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 .
Continued Examination Under 37 CFR 1.114
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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114.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 01/22/2026 has been entered.
Status of the claims
Claims 1-20 were pending, claims 1 and 13 have been amended, claims 6, 15 and 16 have been canceled. Therefore, claims 1-5, 7-14, 17-20 are currently pending for examination.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 02/04/2026 is being considered by the examiner.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-5, 7-14, 17-20 rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1
The claims recite a computer-implemented method and a computer system comprising processors (claims 1 and 13). These claims fall within at least one of the four categories of patentable subject matter.
Step 2A Prong One
The claim recites obtaining,… a query; processing,.., the query with a machine-learned resource retrieval model to generate a model prediction from the machine-learned resource retrieval model , wherein the model prediction directly predicts one or more resources that are predicted to be responsive to the query from a resource corpus containing a plurality of resources, wherein a plurality of resource identifiers are respectively associated with the plurality of resources, and Wherein the machine-learned resource retrieval model maps the query directly to the resource identifiers, wherein the model prediction comprises the resource identifiers for the one or more resources that are predicted to be responsive to the query , Wherein the resource identifiers comprise structured semantic identifiers, wherein the structured semantic identifiers are generated via iterative clustering of a plurality of embeddings respectively associated with the plurality of resources, wherein an embedding for a resource can be a representation of the resource in a lower-dimensional space, and wherein the resource identifiers are generated to reduce a search space after each decoding instance, providing, …, the model prediction as an output.
The limitation of obtaining…processing the query with a machine-learned resource retrieval model and providing the model prediction as an output, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitations fall within the mental process groupings of abstract ideas because they cover concepts performed in the human mind, including observation, evaluation, judgment, and opinion but for the recitation of generic computer components. That is, other than reciting “by a processor” , generating a model prediction involves the techniques rely upon abstract ideas. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation referring to the human mind and therefore recites a judicial exception, namely an abstract idea.
Claim 13 recites limitations that correspond to those recited in claims 1, 7, 9.
Claims 2-5, 7-12, 14, 17-20 recite limitations that are further extensions of the processing the query.
Step 2A Prong Two
This judicial exception is not integrated into a practical application. In particular, the claim
only recites one additional element – using a processor to perform obtaining, processing and providing
steps. The processor in those steps is recited at a high-level of generality (i.e., as a generic processor performing a generic computer processing the query and providing the output) such that it amounts no more than mere instructions to apply the exception using a generic computer component. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea on a computer.
The dependent claims do not recite limitations which would integrate the judicial exception into a practical application from the independent claim. Therefore, the claims do not integrate the judicial exception into a practical application.
Step 2B
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 processor to perform obtaining, processing and providing steps 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. The claim is not patent eligible.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 5, 7, 9 and 10 are rejected under 35 U.S.C. 103 as being unpatentable over Filgueiras et al. (US 20190005069, hereafter Filgueiras) in view of Nicola De Cao et al. “Autoregressive Entity Retrieval”, Published as a conference paper at ICLR 2021, (hereafter Cao) and further in view of Pedersen (US 20200142999).
Regarding claim 1, Filgueiras discloses: A computer-implemented method to perform resource retrieval with improved computational efficiency, the method comprising:
obtaining, by a computing system comprising one or more computing devices, a query (Filgueiras [0021] discloses: One or more computing devices associated with an image retrieval application can provide a query image as input to the trained image descriptor model);
processing, by the computing system, the query with a machine-learned resource retrieval model to generate a model prediction from the machine-learned resource retrieval model (Filgueiras [0038] discloses: In response to receipt of each training image in such first portion of the training dataset, the image descriptor model generates an output. This output of the image descriptor model predicts the remainder of the set of ground-truth data (e.g., the second portion of data associated with each training image),
wherein the model prediction directly predicts one or more resources that are predicted to be responsive to the query from a resource corpus containing a plurality of resources (Filgueiras [0082] discloses: In response to receipt of each training image in such first portion 410 of the training data 402, the image descriptor model 400 generates an output 414. This output 414 of the image descriptor model 400 predicts the remainder of the set of ground-truth data (e.g., the second portion 412 of training data 402 associated with each training image in the first portion 410 of training data 402). After such prediction, the training computing system can apply or otherwise determine a loss function 416 that compares the output 414 of the image descriptor model 400 to the remainder of the ground-truth data (e.g., the second portion 412 of training data 402) which the image descriptor model 400 attempted to predict),
