DETAILED ACTION
Introduction
This office action is in response to applicant’s amendment filed 8/20/2026. Claims 1-20 are currently pending and have been examined. Applicant’s IDS have been considered. There is no claim to foreign priority.
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 .
Response to Arguments
Applicant’s arguments, see remarks, filed 8/20/2026, with respect to the Double Patenting rejection of claims 1-20 have been fully considered and are persuasive, based on the applicant’s Terminal Disclaimer filed 8/30/2026. The Double Patenting rejection of claims 1-20 have been withdrawn.
Applicant’s arguments, see remarks, filed 8/20/2026, with respect to the rejection(s) of claim(s) 1-20 under 35 USC 103 have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made in view of Wehrmann et. al (Wehrmann, Hierarchical Multi-Label Classification Networks), in view of Anonymous et al. (Anonymous, Bert for sequence-to-sequence multi-label text classification). The Examiner notes the neural network architecture of Costello is different than that of Wehrmann and does not fully rely on Costello for teaching the particulars of the neural network, rather the NLP input processing, and output of the neural network, and system components and other elements that are not explicitly disclosed in Wehrmann, processors, non-transitory computer-readable storage devices, and other elements that may not be common found in non-patent literature. The combination of Wehrmann with Anonymous, defining “x”, the input feature vector, as a trained NLP algorithm, such as BERT, as applied to intent detection and input into multiple models. The overall combination of references making obvious the applicant’s current claims (see rejection below).
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.
Claim(s) 1-4, 6-14 and 16-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Costello (US 2020/0065384) in view of Wehrmann et. al (Wehrmann, Hierarchical Multi-Label Classification Networks), in view of Anonymous et al. (Anonymous, Bert for sequence-to-sequence multi-label text classification) and further in view of Bhagavath et al. (Bhagavath, US 2023/0086302).
As per claim 1, Costello teaches a system comprising:
one or more processors (paragraph [0027]-his processor(s), computer-readable media and corresponding instructions);
and one or more non-transitory computer-readable storage devices storing computing instructions configured to run on the one or more processors and cause the one or more processors to perform (ibid):
receiving one or more user utterances (paragraph [0073]-his input speech utterance);
using a trained NLP algorithm as one or more layers in a neural network to generate from an input comprising the one or more user utterances at least one first output of at least one first output layer of the neural network (paragraphs [0077-0079, 0006], Figs. 1A-1B, 3-5-his NLP-intent classification, first output layer, and the utterance also input into a second layer, which generated a second output of the neural network), [wherein the trained NLP algorithm further generates an intermediate output from the input that is passed into at least one of a linear or hidden layer of the neural network, and wherein an output of the at least one of the linear or hidden layer is combined with the at least first output to generate at least one second output of at least one second output layer of the neural network;
combining the at least one first output of the at least one first output layer of the neural network and the at least one second output of the at least one second output layer of the neural network to create a combined output of the neural network, wherein:
the at least one first output layer of the neural network is different than the at least one second output layer of the neural network;] and
coordinating [displaying a customized graphical user interface (GUI) using] the combined output of the neural network (ibid-paragraphs [0076]-his final prediction result output).
Costello lacks teaching that Wehrmann teaches, using a trained NLP algorithm as one or more layers in a neural network to generate from an input comprising the one or more user utterances at least one first output of at least one first output layer of the neural network), wherein the trained NLP algorithm further generates an intermediate output from the input that is passed into at least one of a linear or hidden layer of the neural network, and wherein an output of the at least one of the linear or hidden layer is combined with the at least first output to generate at least one second output of at least one second output layer of the neural network (pages 2-3, section 2, 2.1 and 2.2, Fig. 1-his input, and corresponding intermediate output into a first layer of the neural network, see edge from X, and the output from the first layer, his edge from the global to local FC layers, along with a first output, “the local flows, which also begin in the input layer and pass by their respective global FC layers but also through specific local FC layers, finally ending at the corresponding local output, used to generate a second output from a second layer);
combining the at least one first output of the at least one first output layer of the neural network and the at least one second output of the at least one second output layer of the neural network to create a combined output of the neural network, wherein (ibid-wherein the final prediction, all local outputs are then concatenated and pooled with the global output for consensual prediction):
the at least one first output layer of the neural network is different than the at least one second output layer of the neural network (ibid-global layer output of the neural network being different than the local layer output of the neural network, via his distinct weight matrix for each level).
