DETAILED ACTION
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, 3-4, 7-8, 10, 12-13, 16-17, and 19 are presented for examination.
Response to Amendment
The amendments appear to overcome most, but not all, of the previously-made rejections under 35 USC § 112(b). To the extent that an objection or rejection appears in the previous Office Action(s) but not this Office Action, that objection or rejection is withdrawn. To the extent that it appears both in a previous Office Action(s) and this Office Action, the objection or rejection is maintained.
Claim Rejections - 35 USC § 112
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claims 1, 3-4, 7-8, 10, 12-13, 16-17, and 19 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 applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claims 1, 10, and 17 recite “never before seen” input tokens. It is unclear what the reference point is before which the input tokens were “never seen.”
All claims dependent on a claim rejected hereunder are also rejected for being dependent on a rejected base claim.
Claim Rejections - 35 USC § 101
Claims 1, 3-4, 7-8, 10, 12-13, 16-17, and 19 are rejected under 35 U.S.C. 101 because the claims are directed to an abstract idea without significantly more. The analysis of the claims will follow the 2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. 50 (“2019 PEG”).
Claim 1
Step 1: The claim recites a method; therefore, it is directed to the statutory category of processes.
Step 2A Prong 1: The claim recites, inter alia:
[E]ncoding input tokens from a noisy dataset into input vectors: This limitation could encompass mentally encoding the input tokens.
[P]redicting a label based on the input vectors: This limitation could encompass mentally predicting the label.
[C]alculating a noise robust parameter based on the input vectors and the label …, wherein the noise robust parameter is instance-specific for each … instance, and wherein the noise robust parameter has a lower bound predetermined value between zero and one: This limitation could encompass mentally calculating a noise-robust parameter with a value between zero and one. Alternatively, the calculation represents a mathematical concept.
[J]oint training the classifier model and the input encoder using a noise robust loss function and the training dataset, wherein the noise robust loss function is instance specific and based on the noise robust parameter, wherein the noise robust loss function is regularized using an entropy function that is specific to a distribution d associated with randomness in the noisy dataset, and wherein the noise robust parameter used in the noise robust loss function satisfies a predetermined condition: This limitation recites the mathematical concept of training two networks using a loss function that is regularized using an entropy function and that contains a noise robust parameter that satisfies a condition, as shown by paragraphs 23-26 of the instant specification.
[P]erforming the data annotation associated with the natural language processing task on testing data comprising never before seen input tokens: This limitation could encompass mentally performing the data annotation for the natural language task on input tokens.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites that the input tokens are encoded “using an input encoder”, that the noise robust parameter is calculated “using a trained label quality predictor model”, that “the trained label quality predictor model and the input encoder are initially trained using an auxiliary dataset associated with a training dataset instead of using the training dataset”, that each instance is a “training instance”, and that the natural language task is performed “using the trained classifier model and the input encoder”. However, all of these are mere instructions to apply the judicial exception using a generic computer programmed with generic classes of computer algorithm. MPEP § 2106.05(f). The claim further recites “outputting the trained classifier model and the input encoder that are trained jointly for performing data annotation associated with the natural language processing task”. This limitation recites the insignificant extra-solution activity of mere data gathering and output. MPEP § 2106.05(g).
Step 2B: The claim does not contain significantly more than the judicial exception. The analysis at this step mirrors that of step 2A, prong 2, except insofar as the outputting limitation, in addition to reciting insignificant extra-solution activity, also recites the well-understood, routine, and conventional activity of receiving or transmitting data over a network. MPEP § 2106.05(d)(II); OIP Techs., Inc., v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network). As an ordered whole, the claim is directed to an abstract idea of jointly training two networks based on a noise-robust loss function. Nothing in the claim provides significantly more than this. As such, the claim is not patent eligible.
Claim 3
Step 1: A process, as above.
Step 2A Prong 1: The claim recites the same judicial exceptions as in claim 1.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim recites that “the auxiliary dataset comprises a manual correctness label which indicates a manual correctness of a corresponding label in the noisy dataset.” Training the label quality predictor model and input encoder using the auxiliary dataset remains a mere instruction to apply the judicial exception under these further assumptions. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim recites that “the auxiliary dataset comprises a manual correctness label which indicates a manual correctness of a corresponding label in the noisy dataset.” Training the label quality predictor model and input encoder using the auxiliary dataset remains a mere instruction to apply the judicial exception under these further assumptions. MPEP § 2106.05(f).
Claim 4
Step 1: A process, as above.
Step 2A Prong 1: The claim recites, inter alia:
[E]ncoding input tokens from the auxiliary dataset into auxiliary input vectors: This limitation could encompass mentally encoding the tokens.
[C]alculating an auxiliary noise robust parameter based on the auxiliary input vectors and an annotated label in the auxiliary dataset, wherein the auxiliary noise robust parameter is instance-specific for each … instance: This limitation could encompass mentally calculating the auxiliary parameter based on the auxiliary input vectors and an annotated label. Alternatively, the calculation represents a mathematical concept.
[J]oint training of the input encoder and the trained label quality predictor model using a second noise robust function based on the auxiliary noise robust parameter and the manual correctness label: As noted above, jointly training the models using a noise robust function is a mathematical concept.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites that the input tokens are encoded “using the input encoder” and each instance is a “training instance”. However, these are mere instructions to apply the judicial exception using a generic computer programmed with generic classes of computer algorithm. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites that the input tokens are encoded “using the input encoder” and each instance is a “training instance”. However, these are mere instructions to apply the judicial exception using a generic computer programmed with generic classes of computer algorithm. MPEP § 2106.05(f).