wherein a plurality of resource identifiers are respectively associated with the plurality of resources (Filqueiras [0031] discloses: generate an index including descriptors associated with a plurality of database images. The plurality of database images can be provided as input to the image descriptor model, and corresponding outputs received from the image descriptor model in response to the plurality of database images), and
Filgueiras didn’t disclose, but Cao discloses: Wherein the machine-learned resource retrieval model maps the query directly to the resource identifiers (Cao [section 3.2 “Autoregressive End-to-End Entity linking”] discloses: where, given a document, a system has to both detect entity mentions and link those mentions to their respective KB entities. In this setting, we train the model (as resource retrieval model) to predict the source input again but with annotated spans. We use a Markup annotation where spans boundaries are flagged with special tokens and accompanied by their corresponding entity identifiers (as resource identifiers). At each generation step, the decoder is either generating a mention span, generating a link to a mention, or continuing from the input source. When outside a mention/entity step the decoder has only two options: (i) to continue by copying the next token from the input source, or (ii) to generate the start of mention token (i.e., ‘[’) which makes the decoder enter the mention generating phase. While generating a mention, the decoder has either to continue with the next token in the input source or to generate the end of mention token (i.e., ‘]’) which makes the decoder enter the entity generating phase. Finally, when generating an entity, the decoder employs the entities trie such that it can only output a valid entity identifier (resource identifiers) as in Constrained Beam Search).
Filgueiras and Cao are analogous art because they are in the same field of endeavor, information retrieval. It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Filgueiras, to include the teaching of Cao, in order to generate entity names autoregressively. The suggestion to combine is to provide the autoregressive formulation allows us to directly capture some of these properties, leading to several advantages with respect to current solutions, including an efficient way to cross encode mention context and entity candidates, a much smaller memory footprint, and the ability to compute an exact softmax without the need to subsample negative data.
Filgueiras didn’t disclose, but Pedersen discloses: wherein the model prediction comprises the resource identifiers for the one or more resources that are predicted to be responsive to the query (Pederson [0025;0028] discloses: search queries; features that are used by the trained machine learning model(s) 136 to predict clusters of associated text 138. In this manner, given a relatively large space of words and positions where the words can be placed, the trained machine learning model(s) 136 is trained to make accurate predictions as to text (e.g., words) that can be grouped together into the clusters 138
Wherein the resource identifiers comprise structured semantic identifiers, wherein the structured semantic identifiers are generated via iterative clustering of a plurality of embeddings respectively associated with the plurality of resources, wherein an embedding for a resource can be a representation of the resource in a lower-dimensional space, and wherein the resource identifiers are generated to reduce a search space after each decoding instance (Pedersen [0026] discloses: Word embedding generally involves a mathematical embedding from a space with one dimension per word to a continuous vector space with a much lower dimension; [0028] discloses: the trained machine learning model(s) 136 is trained to make accurate predictions as to text (e.g., words) that can be grouped together into the clusters 138, the text classification framework extremely flexible in that the word embedding component 122 is capable of “catching” creative variants of a particular type of speech that would not otherwise be detected by a system that clusters text based on a semantic understanding of that text (e.g., a system that groups synonyms together); [0036] discloses: train a machine learning model(s) 148 for classifying text, reduction technique well-suited for embedding high-dimensional data for visualization in a low-dimensional space of two or three dimensions).
providing, by the computing system, the model prediction as an output (Pederson [0030; 0047] discloses: A machine learning model(s), once trained, is a learned mechanism that can receive new data as input and estimate or predict a result as output).
Filgueiras as modified and Pedersen are analogous art because they are in the same field of endeavor, systems for classifying and moderating information using a machine learning. It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Filgueiras, to include the teaching of Pedersen, in order for classifying and moderating text using a machine learning. The suggestion to combine is for classifying and moderating text using a machine learning approach that is based on a word embedding process.
Regarding claim 5, Filgueiras as modified discloses: The computer-implemented method of claim 1, wherein the respective resource identifier associated with each resource in the plurality of resources comprises a structured semantic identifier (Filgueiras [0026] discloses: an image descriptor model trained to generate a set of key point descriptors can include layers that are trained to extract local feature descriptors from an image, determine attention scores for the local feature descriptors, and ultimately determine a subset of the local feature descriptors having a highest score for inclusion in a set of key point descriptors. By determining a subset of semantically useful features, as opposed to all features within a dense grid, those local features that are less likely relevant to the given image application (e.g., image retrieval) can be removed. By removing the local features that are more likely to add clutter and/or distract an image application, key point selection can thus advantageously increase accuracy and computational efficiency for a variety of image applications).
