CTNF 17/666,158 CTNF 96434 DETAILED ACTION Continued Examination Under 37 CFR 1.114 07-42-04 A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 03/19/2026 has been entered. Claims 1-12 and 14-30 are pending in the application and have been examined. Notice of Pre-AIA or AIA Status 07-03-aia AIA 15-10-aia The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA. Examiner Recommendation It is the suggestion of the Examiner, to overcome the Rejections under 35 USC 101, please refer to the Examiner Recommendation in the Office Action mailed 12/19/2025. Further, it is the suggestion of the Examiner if the specifics of the fertility model ( based on Fig. 3) with the relevant specificity is claimed in sufficient detail and tied to the current claim language then it is possible it would overcome 101 but will need to indicate the different layers, the different components to the training and regulation of the length. Response to Amendment The response filed on 03/19/2026 has been correspondingly accepted and considered in this Office Action. Claims 1-12, 14-30 have been examined. Applicant’s amendments to claim 1, 7, 19 and 25, indicating the determining based on source language and target language to use the different ordering algorithms with the support in the Specifications [0073] overcome the 35 U.S.C 112(a) rejections previously set forth in the Office Action mailed 12/19/2025. The dependent claims 2-6, 14-18, 8-12, 20-23 and 26-30 overcome the 35 U.S.C 112(a) rejections previously set forth in the Office Action mailed 12/19/2025 based on their dependency to the amended claim 1, 7, 19 and 25 respectively. Therefore, the above referenced rejections under 35 U.S.C. 112(a) are withdrawn. Response to Arguments Applicant's arguments filed 03/19/2026 have been fully considered as follows: Applicant’s arguments with respect to claim 1 (also representative of claims 7, 19, and 25) state that “None of Chen, Gu or Ran, alone or in combine, use a neural network to determine, based at l east in part on a source language of a source text string to be translated and a target language for a translated text string, whether to use a p re-ordering algorithm or a post-ordering algorithm as a re-ordering algorithm for use in translating the source text string.” The examiner respectfully disagrees, Ran teaches processing the decoding for translation based on the pseudo translation based on reordering information and the word probability weighs the pseudo translation to generate the translate the output in the target language. During training the training pair is trained for using Reorder NAT for the deterministic guiding decoding (DGD) loss and then further trained to appropriately update the weights for the non-deterministic guiding decoding (NDGD) strategy for the translation into target sentence. The broadest reasonable interpretation of “determine, based at l east in part on a source language of a source text string to be translated and a target language for a translated text string, whether to use a p re-ordering algorithm or a post-ordering algorithm as a re-ordering algorithm for use in translating the source text string” includes using the combination of both the methods, the Reorder NAT can be trained to choose the words for the translation based on reordered ( pre-ordered) or words from the source sentence ( post-order) based on the training for the specific languages was reported in pg. 13729-13730 of Ran. Therefore, Ran teaches determine, based at least in part on a source language of a source text string to be translated and a target language for a translated text string, whether to use a pre-ordering algorithm or a post-ordering algorithm as a re-ordering algorithm for use in translating the source text string and therefore, the rejection of Claims 1 is rejected under 35 U.S.C. 103 are sustained and further updated accordingly. In response to the art rejection(s) of the remainder of dependent claims are rejected under 35 U.S.C 103, in case said claims are correspondingly discussed and/or argued for at least the same rationale presented in Remarks filed 03/19/2026, Examiner respectfully notes as follows. For completeness, should the mentioned claims be likewise traversed for similar reasons to independent claims 1 7, 19, and 25 correspondingly, Examiner respectfully directs Applicant to the same previous supra reasons provided in the response directed towards claims 1 7, 19, and 25 correspondingly discussed above. For at least the same supra provided reasons, Examiner likewise respectfully disagrees, and Applicant's arguments have been fully considered but they are not persuasive. Applicant’s arguments with respect to Rejection under 35 U.S.C 101 state that “While disagreeing with this rejection, Applicant has amended the claims to further clarify eligibility. The claims are amended to recite use of a neural network to determine, based at least in part on a source language of a source text string to be translated and a target language for a translated text string, whether to use a pre-ordering algorithm or a post-ordering algorithm as a re-ordering algorithm for use in translating the source text string. ….Thus, the claims recite a specific "ordered combination" of features as an inventive concept "significantly more" than any conventional translation approach. Moreover, the determination of a specific re-ordering algorithm based at least in pa rt on a source language of a source text string to be translated and a target language for a translated text string, followed by use of the determined re- ordering algorithm is a practical application.” The examiner respectfully disagrees, the amendments to the claims 1, 7, 19 and 25 represent reordering of translated text string based on the type of algorithm selected using a neural network, the claims as drafted, is a process that, under its broadest reasonable interpretation, recites mental process, nothing in the claim element precludes the step from practically being performed by a computer. Since the added limitations recites the additional element of generating through a “neural network”, which are recited at a high level of generality and amounts to merely using a computer as a tool to perform an abstract idea or mere instructions to apply the exception using a generic computer component. