches wherein the diffusion model input further comprises an estimate of the sequence of discrete tokens generated as of a previous reverse diffusion time step Li Secs 3.3 and 4.1, each iteration (time-step) may use previous iteration’s intermediate variables to generate approximate (estimate) discrete token sequence).
Regarding claim 4, Li and Yu teach the invention as claimed in claim 1 above.
Li does not specifically teach wherein generating the output sequence of discrete tokens using the diffusion model comprises generating unconditional discrete tokens
However Yu teaches wherein generating the output sequence of discrete tokens using the diffusion model comprises generating unconditional discrete tokens (Yu [52, 97] output may be unconditional discrete tokens).
Regarding claim 5, Li and Yu teach the invention as claimed in claim 1 above.
Li further teaches wherein generating the output sequence of discrete tokens using the diffusion model comprises generating discrete tokens conditioned on an input sequence of discrete tokens, and wherein the method comprises: receiving the input sequence of discrete tokens; converting each discrete token in the input sequence into a one-hot vector; and applying an embedding ...function... having pre-trained values to each one-hot vector to embed the one-hot vector into a continuous vector (Li Secs 3.2, 3.3, 4.1, 4.2, 6.4 and F, output may be discrete tokens conditioned on input sequence, input sequence may be discrete tokens, input may be converted to one-hot vectors, embedding function may be applied to input representation to generate embeddings, embeddings can be continuous vector).
Li does not specifically teach an embedding matrix.
However Yu teaches applying an embedding matrix having pre-trained values (Yu [87, 88] encoder may produce embeddings, Yu [61, 89, 90] encoder may employ embedding matrix with learned parameters).
Regarding claim 6, Li and Yu teach the invention as claimed in claim 5 above.
Li further teaches wherein the output sequence of discrete tokens also includes the input sequence of discrete tokens received by the diffusion model (Li Sec 3.3 output token sequence may be reconstruction of input sequence to the model).
Regarding claim 7, Li and Yu teach the invention as claimed in claim 5 above.
Li does not specifically teach applying a linear projection to the continuous vector to generate a projected continuous vector; and processing the projected continuous vector using the diffusion model
However Yu teaches applying a linear projection to the continuous vector to generate a projected continuous vector; and processing the projected continuous vector using the diffusion model (Yu [28, 82, 91, 97] input may be continuous vector, input may be subject to linear projection for encoding by model).
Regarding claim 8, Li and Yu teach the invention as claimed in claim 5 above.
Li does not specifically teach wherein generating the discrete tokens conditioned on the input sequence of discrete tokens comprises using a classifier-free guidance technique
However Yu teaches wherein generating the discrete tokens conditioned on the input sequence of discrete tokens comprises using a classifier-free guidance technique (Yu [113-115] conditional token generation may use classifier-free guidance technique).
Regarding claim 9, Li and Yu teach the invention as claimed in claim 1 above.
Li does not specifically teach wherein the de-embedding matrix has been learned during training of the diffusion model while the embedding matrix is fixed during the training of the diffusion model
However Yu teaches wherein the de-embedding matrix has been learned during training of the diffusion model while the embedding matrix is fixed during the training of the diffusion model (Yu [32, 36, 88] decoder parameters may be determined after freezing encoder parameters).
Regarding claim 10 Li and Yu teach the invention as claimed in claim 1 above.
Li further teaches training the diffusion model on unlabeled discrete token data comprising discrete token inputs to minimize a mean-squared error between each discrete token input and an estimate of the discrete token input generated by the diffusion model as of a sampled reverse diffusion time step (Li Secs 3.1 and 3.3 token data may be unlabeled, training can be to minimize mean-squared error between each discrete token input and an approximate discrete token input for current iteration).
Regarding claim 11, Li and Yu teach the invention as claimed in claim 10 above.
Li further teaches wherein the training also minimizes a cross-entropy loss evaluated with respect to the final latent representation of the sequence of discrete tokens generated by the diffusion model from the discrete token input (Li Sec 4.1, training can minimize cross entropy loss for final latent variables).
Regarding claim 13, Li and Yu teach the invention as claimed in claim 10 above.
Li does not specifically teach wherein the training comprises learning values of the de- embedding matrix while keeping the pre-trained values of the embedding matrix fixed
However Yu teaches wherein the training comprises learning values of the de- embedding matrix while keeping the pre-trained values of the embedding matrix fixed (Yu [32, 36, 88] decoder parameters may be determined after freezing encoder parameters).
Claim 14 is directed towards a system executing instructions similar in scope to the instructions performed by the method of claim 1, and is rejected under the same rationale.
Yu further teaches a system comprising one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations (Yu [127]).
Claim(s) 15-18 and 19 is/are dependent on claim 14 above, is/are directed towards a system executing instructions similar in scope to the instructions performed by the method of claim(s) 2-5 and 9 respectively, and is/are rejected under the same rationale.
Claim 20 is directed towards a system executing instructions similar in scope to the instructions performed by the method of claim 1, and is rejected under the same rationale.
Yu further teaches a computer storage medium encoded with instructions that, when executed by one or more computers, cause the one or more computers to perform operations (Yu [127]).
Claim(s) 12 is/are rejected under 35 U.S.C. 103 as being unpatentable over Li et al in view of Yu (US 20240112088 A1), and further in view of Mittal et al “SYMBOLIC MUSIC GENERATION WITH DIFFUSION MODELS” dated 25 Nov 2021, and retrieved from arXiv:2103.16091v2.
Regarding claim 12, Li and Yu teach the invention as claimed in claim 10 above.
Li further teaches training the diffusion model on the masked sequence of discrete token inputs to generate an estimate of the sequence of discrete token inputs that includes the infilling tokens in the sequence (Li Sec 3.1, 4.1, 4.2 and 6.4, diffusion model may be trained based on masked token sequence to generate approximate token sequence and may use infilled tokens).
Li does not specifically teach applying binary masks to a sequence of discrete token inputs, the binary masks comprising one or more first masks defining conditioning tokens in the sequence and one or more second masks defining infilling tokens in the sequence
However Mittal teaches applying binary masks to a sequence of discrete token inputs, the binary masks comprising one or more first masks defining conditioning tokens in the sequence and one or more second masks defining infilling tokens in the sequence (Mittal Sec 3.3 conditional infilling may use masks that define conditioning and tokens to use for infilling, trajectory of the reverse process can be steered and arbitrarily conditioned without the need for retraining the diffusion model).
It would have been obvious to one of an ordinary skill in the art before the effective filing date of the claimed invention, to have incorporated the concept taught by Mittal of applying binary masks to a sequence of discrete token inputs, the binary masks comprising one or more first masks defining conditioning tokens in the sequence and one or more second masks defining infilling tokens in the sequence, into the invention suggested by Li and Yu; since both inventions are directed towards using diffusion models to generate estimate of discrete token sequence(s), and incorporating the teaching of Mittal into the invention suggested by Li and Yu would provide the added advantage that trajectory of the reverse process can be steered and arbitrarily conditioned without the need for retraining the diffusion model, and the combination would perform with a reasonable expectation of success (Mittal Sec 3.3).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Torrado (US 20220414429 A1) discloses a diffusion model to generate an output sequence based on encoder embeddings.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SANCHITA ROY whose telephone number is (571)272-5310. The examiner can normally be reached Monday-Friday 12-8.
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, Usmaan Saeed can be reached at (571) 272-4046. 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.
SANCHITA ROY
Primary Examiner
Art Unit 2146
/SANCHITA ROY/Primary Examiner, Art Unit 2146
Prosecution Insights