Regarding claim 7, Filgueiras as modified discloses: The computer-implemented method of claim 1, wherein the machine-learned resource retrieval model has been trained using an indexing loss function, wherein the indexing loss function evaluates an ability of the machine- learned resource retrieval model to output the resource identifier associated with a particular resource when provided with data descriptive of the particular resource as an input (Filgueiras [0038] discloses: In response to receipt of each training image in such first portion of the training dataset, the image descriptor model generates an output. This output of the image descriptor model predicts the remainder of the set of ground-truth data (e.g., the second portion of data associated with each training image). After such prediction, the training computing system can apply or otherwise determine a loss function that compares the output of the image descriptor model to the remainder of the ground-truth data which the image descriptor model attempted to predict).
Regarding claim 9, Filgueiras as modified discloses: The computer-implemented method of claim 1, wherein the machine-learned resource retrieval model has been trained using a retrieval loss function, wherein the retrieval loss function evaluates an ability of the machine- learned resource retrieval model to output the resource identifier associated with a particular resource when provided with a training query for which the particular resource has been labeled as a response (Filgueiras [0038] discloses: In response to receipt of each training image in such first portion of the training dataset, the image descriptor model generates an output. This output of the image descriptor model predicts the remainder of the set of ground-truth data (e.g., the second portion of data associated with each training image). After such prediction, the training computing system can apply or otherwise determine a loss function that compares the output of the image descriptor model to the remainder of the ground-truth data which the image descriptor model attempted to predict).
Regarding claim 10, Filgueiras as modified discloses: The computer-implemented method of claim 1, wherein the machine-learned resource retrieval model has been trained using an indexing loss function and a retrieval loss function in a multi-task training approach (Filgueiras [0038] discloses: In response to receipt of each training image in such first portion of the training dataset, the image descriptor model generates an output. This output of the image descriptor model predicts the remainder of the set of ground-truth data (e.g., the second portion of data associated with each training image). After such prediction, the training computing system can apply or otherwise determine a loss function that compares the output of the image descriptor model to the remainder of the ground-truth data which the image descriptor model attempted to predict).
Claims 2-4, 8, 11-14, 17-20 are rejected under 35 U.S.C. 103 as being unpatentable over Filgueiras et al. (US 20190005069, hereafter Filgueiras) in view of Nicola De Cao et al. “Autoregressive Entity Retrieval”, Published as a conference paper at ICLR 2021, (hereafter Cao) and in view of Pedersen (US 20200142999) and further in view of Sharef et al. “Enhancements to the Sequence-to-Sequence-Based Natural Answer Generation Models”, IEEE Access, publication 03-05-2020 (hereafter Sharef).
Regarding claim 2, Filgueiras as modified didn’t disclose, but Sharef discloses: The computer-implemented method of claim 1, wherein: the machine-learned resource retrieval model comprises a sequence-to-sequence model that receives and processes the query as an input sequence to generate one or more predicted output sequences as the model prediction (Sharef [page 45741] discloses: Algorithm 2 Seq2Seq Model prediction (Beam search ) for generating output to question);and
the one or more predicted output sequences correspond to the resource identifiers of the one or more resources that are predicted to be responsive to the query (Sharef [page 45741] discloses: Algorithm 2 Seq2Seq Model prediction (Beam search ) for generating output to question and predicted for each of the tokens).
Filgueiras as modified and Sharef are analogous art because they are in the same field of endeavor, information retrieval. It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Filgueiras, to include the teaching of Sharef, in order to provide the answer to the query. The suggestion to combine is to generate an accurate answer in natural language for the question.
Regarding claim 3, Filgueiras as modified discloses: The computer-implemented method of claim 1, wherein the respective resource identifier associated with each resource in the plurality of resources comprises an unstructured atomic identifier (Sharef [page 45739, section II Seq2Seq model] discloses the list of tokens, token can be a word or a sub-word of one or more characters/number and symbols)).
Regarding claim 4, Filgueiras as modified discloses: The computer-implemented method of claim 1, wherein the respective resource identifier associated with each resource in the plurality of resources comprises an unstructured string identifier (Sharef [page 45739, section II Seq2Seq model] discloses the list of tokens).
Regarding claim 8, Filgueiras as modified discloses: The computer-implemented method of claim 7, wherein the data descriptive of the particular resource comprises: direct indexing tokens extracted from the particular resource; set indexing tokens extracted from the particular resource; or inverted indexing tokens extracted from the particular resource (Sharef [page 45739, section II Seq2Seq model] discloses generate or extract the tokens and get the k top tokens)).