Thus, it would have been obvious to one of ordinary skill in the linguistics art, before the effective filing date of the invention, as all the claimed elements were known in the prior art and one skilled in the art could have combined the elements as claimed by known methods (computer implemented techniques and algorithms combining processes and steps in natural language processing), in view of the teachings of Costello and Wehrmann to combine the prior art element of a system and intent classifier and determined intent using a neural network as taught by Costello with the Hierarchical Multi-Label Classification Network, that generates an intermediate output neural network, global and local outputs which are combined into a final concatenated output as taught by Wehrmann as each element performs the same function as it does separately, as the combination would yield predictable results, KSR International Co. v. Teleflex Inc., 550 US. -- 82 USPQ2nd 1385 (2007), wherein the predictable result would be to classify data in which classes are hierarchically structured, and objects are assigned to multiple paths of the class hierarchy at the same time (ibid-Wehrmann, abstract).
Costello with Wehrmann lack explicitly teaching that which Anonymous teaches, wherein the trained NLP algorithm further generates an intermediate output from the input that is passed into at least one of a linear or hidden layer of the neural network, and wherein an output of the at least one of the linear or hidden layer is combined with the at least first output to generate at least one second output of at least one second output layer of the neural network (pages 3, 4, Fig. 1-his BERT+SGM, wherein his NLP algorithm, BERT model, see his BERT language, and text encoder discussion, is used to generate a representation for input into multiple models, also having an intermediate output and combination of an output from a layer and a first output, in order to generate an output from a second layer, and a final output, however, herein is relied upon for the particulars of the BERT model, as the NLP algorithm that provides that input into multiple models, see his hybrid method, and “Wehrmann HCMN discussion of jointly optimizing both local (hierarchical) and global classifiers, discussion and adaptation of the BERT model, as the input representation).
Thus, it would have been obvious to one of ordinary skill in the linguistics art, before the effective filing date of the invention, as all the claimed elements were known in the prior art and one skilled in the art could have combined the elements as claimed by known methods (computer implemented techniques and algorithms combining processes and steps in natural language processing), in view of the teachings of Costello and Wehrmann and Anonymous to combine the prior art element of a system and intent classifier and determined intent using a neural network as taught by Costello with the Hierarchical Multi-Label Classification Network, that generates an intermediate output neural network, global and local outputs which are combined into a final concatenated output as taught by Wehrmann with the ensemble and input BERT NLP algorithm, for intermediate representation output and hierarchical neural model, as taught by Anonymous, as each element performs the same function as it does separately, as the combination would yield predictable results, KSR International Co. v. Teleflex Inc., 550 US. -- 82 USPQ2nd 1385 (2007), wherein the predictable result would be to classify data in which classes are hierarchically structured, and objects are assigned to multiple paths of the class hierarchy at the same time, using an input to the different layers from a NLP algorithm thus handling multi-label task classification, including intent detection in dialogue system (ibid-Wehrmann, abstract, Anonymous page 1, section 1 introduction).
Costello lacks explicitly teaching that which Bhagavath teaches, coordinating displaying a customized graphical user interface (GUI) using the combined output of the neural network (paragraphs [0023, 0053]-his configured display for presenting return item results, based on the output of the intent classification model).
Thus, it would have been obvious to one of ordinary skill in the linguistics art, before the effective filing date of the invention, as all the claimed elements were known in the prior art and one skilled in the art could have combined the elements as claimed by known methods (computer implemented techniques and algorithms combining processes and steps in natural language processing), in view of the teachings of Costello and Bhagavath to combine the prior art element of a hierarchical intent classifier and determined intent using a neural network as taught by Costello with using the generated intent from a neural network to configure a display interface and present results to a user as taught by Bhagavath as each element performs the same function as it does separately, as the combination would yield predictable results, KSR International Co. v. Teleflex Inc., 550 US. -- 82 USPQ2nd 1385 (2007), wherein the predictable result would be to allow a chatbot to generate a customized display to respond to different determined intents as needed (ibid-Bhagavath, paragraph [0023]).
As per claims 2 and 12, Costello with Wehrmann with Anonymous with Bhagavath make obvious the system of claim 1, Bhagavath further teaching that which Costello lacks, wherein the customized graphical user interface (GUI) comprises a reply from a chat bot generated based on the combined output of the neural network (ibid-see claim 1, Bhagavath, paragraphs [0023, 0053]-his configured display for presenting return item results, based on the output of the intent classification model, as similarly motivated in combination).