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the insignificant extra-solution activities abstract idea but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Based on MPEP 2106(c) A Claim That Requires a Computer May Still Recite a Mental Process, the claim itself is a mental process and hence improving mental process is still a mental process. The rejections of claims 1-12, 13-30 under 35 U.S.C. 101 are sustained and updated accordingly. Claim Rejections - 35 USC § 101 07-04-01 AIA 07-04 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. Claims 1-30 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. According to USPTO guidelines, a claim is directed to non-statutory subject matter if: STEP 1: the claim does not fall within one of the four statutory categories of invention (process, machine, method of manufacture, or composition of matter), or STEP 2: the claim recites a judicial exception (e.g. an abstract idea) without reciting additional elements that amount to significantly more than the judicial exception, as determined using the following analysis: STEP 2A (Prong 1): Does the claim recite an abstract idea, law of nature, or natural phenomenon? The guidelines provide three groupings of subject matter that are considered abstract ideas: Mathematical concepts- mathematical relationships, formulas or equations, calculations Certain methods of organizing human activity- fundamental economic principles or practices, commercial or legal interactions, managing personal behavior or relationships or interactions between people Mental processes- concepts that are practicably performed in the human mind (including an observation, evaluation, judgement, or opinions) STEP 2A (Prong 2): Does the claim recite additional elements that integrate the judicial exception into a practical application? The guidelines provide the following exemplary considerations that are indicative than an additional element (or combination of elements) may have integrated the judicial exception into a practical application: an additional element reflects an improvement in the functioning of a computer, or an improvement to other technology or technical field; an additional element that applies or uses a judicial exception to affect a particular treatment or prophylaxis for a disease or medical condition; an additional element implements a judicial exception with, or uses a judicial exception in conjunction with, a particular machine or manufacture that is integral to the claim; an additional element effects a transformation or reduction of a particular article to a different state or thing; and an additional element applies or uses the judicial exception in some other meaningful way beyond generally linking the use of the judicial exception to a particular technological environment, such that the claim as a whole is more than a drafting effort designed to monopolize the exception. While the guidelines further state that the exemplary considerations are not an exhaustive list and that there may be other examples of integrating the exception into a practical application, the guidelines also list examples in which a judicial exception has not been integrated into a practical application: an additional element merely recites the words “apply it” (or an equivalent) with the judicial exception, or merely includes instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea; an additional element adds insignificant extra-solution activity to the judicial exception; and an additional element does no more than generally link the use of a judicial exception to a particular technological environment or field of use. STEP 2B: Does the claim recite additional elements that amount to significantly more than the judicial exception? Consider whether an additional element or combination of elements: adds a specific limitation or combination of limitations that are not well-understood, routine, or conventional activity in the field, which is indicative that an inventive concept may be present; or simply appends well-understood, routine and conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception, which is indicative that an inventive concept may not be present. Using the two-step inquiry, claim 1 is directed to an abstract idea as show below: STEP 1: Does the claim recite an abstract idea, law of nature, or natural phenomenon? YES. Claim 1 is directed to a processor system. STEP 2A (Prong 1): Does the claim recite an abstract idea, law of nature, or natural phenomenon? YES. The claim recites an abstract idea: The limitation of determine based at least in part on a source language of a source text string to be translated and a target language for a translated text string, whether to use a pre- ordering algorithm or a post-ordering algorithm as a re- ordering algorithm for use in translating, the source text string, wherein the pre- ordering algorithm is to reorder words of the source text string prior to translation of the source text string, and the post-ordering algorithm is to