Regarding claim 11, Filgueiras as modified discloses: The computer-implemented method of claim 1, wherein the model prediction from the machine-learned resource retrieval model comprises a softmax output over the plurality of resources (Sharef [page 45741] discloses: Algorithm 2 Seq2Seq Model prediction (Beam search ) for generating output to question).
Regarding claim 12, Filgueiras as modified discloses: The computer-implemented method of claim 1, wherein the model prediction from the machine-learned resource retrieval model comprises one or more beam search results generated by performance of a sequential beam search (Sharef [page 45741] discloses: Algorithm 2 Seq2Seq Model prediction (Beam search ) for generating output to question).
Regarding claim 13, Filgueiras as modified discloses: A computing system for training a model to perform resource retrieval with improved computational efficiency, the computing system comprising:
one or more processors (Filgueiras[0049]); and
one or more non-transitory computer-readable media that collectively store instructions that, when executed by the one or more processors, cause the computing system to perform operations, the operations comprising (Filgueiras[0049]);:
obtaining, by the computing system, a resource corpus comprising a plurality of resources, wherein a respective resource identifier is associated with each of the plurality of resources (Filqueiras [0031] discloses: generate an index including descriptors associated with a plurality of database images. The plurality of database images can be provided as input to the image descriptor model, and corresponding outputs received from the image descriptor model in response to the plurality of database images); and
for each of one or more input resources of the plurality of resources (Filgueiras [0021] discloses: One or more computing devices associated with an image retrieval application can provide a query image as input to the trained image descriptor model):
evaluating an indexing loss function that compares the predicted resource identifier to the actual resource identifier for the input resource (Filgueiras [0038] discloses: In response to receipt of each training image in such first portion of the training dataset, the image descriptor model generates an output. This output of the image descriptor model predicts the remainder of the set of ground-truth data (e.g., the second portion of data associated with each training image). After such prediction, the training computing system can apply or otherwise determine a loss function that compares the output of the image descriptor model to the remainder of the ground-truth data which the image descriptor model attempted to predict).
Filgueiras didn’t disclose, but Cao discloses: processing, by the computing system, data descriptive of the input resource with a resource retrieval model to generate a predicted resource identifier for the input resource via autoregressive generation, wherein the resource retrieval model comprises a transformer model, wherein the transformer model maps from encoding to resource identifiers (Cao [section 3.2 “Autoregressive End-to-End Entity linking”] discloses: where, given a document, a system has to both detect entity mentions and link those mentions to their respective KB entities. In this setting, we train the model (as resource retrieval model) to predict the source input again but with annotated spans. We use a Markup annotation where spans boundaries are flagged with special tokens and accompanied by their corresponding entity identifiers (as resource identifiers). At each generation step, the decoder is either generating a mention span, generating a link to a mention, or continuing from the input source. When outside a mention/entity step the decoder has only two options: (i) to continue by copying the next token from the input source, or (ii) to generate the start of mention token (i.e., ‘[’) which makes the decoder enter the mention generating phase. While generating a mention, the decoder has either to continue with the next token in the input source or to generate the end of mention token (i.e., ‘]’) which makes the decoder enter the entity generating phase. Finally, when generating an entity, the decoder employs the entities trie such that it can only output a valid entity identifier (resource identifiers) as in Constrained Beam Search).
Filgueiras and Cao are analogous art because they are in the same field of endeavor, information retrieval. It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Filgueiras, to include the teaching of Cao, in order to generate entity names autoregressively. The suggestion to combine is to provide the autoregressive formulation allows us to directly capture some of these properties, leading to several advantages with respect to current solutions, including an efficient way to cross encode mention context and entity candidates, a much smaller memory footprint, and the ability to compute an exact softmax without the need to subsample negative data.
Filgueiras as modified didn’t disclose, but Pedersen discloses: Wherein the resource identifiers comprise structured semantic identifiers, wherein the structured semantic identifiers are generated via iterative clustering of a plurality of embeddings respectively associated with the plurality of resources, wherein an embedding for a resource can be a representation of the resource in a lower-dimensional space, and wherein the resource identifiers are generated to reduce a search space after each decoding instance (Pedersen [0026] discloses: Word embedding generally involves a mathematical embedding from a space with one dimension per word to a continuous vector space with a much lower dimension; [0028] discloses: the trained machine learning model(s) 136 is trained to make accurate predictions as to text (e.g., words) that can be grouped together into the clusters 138, the text classification framework extremely flexible in that the word embedding component 122 is capable of “catching” creative variants of a particular type of speech that would not otherwise be detected by a system that clusters text based on a semantic understanding of that text (e.g., a system that groups synonyms together); [0036] discloses: train a machine learning model(s) 148 for classifying text, reduction technique well-suited for embedding high-dimensional data for visualization in a low-dimensional space of two or three dimensions).