As per claims 3 and 13, Costello with Wehrmann with Anonymous with Bhagavath make obvious the system of claim 1, wherein the one or more user utterances comprise messages entered into a chat bot (ibid-Costello, paragraph [0155]-his chatbot).
As per claims 4 and 14, Costello with Wehrmann with Anonymous with Bhagavath make obvious the system of claim 1, wherein the trained NLP algorithm, as used in the neural network, feeds into: the at least one first output layer of the neural network and the at least one second output layer of the neural network through two different pathways (ibid-Anonymous, Fig, 1-his BERT as the trained NLP algorithm, feeding into a first output layer, and one second output layer through two different pathways, as similarly motivated and combined).
As per claims 6 and 16, Costello further makes obvious the system of claim 1, wherein: the computing instructions are further configured to run on the one or more processors and cause the one or more processors to perform:
after using the trained NLP algorithm as the one or more layers in the neural network, receiving a new user utterance from a user (Costello, Figs. 6, S100, S200-his trained model, and subsequent user input utterance); and the combined output of the neural network is correlated with an intent of the new user utterance (ibid-Figs. 6, his output intent from neural network, as correlated with the new user utterance).
Costello lacks teaching that which Anonymous teaches, after using the trained NLP algorithm as the one or more layers in the neural network, [receiving a new user utterance from a user] (ibid-see Anonymous, BERT as the used NLP algorithm in the neural network, see claim 1, NLP discussion, and user input utterance, and intent detection in dialogue systems, page 1, section 1); and
the combined output of the neural network is correlated with an intent of the new user utterance (ibid, the combined ensemble, and hybrid hierarchical neural network, and intent detection in dialogue, as the final output, from the neural network, as discussed in claim 1).
Thus, it would have been obvious to one of ordinary skill in the linguistics art, before the effective filing date of the invention, as all the claimed elements were known in the prior art and one skilled in the art could have combined the elements as claimed by known methods (computer implemented techniques and algorithms combining processes and steps in natural language processing), in view of the teachings of Costello and Wehrmann and Anonymous to combine the prior art element of a system and intent classifier and determined intent, from new user utterances, after training a model, using a neural network as taught by Costello with the Hierarchical Multi-Label Classification Network, that generates an intermediate output neural network, global and local outputs which are combined into a final concatenated output as taught by Wehrmann with the ensemble and input BERT NLP algorithm, for intermediate representation output and hierarchical neural model, for intent detection in dialogue, as taught by Anonymous, as each element performs the same function as it does separately, as the combination would yield predictable results, KSR International Co. v. Teleflex Inc., 550 US. -- 82 USPQ2nd 1385 (2007), wherein the predictable result would be to classify data in which classes are hierarchically structured, and objects are assigned to multiple paths of the class hierarchy at the same time, using an input to the different layers from a NLP algorithm thus handling multi-label task classification, including intent detection in dialogue system, for a new user utterance, (ibid-Costello-utterance discussion, Wehrmann, abstract, Anonymous page 1, section 1 introduction).
As per claims 7 and 17, Costello with Wehrmann with Anonymous with Bhagavath make obvious the system of claim 1, Wehrmann further teaching, that which the others lack, wherein the neural network comprises a hybrid neural network comprising at least a portion of two different types of hierarchical multi-label classification networks (ibid, Wehrmann, pages 1, section 1-his two different types of HMCN, global and local, see his “hybrid” HMC discussion).
Thus, it would have been obvious to one of ordinary skill in the linguistics art, before the effective filing date of the invention, as all the claimed elements were known in the prior art and one skilled in the art could have combined the elements as claimed by known methods (computer implemented techniques and algorithms combining processes and steps in natural language processing), in view of the teachings of Costello and Wehrmann to combine the prior art element of a system and intent classifier and determined intent using a neural network as taught by Costello with the Hierarchical Multi-Label Classification Network, that generates an intermediate output neural network, global and local outputs which are combined into a final concatenated output as taught by Wehrmann as each element performs the same function as it does separately, as the combination would yield predictable results, KSR International Co. v. Teleflex Inc., 550 US. -- 82 USPQ2nd 1385 (2007), wherein the predictable result would be to classify data in which classes are hierarchically structured, and objects are assigned to multiple paths of the class hierarchy at the same time (ibid-Wehrmann, abstract).