reorder words of an output text string after translation of the source text string , as drafted, is a process that, under its broadest reasonable interpretation, recites determining the selection of the mathematical formula or calculation based on the languages for translation, the selection of the appropriate method can be determined by a human and the calculations can be performed using computers, hence the claim as drafted is a process that, under its broadest reasonable interpretation, recites mathematical formula or calculation but for the recitation of generic computer components. The limitation of generate the translated text string based, at least in part, on re-ordering of the source text string or the output text string according to the determined re-ordering algorithm , as drafted, is a process that, under its broadest reasonable interpretation, recites mathematical formula or calculation but for the recitation of generic computer components. STEP 2A (Prong 2): Does the claim recite additional elements that integrate the judicial exception into a practical application? NO. Claim 1 recites the additional element of generating through a “neural network”, which are recited at a high level of generality and amounts to merely using a computer as a tool to perform an abstract idea or mere instructions to apply the exception using a generic computer component. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the insignificant extra-solution activities abstract idea but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. This judicial exception is not integrated into a practical application. STEP 2B: Does the claim recite additional elements that amount to significantly more than the judicial exception? NO. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the additional element of using a neural network amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept. Claim 1 is not patent eligible. Claims 7, 19 and 23 are analogous to claims 1 respectively as directed to a system using processors and methods, the processing device to perform the operations set forth in claim 1, and are subjected to the same rejection as claims 1 respectively. Claims 2-6 further specifies providing the processing of the text strings or translated text strings by the neural network and is a process that, under its broadest reasonable interpretation, is a data gathering process (insignificant extra-solution activity) and does not reflect an improvement in the functioning of a technology or computer . The claim is not patent eligible. Claims 8-12 further specifies providing the processing of the text strings or translated text strings by the neural network and is a process that, under its broadest reasonable interpretation, is a data gathering process (insignificant extra-solution activity) and does not reflect an improvement in the functioning of a technology or computer . The claim is not patent eligible. Claims 14-18 further specifies providing the processing of the text strings or translated text strings by the neural network and is a process that, under its broadest reasonable interpretation, is a data gathering process (insignificant extra-solution activity) and does not reflect an improvement in the functioning of a technology or computer . The claim is not patent eligible. Claims 20-24 further specifies providing the processing of the text strings or translated text strings by the neural network and is a process that, under its broadest reasonable interpretation, is a data gathering process (insignificant extra-solution activity) and does not reflect an improvement in the functioning of a technology or computer . The claim is not patent eligible. Claims 25 further specifies a training method to implement claim 1, and is analogous to claims 1 respectively as directed to a system using processors and methods, the processing device to perform the operations set forth in claim 1, and are subjected to the same rejection as claims 1 respectively. The claim is not patent eligible. Claims 26-30 further specifies providing the formatting of the text strings by the neural network and is a process that, under its broadest reasonable interpretation, is a data gathering process (insignificant extra-solution activity) and does not reflect an improvement in the functioning of a technology or computer . The claim is not patent eligible. Claim Rejections - 35 USC § 103 07-103 AIA The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action. 07-21-aia AIA Claim s 1-7, 10, 12, 14-20, 22 and 24-30 are rejected under 35 U.S.C. 103 as being unpatentable over Ran, Qiu, et al. " Guiding non-autoregressive neural machine translation decoding with reordering information." Proceedings of the AAAI Conference on Artificial Intelligence. Vol. 35. No. 15. 2021 in view of Gu, Jiatao, et al. " Non-autoregressive neural machine translation." arXiv preprint arXiv:1711.02281 (2017) . Regarding claim 1, Ran teaches one or more circuits to use one or more neural networks to: determine select, for based at least in part on a source language of a source text string to be translated and a target language for a translated text string, whether to use a re-ordering algorithm from among a pre- ordering algorithm [[and]] or a post-ordering algorithm as a re- ordering algorithm for use in translatin g ( see Ran, Fig. 1, adds a reordering module between the encoder and decoder modules to explicitly model the reordering information. For original NAT models, the decoder inputs are the copied embeddings of source sentence (No.1 dashed arrow), and for our ReorderNAT model, the decoder inputs are the embeddings of pseudo-translation generated by