Claim 7
Step 1: A process, as above.
Step 2A Prong 1: The claim recites that “prior to calculating the noise robust parameter based on the input vectors and the label, the noise robust parameter is set to one for a limited number of epochs.” This limitation could encompass mentally setting the noise robust parameter to one. Alternatively, setting a value to one is a mathematical concept.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. See claim 1 analysis.
Step 2B: The claim does not contain significantly more than the judicial exception. See claim 1 analysis.
Claim 8
Step 1: A process, as above.
Step 2A Prong 1: The claim recites that “the manual correctness label is set to 1 if the corresponding label in the noisy dataset is accurate.” This limitation could encompass mentally setting the correctness label to 1 if the label in the dataset is accurate.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. See claim 1 analysis.
Step 2B: The claim does not contain significantly more than the judicial exception. See claim 1 analysis.
Claims 10, 12-13
Step 1: The claims recite an apparatus comprising a memory and a processor; therefore, they are directed to the statutory category of machines.
Step 2A Prong 1: The claims recite the same judicial exceptions as in claims 1 and 3-4, respectively.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The analysis at this step mirrors that of claims 1 and 3-4, respectively, except insofar as these claims recite an “apparatus for joint training using neural networks with noise-robust losses, the apparatus comprising: at least one memory configured to store computer program code; at least one processor configured to access the computer program code and operate as instructed by the computer program code”. However, this is a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The analysis at this step mirrors that of claims 1 and 3-4, respectively, except insofar as these claims recite an “apparatus for joint training using neural networks with noise-robust losses, the apparatus comprising: at least one memory configured to store computer program code; at least one processor configured to access the computer program code and operate as instructed by the computer program code”. However, this is a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Claim 16
Step 1: A machine, as above.
Step 2A Prong 1: The claim recites that the “noise robust parameter is higher than a threshold”. Training the networks using the noise robust parameter remains a mathematical concept under these further assumptions.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. See claim 10 analysis.
Step 2B: The claim does not contain significantly more than the judicial exception. See claim 10 analysis.
Claims 17, 19
Step 1: The claims recite a non-transitory computer-readable medium; therefore, they are directed to the statutory category of articles of manufacture.
Step 2A Prong 1: The claims recite the same abstract ideas as in claims 1 and 3, respectively.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The analysis at this step mirrors that of claims 1 and 3, respectively, except insofar as these claims recite a “non-transitory computer readable medium storing a program to execute [the] process”. However, this is a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The analysis at this step mirrors that of claims 1 and 3, respectively, except insofar as these claims recite a “non-transitory computer readable medium storing a program to execute [the] process”. However, this is a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Response to Arguments
Applicant's arguments filed December 22, 2025 (“Remarks”) have been fully considered but they are not persuasive.
Applicant argues that the claims as amended are eligible under 35 USC § 101 because (a) outputting a trained classifier model and input encoder and performing data annotation for a natural language processing task are allegedly not practically mentally performable and are not mathematical concepts; (b) any judicial exception recited is allegedly integrated into a practical application because the specification discloses an improvement in automated annotation of natural language processing datasets and the independent claims’ recitation of joint training, outputting the models, and performing data annotation reflect this alleged improvement; and (c) the instant claims are allegedly analogous to claim 3 of Example 47 because they allegedly recite a practical application. Remarks at 10-14.
Regarding (a), while Examiner agrees in principle that outputting machine learning models is not practically mentally performable, the outputting limitation nonetheless does not amount to significantly more than the judicial exception because it amounts to insignificant extra-solution activity that is well-understood, routine, and conventional, as shown in the rejection itself. And there is nothing innately not mentally performable about performing data annotation for a natural language processing task. But for the recitation that this step is performed using a classifier model and an encoder, the annotation could be performed in the mind, and indeed often is performed manually in practice.
Regarding (b), even assuming arguendo that the specification does disclose an improvement in technology, which Examiner does not concede, any improvement reflected in the claims themselves is to the abstract idea of performing a mathematical model training algorithm and annotating datasets. Specifically, the joint training and data annotation limitations, as noted in the rejection itself, are part of the abstract idea itself and thus by definition cannot provide an inventive concept. MPEP § 2106.05(I). And as noted above, the outputting limitation also does not amount to significantly more than this judicial exception.
Regarding (c), the suggestion that the instant claims reflect a practical application of any judicial exception recited has been responded to above. The instant claims are more appropriately analogized to claim 2 of Example 47. Like claim 2, the claims recite a specific mathematical algorithm for training (using backpropagation with gradient descent in claim 2, using a noise-robust loss function regularized using an entropy function in the instant claims), and the specification specifically recites the mathematical calculations involved in performing the training in paragraphs 23-26. Thus, insofar as the claims recite an improvement in a model training algorithm, such improvement is only to an abstract idea.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to RYAN C VAUGHN whose telephone number is (571)272-4849. The examiner can normally be reached M-R 7:00a-5:00p ET.
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, Kamran Afshar, can be reached at (571) 272-7796. 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.
/RYAN C VAUGHN/Primary Examiner, Art Unit 2125
Prosecution Insights