Filgueiras and Pedersen are analogous art because they are in the same field of endeavor, systems for classifying and moderating information using a machine learning. It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Filgueiras, to include the teaching of Pedersen, in order for classifying and moderating text using a machine learning. The suggestion to combine is for classifying and moderating text using a machine learning approach that is based on a word embedding process.
Filgueiras as modified didn’t disclose, but Sharef discloses: modifying one or more parameters of the resource retrieval model based on the indexing loss function (Sharef ,see algorithm 1 for generating the output based on calculate the cross entropy loss function and update the model parameters).
Filgueiras and Sharef are analogous art because they are in the same field of endeavor, information retrieval. It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Filgueiras, to include the teaching of Sharef, in order to provide the answer to the query. The suggestion to combine is to generate an accurate answer in natural language for the question.
.
Regarding claim 14, Filgueiras as modified discloses: The computing system of claim 13, wherein the data descriptive of the input resource comprises: direct indexing tokens extracted from the input resource; set indexing tokens extracted from the input resource; or inverted indexing tokens extracted from the input resource (Sharef [page 45741] discloses: algorithm 2, generating/extracting tokens/indexing form the query and generating a list of top tokens).
Regarding claim 17, Filgueiras as modified discloses: The computing system of any of claim 13, wherein the operations further comprise:
training the resource retrieval model using a retrieval loss function, wherein the retrieval loss function evaluates an ability of the resource retrieval model to output the resource identifier associated with a particular resource when provided with a training query for which the particular resource has been labeled as a response (Filgueiras [0084] discloses: the training computing system can train an image descriptor model 400 based on a first training process to learn determination of the one or more local feature descriptors and a second training process to learn determination of the attention scores for each of the one or more local feature descriptors given the determined local feature descriptors. In some implementations, the first training process can include determining a first loss function that can be backpropagated through the image descriptor model 400 to train the feature extraction layers (e.g., selected shared layers 403 and the one or more feature extraction layers 404) in the image descriptor model 400 (e.g., by modifying one or more weights associated with the feature extraction layers within the image descriptor model 400). In some implementations, the second training process can determine a second loss function that can be backpropagated through the image descriptor model 400 to train the attention-based key point selection layers (e.g., selected shared layers 403 and the one or more attention-based key point selection layers 406) in the image descriptor model 400 (e.g., by modifying one or more weights associated with the attention-based key point selection layers within the image descriptor model 400).
Regarding claim 18, Filgueiras as modified discloses: The computing system of claim 17, wherein the operations comprise training the resource retrieval model using the indexing loss function and the retrieval loss function in a multi-task training approach (Filgueiras [0038] discloses: In response to receipt of each training image in such first portion of the training dataset, the image descriptor model generates an output. This output of the image descriptor model predicts the remainder of the set of ground-truth data (e.g., the second portion of data associated with each training image). After such prediction, the training computing system can apply or otherwise determine a loss function that compares the output of the image descriptor model to the remainder of the ground-truth data which the image descriptor model attempted to predict).
Regarding claim 19, Filgueiras as modified discloses: The computing system of any of claim 13, wherein the model prediction from the resource retrieval model comprises a softmax output over the plurality of resources (Sharef [page 45741] discloses: Algorithm 2 Seq2Seq Model prediction (Beam search ) and sort beam search list ).
Regarding claim 20, Filgueiras as modified discloses: The computing system of claim 13, wherein the model prediction from the resource retrieval model comprises one or more beam search results generated by performance of a sequential beam search (Sharef [page 45741] discloses: Algorithm 2 Seq2Seq Model prediction (Beam search ) for generating output to question).
Response to Arguments
Applicant’s arguments have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Contact Information
Any inquiry concerning this communication or earlier communications from the examiner should be directed to CINDY NGUYEN whose telephone number is (571)272-4025. The examiner can normally be reached M-F 8:00-4:30.
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, Bhatia Ajay can be reached at 571-272-3906. 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.
/CINDY NGUYEN/Examiner, Art Unit 2156