reordering module (No. 2 dashed arrow); pg. 13729, To be specific, as shown in Figure 1, the major difference between DGD and NDGD strategy is the inputs of decoder module (No. 2 dashed arrow), where the DGD strategy directly utilizes the word embeddings of generated pseudo-translation and the NDGD strategy utilizes the word embeddings weighted by the word probability of pseudo translation; interpreted as determining based on the source language or target language etymologies of words, the weights of the decoder block are adjusted to use path 1 and path 2 embeddings; the weights for the decoder during training to determine the reorder(DGD) or probabilities of the words ( NDGD) is interpreted as the determining which algorithm ( preorder) or (post order) to be used based on the source language of source text string) ) , the source text string to be translated, wherein the pre- ordering algorithm is to reorder words of the source text string prior to translation of the source text string ( see Ran, pg. 13729, Different from original NAT, the input of our decoder module is the embeddings of pseudo-translation instead of copied embeddings of source sentence, which is used to guide the word selection; Ran pg. 13758, Reordering module, Reordering Module The reordering module determines the source-side information of each target word by learning to translate the source sentence into a pseudo-translation(preorder) ) , and the post-ordering algorithm is to reorder words of [[the]] an output text string after translation of the source text string ( see Ran , pg. 13728, copy source word representations as the input of the decoder. Hence, when translating a sentence, NAT models could predict all target words with their maximum likelihood individually by breaking the dependency among them, and therefore the decoding procedure of NAT ( as path 1 in Fig. 1 to decoder block ) ) ; and generate [[a]] the translated text string based, at least in part, on re-ordering of the source text string or the output text string according to the determined re-ordering algorithm selected for the text string to be translated ( see Ran, Fig. 1 output after Softmax Layer, as translated string ) . Ran teaches post ordering of output text string using the Non-autoregressive neural machine translation (NAT) is first proposed by Gu, Gu further teaches the post-ordering algorithm is to reorder words of [[the]] an output text string after translation of the source text string ( see Gu, sect 3.3, Fertility prediction (post ordering) As shown in Fig. 2, we model the fertility pF (ft 0 |x1:T0 ) at each position independently using a one-layer neural network with a softmax classifier (L = 50 in our experiments) on top of the output of the last encoder layer. This models the way that fertility values are a property of each input word but depend on information and context from the entire sentence ) ; and generate [[a]] the translated text string based, at least in part, on the output text string for the text string to be translated ( see Gu, sect 3.4, At inference time, the model can identify the translation with the highest conditional probability (see Eq. 5) by marginalizing over all possible latent fertility sequences. Given a fertility sequence, however, identifying the optimal translation only requires independently maximizing the local probability for each output position ) . Ran teaches a translation model based on the Reordering module in a Transformer network. Using the known technique of non-autoregressive translation model based on the Transformer network ( see Gu, sect.1 ), to provide the improved computation speed in the reference of Ran would have been obvious to one of ordinary skill in the art ( see Ran, pg. 13727) . Regarding claim 2, Ran in view of Gu teach the processor of claim 1 . Ran teaches re-order, based at least in part on a source language of the text string, one or more elements of the text string to generate a re-ordered source text string, wherein the determined re-ordering algorithm is the pre-ordering algorithm ( see Ran, pg. 13729, Different from original NAT, the input of our decoder module is the embeddings of pseudo-translation instead of copied embeddings of source sentence, which is used to guide the word selection; Ran Fig. 1 Reordering module, paths in dashed lines 1, 2, 3; pg. 13729, To be specific, as shown in Figure 1, the major differ ence between DGD and NDGD strategy is the inputs of decoder module (No. 2 dashed arrow), where the DGD strategy directly utilizes the word embeddings of generated pseudo-translation and the NDGD strategy utilizes the word embeddings weighted by the word probability of pseudo translation; interpreted as determining based on the source language or target language etymologies of words, the weights of the decoder block are adjusted to use path 1 and path 2 embeddings ) ; and calculate the length of the translated text string based at least in part on the re- ordered text string before the text string is to be translated ( see Ran, pg. 13729 During inference, the NAT reordering module needs to determine the length of the pseudo-translation. To this end, we use a length predictor and copy the embeddings of the source input of the reordering module similar to existing NAT models. ). Regarding claim 3, Ran in view of Gu teach the processor of claim 2 , Gu further teaches to re-order the one or more elements of the source text string by at least parsing the text string to calculate one or more dependencies among the one or more elements of the source text string ( see Gu, sec 3.2 The positional attention is provided to the decoder layer, this this additional information