As per claims 8 and 18, Costello with Wehrmann with Anonymous with Bhagavath make further makes obvious the system of claim 1, wherein the neural network comprises a hybrid neural network comprising at least one chained portion and at least one unchained portion (ibid-Anonymous, Fig. 1, his hybrid neural network, comprising BERT + SGM, as one trained model unchained portion, along with a chained portion, together as an ensemble).
Thus, it would have been obvious to one of ordinary skill in the linguistics art, before the effective filing date of the invention, as all the claimed elements were known in the prior art and one skilled in the art could have combined the elements as claimed by known methods (computer implemented techniques and algorithms combining processes and steps in natural language processing), in view of the teachings of Costello and Wehrmann and Anonymous to combine the prior art element of a system and intent classifier and determined intent using a neural network as taught by Costello with the Hierarchical Multi-Label Classification Network, that generates an intermediate output neural network, global and local outputs which are combined into a final concatenated output as taught by Wehrmann with the ensemble, hybrid model having a chained and unchained portion, as taught by Anonymous, as each element performs the same function as it does separately, as the combination would yield predictable results, KSR International Co. v. Teleflex Inc., 550 US. -- 82 USPQ2nd 1385 (2007), wherein the predictable result would be to classify data in which classes are hierarchically structured, and objects are assigned to multiple paths of the class hierarchy at the same time, using an input to the different layers from a NLP algorithm thus handling multi-label task classification, via the hybrid model, including intent detection in dialogue system (ibid-Wehrmann, abstract, Anonymous page 1, section 1 introduction).
As per claims 9 and 19, Costello with Wehrmann with Anonymous with Bhagavath further makes obvious the system of claim 1, wherein the one or more user utterances comprise requests to return, exchange, or refund one or more items (ibid-see claim 1, Costello, paragraphs [0023, 0053]-his configured display for presenting return item results, based on the output of the intent classification model, as similarly combined and motivated).
As per claim 10, Costello with Bhagavath make obvious the system of claim 1, wherein coordinating displaying the customized GUI using the combined output of the neural network comprises:
calculating a total loss using the at least one first output of the at least one first output layer, the at least one second output of the at least one second output layer, and the combined output of the neural network (ibid-Wehrmann, see pages 4, 5, section 2.3, his final loss function, including a first layer, second layer, and combined layer, as discussed in claim 1, as similarly combined and motivated); and
[coordinating displaying the customized GUI using the total loss] (his output, final prediction result based on the total prediction results).
Costello, with Wehrmann, lack explicitly teaching that which Bhagavath teaches, calculating a total loss, and coordinating displaying the customized GUI using the total loss (ibid-see claim 1, display discussion based on the output of the neural network, paragraph [0015]-his multiple types of loss functions, in order to display the customized GUI).
Thus, it would have been obvious to one of ordinary skill in the linguistics art, before the effective filing date of the invention, as all the claimed elements were known in the prior art and one skilled in the art could have combined the elements as claimed by known methods (computer implemented techniques and algorithms combining processes and steps in natural language processing), in view of the teachings of Costello and Bhagavath to combine the prior art element of a hierarchical intent classifier and determined intent using a neural network (wherein Costello does not explicitly state any of loss functions in his regression analysis, prediction network, or any of his multiple types of neural networks for classification) as taught by Costello with using the generated intent, utilizing a loss function for each prediction analysis, from a neural network to configure a display interface and present results to a user as taught by Bhagavath as each element performs the same function as it does separately, as the combination would yield predictable results, KSR International Co. v. Teleflex Inc., 550 US. -- 82 USPQ2nd 1385 (2007), wherein the predictable result would be to allow a chatbot to generate a customized display to respond to different determined intents, based on a calculated prediction, as needed (ibid-Bhagavath, paragraph [0015, 0023]).