improves the decoder’s ability to perform local reordering ) . The motivation to combine as claim 1 applies here Regarding claim 4, Ran in view of Gu teach the processor of claim 1 . Gu further teaches use the one or more neural networks to translate the source text string based at least in part on the length of the translated text string ( see Gu, pg. 4-5 • Copy source inputs using fertilities: In this case the source inputs are scanned from left to right at a “speed” that varies inversely with the fertility of each input; the decoding process is now conditioned on the sequence of fertilities, while the resulting output length is determined by the sum of all fertility values. At inference time, the model can identify the translation with the highest conditional probability (see Eq. 5) by marginalizing over all possible latent fertility sequences ). ) . The motivation to combine as claim 1 applies here Regarding claim 5, Ran in view of Gu teach the processor of claim 4 . Gu further teaches wherein the length comprises a number of words for a predicted translated text string ( see Gu, pg. 5 We propose the use of fertilities instead. These are integers for each word in the source sentence that correspond to the number of words in the target sentence that can be aligned to that source word using a hard alignment algorithm like IBM Model 2 ). The motivation to combine as claim 1 applies here Regarding claim 6, Ran in view of Gu teach the processor of claim 1 . Gu further teaches wherein the PNG media_image1.png 383 819 media_image1.png Greyscale one or more neural networks are further to generate a translation of the text string independent of word order ( see Gu, Fig. 2 shows the translation of input string using the NAT ). The motivation to combine as claim 1 applies here Regarding claim 7, is directed to a system claim corresponding to the processor claim presented in claim 1 and is rejected under the same grounds stated above regarding claim 1. Regarding claim 10, Ran in view of Gu teach the system of claim 7 , Ran further teaches obtain the source text string based at least in part on input from one or more translation software applications executing on one or more devices ( Ran, pg. 13730, discusses the processing of the datasets for machine translations ) . Regarding claim 12, Ran in view of Gu teach the system of claim 7 . Gu teaches use a first neural network of the one or more neural networks to predict a length of the translated text string ( see Gu, Fig. 2 Encoder stack and Fertility predictor comprise one or more neural network to predict the length ) ; and use a second neural network of the one or more neural networks to generate the translated text string based, at least in part, on the predicted length ( see Gu, Fig. Decoder stack ( fertility information is predicted length) ). The same motivation to combine as claim 1 applies here Regarding claim 14, Ran in view of Gu teach the processor of claim 1 . Gu further teaches to re-order one or more words of the source text string according to the determined re-ordering algorithm based at least in part on a word order of the target language corresponding to the translated text string ( Gu, sect 3.2, 3,3: including both fertilities and reordering in the latent variable would provide complete alignment statistics. This would make the decoding function trivially easy to approximate given the latent variable and force all of the modeling complexity into the encoder. Using fertilities alone allows the decoder to take some of this burden off of the encoder. At inference time, the model can identify the translation with the highest conditional probability (see Eq. 5) by marginalizing over all possible latent fertility sequences ). Ran further teaches to re-order one or more words of the text string according to the selected re-ordering algorithm based at least in part on a word order of a language corresponding to the translated text string ( see Ran, pg. 13728, ReorderNAT As shown in Figure 1, ReorderNAT employs a reordering module to explicitly model the reordering information in the decoding1 . Formally, ReorderNAT first employs the reordering module to translate the source sentence X into a pseudo translation Z = {z1, · · · , zm} which reorganizes source sentence structure into the target language, and then uses the decoder module to generate target translation Y based on the pseudo-translation. )) . the same motivation to combine as claim 1. Regarding claim 15, Ran in view of Gu teach the processor of claim 1 , Ran further teaches wherein the word order indicates how words are arranged in a sequence in a sentence ( see Ran, pg. 13728, Reordering Module The reordering module determines the source-side information of each target word by learning to translate the source sentence into a pseudo-translation. Ran, Fig.1 ) . Regarding claim 16, Ran in view of Gu teach the processor of claim 1 . Gu teaches to use at least one or more encoder and decoder layers to translate the text string ( see Gu, Fig. 2 Encoder and Decoder Stack ) . The same motivation to combine as claim 1 applies here. Regarding claim 17, Ran in view of Gu teach the processor of claim 1 . Gu further teaches to calculate one or more embeddings based at least in part on the text string, wherein the one or more embeddings correspond to at least one or more positions of one or more words in the source text string ( see Gu, sect 3.2 Copy source inputs using fertilities: A more powerful way, depicted in Fig. 2 and discussed in more detail below, is to copy each encoder input as a decoder input zero or more times, with the number of times each input is copied referred to as that input word’s “fertility.” ) . The same motivation to combine as claim 1 applies here. Regarding claim 18, is directed to a machine-readable medium claim corresponding to the system claim presented in claim 12 and is rejected under the same grounds stated above regarding claim 12. Regarding claim 19, is directed to a processor claim corresponding to the processor claim presented in claim 1, Gu further teaches to train one or more neural networks ( see Gu, sect 4 ): and is rejected under the same grounds stated above regarding claim 1. Regarding claim 20, Ran in view of Gu teach the processor of claim 19 . Gu teaches to calculate the length of the translated text string as a set of integers comprising one or more numbers of words for one or more words of the source text string ( see Gu, pg. 2 the target sequence length T can be modeled with a separate conditional distribution p L , eq. (3), pg. 4 We propose the use of fertilities instead. These are integers for each word in the source sentence that correspond to the number of words in the target sentence that can be aligned to that source word using a hard alignment algorithm like IBM Model 2 (Brown et al., 1993). ) and train the one or more neural networks by at least comparing the set of integers with a set of reference integers corresponding to a translated text string ( see Gu, pg. 6, We choose a proposal distribution q defined by a separate, fixed fertility model. Possible options include the output of an external aligner, which produces a deterministic sequence of integer fertilities for each (source, target) pair in a training corpus, or fertilities computed from the attention weights used in our fixed autoregressive teacher model. This simplifies the inference process considerably, as the expectation over q is deterministic. The resulting loss function, consisting of the two bracketed terms in Eq. 9, allows us to train the entire model in a supervised fashion, using the inferred fertilities to simultaneously train the translation model p and supervise the fertility neural network model p F ) . The same motivation to combine as claim 1 applies here. Regarding claim 22, Ran in view of Gu teach the processor of claim 19 . Gu teaches to calculate one or more embeddings based at least in part on the source text string indicating one or more classifications of one or more words in the source text string ( see Gu, sect 3.3. As shown in Fig. 2, we model the fertility pF (ft 0 |x1:T0 ) at each position independently using a one-layer neural network with a softmax classifier (L = 50 in our experiments) on top of the output of the last encoder layer. This models the way that fertility values are a property of each input word but depend on information and context from the entire sentence. ) . The same motivation to combine as claim 1 applies here. Regarding claim 24, Ran in view of Gu teach the processor of claim 20 . Gu teaches train a first portion of the one or more neural networks to infer the length of the translated text string ( see Gu, sect 4.1, 4.2 teaches the training the fertility model ) ; and train a second portion of the one or more neural networks to generate a translated text string using the inferred length ( see Gu, sect 4 The resulting loss function, consisting of the two bracketed terms in Eq. 9, allows us to train the entire model in a supervised fashion, using the inferred fertilities to simultaneously train the translation model p and supervise the fertility neural network model pF (network shown in Fig. 2) ). The same motivation to combine as claim 1 applies here. Regarding claim 25, is directed to a method claim corresponding to the processor claim presented in claim 19 and is rejected under the same grounds stated above regarding claim 19. Ran further teaches training one or more neural networks ( see Ran, pg. 13730 ). Regarding claim 26, Ran in view of Gu teach the method of claim 25 . Gu teaches to predict a length of the translated text string by at least processing one or more embeddings corresponding to the source text string ( see Gu, pg. 4 In this case the source inputs are scanned from left to right at a “speed” that varies inversely with the fertility of each input; the decoding process is now conditioned on the sequence of fertilities, while the resulting output length is determined by the sum of all fertility values ). The same motivation to combine as claim 1 applies here. Regarding claim 27, Ran in view of Gu teach the method of claim 26 .Gu further teaches computing loss based at least in part on the calculated length of the translated text string and a reference length corresponding to a translated text string ( see Gu, pg. 7 equation 11 ) ; and updating the one or more neural networks based at least in part on the computed loss ( see Gu, pg. 7 Then we train the whole model jointly with a weighted sum of the original distillation loss and two such terms, one an expectation over the predicted fertility distribution, normalized with a baseline, and the other based on the external fertility inference model ). The same motivation to combine as claim 1 applies here. Regarding claim 28, is directed to a method claim corresponding to the processor claim presented in claim 24 and is rejected under the same grounds stated above regarding claim 24. Regarding claim 29, Ran in view of Gu teach the method of claim 25 . Ran further teaches wherein the determined re-ordering algorithm is the pre-ordering algorithm, the method further comprising: calculating dependencies among words of the source text string ( see Ran, pg. 13729, During inference, the NAT reordering module needs to determine the length of the pseudo-translation. To