As per claim 11, claim 11 sets forth limitations similar to claim 1 and is thus rejected under similar reasons and rationale, wherein the system is deemed to embody the method, such that Costello with Wehrmann with Anonymous with Bhagavath make obvious a method implemented via execution of computing instructions configured to run at one or more processors and configured to be stored at non-transitory computer-readable media, the method comprising (ibid-see Costello, claim 1, computer-readable media discussion, his abstract-method discussion):
receiving one or more user utterances (ibid-see claim 1, corresponding and similar limitation); using a trained NLP algorithm as one or more layers in a neural network to generate from an input comprising the one or more user utterances at least one first output of at least one first output layer of the neural network, wherein the trained NLP algorithm further generates an intermediate output from the input that is passed into at least one of a linear or hidden layer of the neural network, and wherein an output of the at least one of the linear or hidden layer is combined with the at least first output to generate at least one second output of at least one second output layer of the neural network (ibid); combining the at least one first output of the at least one first output layer of the neural network and the at least one second output of the at least one second output layer of the neural network to create a combined output of the neural network (ibid), wherein: the at least one first output layer of the neural network is different than the at least one second output layer of the neural network (ibid); and coordinating displaying a customized graphical user interface (GUI) using a final version of the combined output of the neural network (ibid).
As per claim 20, claim 1 sets forth limitations similar to claim 1 and is thus rejected under similar reasons and rationale, wherein the non-transitory computer-readable medium storing instructions is deemed to embody the method, such that Costello with Wehrmann with Anonymous with Bhagavath makes obvious teaches a non-transitory computer-readable medium storing instructions, wherein the instructions, upon execution by a processor, cause the processor to perform operations comprising (ibid-see claims 1 and 11, computer-readable medium and instructions discussion): receiving one or more user utterances (ibid-see claim 1, corresponding and similar limitation); using a trained NLP algorithm as one or more layers in a neural network to generate from an input comprising the one or more user utterances at least one first output of at least one first output layer of the neural network, wherein the trained NLP algorithm further generates an intermediate output from the input that is passed into at least one of a linear or hidden layer of the neural network, and wherein an output of the at least one of the linear or hidden layer is combined with the at least first output to generate at least one second output of at least one second output layer of the neural network (ibid); combining the at least one first output of the at least one first output layer of the neural network and the at least one second output of the at least one second output layer of the neural network to create a combined output of the neural network (ibid), wherein: the at least one first output layer of the neural network is different than the at least one second output layer of the neural network (ibid); and coordinating displaying a customized graphical user interface (GUI) using a version of the combined output of the neural network (ibid).
Claim(s) 5 and 15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Costello (US 202/0065384) in view of Wehrmann, in view of Anonymous, in view of Bhagavath et al. (Bhagavath, US 2023/0086302), as applied to claim 1, and further in view of Chen et al. (Chen, US 2022/0129556).
As per claims 5 and 15, Costello with Wehrmann with Anonymous with Bhagavath make obvious the system of claim 1, Wehrmann further teaches wherein: the at least one first output of the at least one first output layer comprises a first cross-entropy loss (ibid-see claim 1, Wehrmann output layer discussion, pages 2-5, see also section 2.3-his cross entropy loss for the at least one first output, and second output);
the at least one second output of the at least one second output layer comprises a second cross-entropy loss (ibid); and
the combined output of the neural network comprises mean squared error (ibid-see section 2.3, formulas (13)-(17), and final loss function, which include the combined output which comprises mean squared error).
Thus, it would have been obvious to one of ordinary skill in the linguistics art, before the effective filing date of the invention, as all the claimed elements were known in the prior art and one skilled in the art could have combined the elements as claimed by known methods (computer implemented techniques and algorithms combining processes and steps in natural language processing), in view of the teachings of Costello and Bhagavath to combine the prior art element of a hierarchical intent classifier and determined intent using a neural network as taught by Costello with using a well-known loss function, including mean squared error, to determine prediction confidence, of the local path, global pathway and overall classification as taught by Wehrmann with using a well-known loss function to determine prediction quality of the intent as classified as taught by Bhagavath as each element performs the same function as it does separately, as the combination would yield predictable results, KSR International Co. v. Teleflex Inc., 550 US. -- 82 USPQ2nd 1385 (2007), wherein the predictable result would be allow a chatbot to generate a customized display to respond to different determined intents, using loss function for prediction accuracy analysis, as needed (ibid, Wehrmann, Bhagavath, paragraph [0015, 0023]).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure (See PTO-892).
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 LAMONT M SPOONER whose telephone number is (571)272-7613. The examiner can normally be reached 8:00 AM -5:00 PM.
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, Daniel Washburn can be reached at (571)272-5551. 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 0Patent 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.
/LAMONT M SPOONER/ Primary Examiner, Art Unit 2657
9/25/2026