this end, we use a length predictor and copy the embeddings of the source sentence as the input of the reordering module similar to existing NAT model ) ) ; using the dependencies to re-order the words of the source text string to generate a re- ordered text string ( see Ran, pg. 13729, AT Reordering Module: We find that AT models are more suitable for modeling the reordering information compared to NAT models, and even a light AT model with similar decoding speed to a large NAT model could achieve better performance in modeling reordering information. Hence, we also introduce a light AT model to model the pseudo translation probability ) ; and using at least the re-ordered text string and a predicted length to translate the text string to generate the translated text string ( see Ran, pg. 13729, The decoder module translates the source sentence into the target translation with the guiding of pseudo-translation, which regards the translation of each word as NAT. Fig. 1, Decoder module ). Regarding claim 30, Ran in view of Gu teaches the method of claim 25 . Ran teaches performing training using at least one of. an order of subject-predicate ordering distortion prediction task, an order of predicate-object order distortion prediction task, and an order of n-tuples of predicate-argument structures prediction task ( see Ran, pg. 13733, Zhang et al. (2017) presented three distortion models to further incorporate reordering knowledge into attention-based NMT models. This work empirically justifies reordering information is essential for NAT ) . 07-21-aia AIA Claim s 8, 9, 11 are rejected under 35 U.S.C. 103 as being unpatentable over Ran, Qiu, et al. " Guiding non-autoregressive neural machine translation decoding with reordering information." Proceedings of the AAAI Conference on Artificial Intelligence. Vol. 35. No. 15. 2021 in view of Gu, Jiatao, et al. " Non-autoregressive neural machine translation." arXiv preprint arXiv:1711.02281 (2017) further in view of Zhang et. al. US 2021/0042475 . Regarding claim 8, Ran in view of Gu teach the system of claim 7 , Gu further teaches predict a length of the translated text string based on the identified language ( see Gu, pg. 4 • In this case the source inputs are scanned from left to right at a “speed” that varies inversely with the fertility of each input; the decoding process is now conditioned on the sequence of fertilities, while the resulting output length is determined by the sum of all fertility values ) ; translate the source text string to generate an initially translated text string based on the length ( see Gu, Fig. 2 and pg. 4-5 At inference time, the model can identify the translation with the highest conditional probability (see Eq. 5) by marginalizing over all possible latent fertility sequences ) ; and re-order one or more elements of the initially translated text string, wherein the determined re-ordering algorithm is the post-ordering algorithm ( see Gu, Fig. 5 ) . However, Ran in view of Gu fails to teach identify the language corresponding to the text string. However, Zhang teaches identify the language corresponding to the text string ( see Zhang, [0178- 0179] the classification and selection ( identify) is made by machine translation selector 306 ) and re-order one or more elements of the translated text string ( see Zhang, [0181] The post-editor module 310 generates a predicted post-edit to the first translated text element 312 to produce the second translated text element 314 ). Ran in view of Gu teaches a translation model based on the Transformer network. Using the known technique of identifying the language corresponding to the text string teachings by Zhang ( see Zhang, [0178-0179] ), to provide the identification of the source language in the reference Ran in view of Gu would have been obvious to one of ordinary skill in the art. Regarding claim 9, Ran in view of Gu further in view of Zhang teach the system of claim 8 . Gu further teaches re-order the one or more elements of the initially translated text string based on the source language corresponding to the source text string and the target language corresponding to the translated text string ( see Gu, Fig. 2, output of Translation Predictor, Gu, Fig. 5; Gu, sect 3.4 ) . Zhang further teaches re-order the one or more elements of the initially translated text string based on the source language corresponding to the source text string and the target language corresponding to the translated text string ( see Zhang, [0203] discusses the post-edit of the first translation ( reordering based on the language of the translation) ). The same motivation to combine as claim 8 applies here. Regarding claim 11, Ran in view of Gu further in view of Zhang teach the system of claim 8 . Gu further teaches the length indicates at least a maximum number of words for a translation of a word of the source text string ( see Gu, pg. 5 Using fertilities as a latent variable makes significant progress towards solving the multimodality problem by providing a natural factorization of the output space(interpreted as maximum number of output space). Given a source sentence, restricting the output distribution to those target sentences consistent with a particular fertility sequence dramatically reduces the mode space. Furthermore, the global choice of mode is factored into a set of local mode choices: namely, how to translate each input word. These local mode choices can be effectively supervised because the fertilities provide a fixed “scaffold.” ) . 07-21-aia AIA Claim 21 is rejected under 35 U.S.C. 103 as being unpatentable over Ran, Qiu, et al. " Guiding non-autoregressive neural machine translation decoding with reordering information." Proceedings of the AAAI Conference on Artificial Intelligence. Vol. 35. No. 15. 2021 in view of Gu, Jiatao, et al. " Non-autoregressive neural machine translation." arXiv preprint arXiv:1711.02281 (2017) further in view of Shamir US Patent 11,436,496 . Regarding claim 21, Ran in view of Gu teach the processor of claim 19 , however Ran in view of Gu fails to teach train the one or more neural networks using at least one or more minimum square error (MSE) loss functions. However, Shamir teaches train the one or more neural networks using at least one or more minimum square error (MSE) loss functions ( see Samir, col 11 lines 5-10 The innovation loss is designed to push the network to maximize the innovation of each of the neurons in the layer. This implies that the (minimal) mean square prediction errors should be maximized). Ran in view of Gu teaches a translation model based on the Transformer network. Using the known technique of training neural networks using MSE loss function teachings by Shamir ( see Shamir col 11 lines 5-10 ), to provide the training based on MSE loss in the reference Ran in view of Gu would have been obvious to one of ordinary skill in the art . 07-21-aia AIA Claim 23 is rejected under 35 U.S.C. 103 as being unpatentable over Ran, Qiu, et al. " Guiding non-autoregressive neural machine translation decoding with reordering information." Proceedings of the AAAI Conference on Artificial Intelligence. Vol. 35. No. 15. 2021 in view of Gu, Jiatao, et al. " Non-autoregressive neural machine translation." arXiv preprint arXiv:1711.02281 (2017) further in view of Wang et. al. US PgPub. 2023/0076471 . Regarding claim 23, Ran in view of Gu teaches the processor of claim 19 , however fails to teach to perform training using one or more tasks corresponding to at least one of: predicate masking, subject phrase masking, and object phrase masking. However, Wang teaches to perform training using one or more tasks corresponding to at least one of: predicate masking, subject phrase masking, and object phrase masking ( see Wang, [0057] According to embodiments of the present disclosure, the source sample text data may be obtained by masking at least one target object(object phrase) in original source sample text data, and the sample feature vector sequence may include a sample feature vector corresponding to each of the at least one target object). Ran in view of Gu teaches a translation model based on the Transformer network. Using the known technique of target object masking teachings by Wang ( see Wang[0057] ), to provide the training based on object phrase masking in the reference Ran in view of Gu would have been obvious to one of ordinary skill in the art . Conclusion 07-96 AIA The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Sudo et. al. JP 2014170296 teaches word order rearrangement and translation method ( see Sudo, abstract ). Sudoh, Katsuhito, et al . "Post-ordering in statistical machine translation." Proceedings of Machine Translation Summit XIII: Papers. 2011 teaches the translation in the opposite direction and proposes post-ordering; foreign sentences are first translated into foreign-ordered English, and then reordered into correctly-ordered English( see Sudoh, abstract ). Hopkins et. al. US PgPub. 2012/0323554 teaches post-processing to reorder words (phrases), generate the sentence in the target language ( Hopkins, Fig. 3 ). Song et. al. US Patent 10,108,607 teaches a reordering model is which is trained to score the reordering based results of alignment between a sentence pair and a word pair in two languages ( see Song, col 6 lines 7-36 ). Any inquiry concerning this communication or earlier communications from the examiner should be directed to NANDINI SUBRAMANI whose telephone number is (571)272-3916. The examiner can normally be reached Monday - Friday 12:00pm - 5:00 pm EST. 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, Bhavesh M Mehta can be reached at (571)272-7453. 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. /NANDINI SUBRAMANI/ Examiner, Art Unit 2656 Application/Control Number: 17/666,158 Page 2 Art Unit: 2656 Application/Control Number: 17/666,158 Page 3 Art Unit: 2656 Application/Control Number: 17/666,158 Page 4 Art Unit: 2656 Application/Control Number: 17/666,158 Page 5 Art Unit: 2656 Application/Control Number: 17/666,158 Page 6 Art Unit: 2656 Application/Control Number: 17/666,158 Page 7 Art Unit: 2656 Application/Control Number: 17/666,158 Page 8 Art Unit: 2656 Application/Control Number: 17/666,158 Page 9 Art Unit: 2656 Application/Control Number: 17/666,158 Page 10 Art Unit: 2656 Application/Control Number: 17/666,158 Page 11 Art Unit: 2656 Application/Control Number: 17/666,158 Page 12 Art Unit: 2656 Application/Control Number: 17/666,158 Page 13 Art Unit: 2656 Application/Control Number: 17/666,158 Page 14 Art Unit: 2656 Application/Control Number: 17/666,158 Page 15 Art Unit: 2656 Application/Control Number: 17/666,158 Page 16 Art Unit: 2656 Application/Control Number: 17/666,158 Page 17 Art Unit: 2656 Application/Control Number: 17/666,158 Page 18 Art Unit: 2656 Application/Control Number: 17/666,158 Page 19 Art Unit: 2656 Application/Control Number: 17/666,158 Page 20 Art Unit: 2656 Application/Control Number: 17/666,158 Page 21 Art Unit: 2656 Application/Control Number: 17/666,158 Page 22 Art Unit: 2656 Application/Control Number: 17/666,158 Page 23 Art Unit: 2656 Application/Control Number: 17/666,158 Page 24 Art Unit: 2656 Application/Control Number: 17/666,158 Page 25 Art Unit: 2656 Application/Control Number: 17/666,158 Page 26 Art Unit: 2656
Prosecution Insights