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 Amendment
Claims 1, 9, 17, and 19 are amended. Claims 21-24 are cancelled. As such, claims 1-2, 6-10, and 14-20 are presented for examination.
Response to Arguments
Rejection under 35 U.S.C. 101
Applicant’s arguments have been fully considered and are persuasive. The amended independent claims recite generating a first input sequence based on an input sentence, generating a second input sequence using predefined categories of an external dictionary and based on the tokens in the first input sequence, performing an embedding on the first and second input sequences to generate embedding vectors, concatenating the first and second embedding vectors to generate a concatenated embedding vector, encoding the concatenated embedding vector, encoding the second input sequence, merging the encoded vectors, performing slot tagging based on the merged vectors, and outputting a control signal to control a device based on the slot tagging. Therefore, the claim integrates the judicial exception (mental process) into a practical application by controlling a device based on the slot tagging.
Rejection under 35 U.S.C. 103
Applicant's arguments regarding the Dubey reference have been fully considered but they are not persuasive. Applicant argues “Referring to paragraphs 0078 and 0168 of Dubey, it is described that specific words (e.g., "Windows") may be searched in a dictionary 606. In contrast, according to the claimed invention, a specific value is discarded, and only category information is used.” However, Dubey teaches generating a “tag sentence”, which is a sequence of classes (categories) that represent an input sentence, as described in paragraphs [0136] and [0150] of Dubey. Further, the mapped classes in the tag sentence do not represent specific values. For example, paragraph [0167] of Dubey describes how the word “password” can be mapped to an “entity” class or how the word “slow” can be mapped to a “qualifier” class.
Applicant's arguments regarding the Liu reference have been fully considered but they are not persuasive. Applicant argues, “there is no teaching or suggestion in the proposed combination of utilizing first and second encoders ‘of different types’ as claimed.” However, the Liu reference teaches using a BiLSTM encoder in paragraph [0119], while the Henderson reference teaches using a feed-forward network as part of an encoder in [col 14, lines 10-23].
Applicant’s remaining 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.
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 1-2, 8-10, and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Steedman Henderson et al. (US 11132988 B1; hereinafter referred to as Henderson) in view of Dubey et al. (US 20180173698 A1; hereinafter referred to as Dubey), Wang et al. (US 20220303288 A1; hereinafter referred to as Wang), and Vu et al. (US 20220229993 A1; hereinafter referred to as Vu).
Regarding claim 1, Henderson discloses: a method for training a slot tagging model by a computing device including at least a processor and a memory ([col 7, lines 49-52] The processor 105 is coupled to the storage 107 and accesses the working memory 111. The processor 105 may comprise logic circuitry that responds to and processes the instructions in code stored in the working memory 111), the method comprising: generating, by the processor, a first input sequence based on an input sentence… ([col 2, lines 21-25] receiving input data relating to a speech or text signal originating from a user; representing the input data as a first sequence of first representations, each representing a unit of the input data);
performing, by the processor, a second encoding on the second embedding vector ([col 14, lines 20-24, 34-36] Each embedding output from the previous layer is taken as input to the input FFN 404 and the template FFN 404 in turn, and a sequence of reduced dimension embeddings is output from each of the input FFN 404 and the template FFN 406… The template FFN 406 outputs a second sequence of second vector representations, each vector representing one of the units of the input data. The reduced dimension embedding can be a second encoding.);
merging, by the processor, a first result of the first encoding and a second result of the second encoding using a second encoder… ([col 19, lines 20-31] the input and template sequences are combined using the repeated attention layer 414. However, various other options for combining the sequence pairs may be used. For example, only a single attention layer 414 or more than two attention layers 414 may be used. One or more multi-headed attention layers may additionally or alternatively be used. Alternatively, the two sequences may be combined by a linear or non-linear mapping using an FFN layer taking the concatenated representations as input and mapping to another joint representation. The second encoder can be a feed-forward network.);
performing, by the processor, slot tagging on the input sentence based on a result of the merging… ([col 13, lines 7-17] A third model 409 is run for each slot, checking if a value for this slot has been mentioned in the current utterance. For example, the input utterance “2 people for 6 pm please” comprises a value for the slot “party size” (two) and a value for the slot “time” (6 pm). A set of slot-specific third models 409 all share the parameters of the encoder (the first model 400). This way, extraction of more than one (slot, value) pair from each utterance is enabled, with a reduced memory footprint. A shared set of first model 400 parameters, which constitutes the majority of parameters, are used for all slot-specific labellers, i.e. for all third models 409);
and generating, by the processor, a control signal ([col 42, lines 62-67] As described in relation to the example above, rules are applied in S204 which determine the system response to the input user utterance. Other approaches, such as data driven approaches, may be used however. A text signal corresponding to the selected dialogue act or acts is then generated in S205) to control an operation ([col 40, lines 28-34] a dialogue management step is performed in S204 based on the determined slot value. In this step, a system act is determined. Various example dialogue management methods which may be used in the method of FIG. 2 will now be described. The dialogue manager chooses an appropriate system response following the latest user utterance) of an electronic device corresponding to the result of the slot tagging… ([col 8, lines 1-5] The output module 103 provides the response generated by the processor 105 to an output such as a speaker or screen, or a transmitter for transmitting data to an external storage medium or a network for example).
Henderson does not explicitly, but Dubey discloses: generating, by the processor, a second input sequence ([0136] the phrase-filtering module 628 can be configured to additionally or alternatively determine characteristic pattern(s) associated with respective word(s) or phrases(s) of the free-form user text. For example, the characteristic pattern for a text segment can include a tag sentence, as described below, a sequence of part-of-speech tags, or other forms described herein) by, for each of a plurality of tokens of the first input sequence, determining whether the token corresponds to a predefined category in an external dictionary ([0160] The dictionary 606, e.g., implemented using a trie or hash map, can provide class(es) of the model, e.g., ontology 604. The identified occurrences of the words or phrases can be tagged, e.g., in the free-form text or in sentences extracted therefrom, according to the provided class(es)), and generating an element in the second input sequence that represents the category without including a specific value associated with the token… ([0150] the mapping module 632 can determine a collection of class(es) of the model associated with individual word(s) or phrase(s) of the free-form user text based at least in part on the association determined by the classification module 630… the term “tag sentence” refers to a sequence of tags corresponding to some or all of the free-form user text, arranged in the same order in which the corresponding words or phrases are presented in the user text).
Henderson and Dubey are considered analogous in the field of text processing. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Henderson to combine the teachings of Dubey because doing so would allow for the use of an external dictionary to track and determine categories of input text for more accurate text analysis and speech recognition (Dubey [0004] The knowledge base can include a dictionary, which can be associated with a plurality of entries. The plurality of entries can respectively correspond with individual classes of a plurality of classes of a model, e.g., an ontology. A first text segment, e.g., a word or a phrase, that is not found in a dictionary can be received. A set of features associated with the first text segment can be determined and can be provided to a classifier).
Henderson in view of Dubey does not explicitly, but Wang teaches: performing, by the processor, embedding on the first input sequence to generate a first embedding vector ([0077] a part of the input data 501 corresponding to the embedded feature is provided to a word embedding module 505. The word embedding module 505 is configured to perform word embedding to produce a fixed dimensionality numerical vector representation… The vectorized embedded feature data is provided to a concatenation module 509) and on the second input sequence to generate a second embedding vector… ([0074, 0078] The set of possible words that can occur in domain names is an example of a categorical feature that would require embedding… the part of the input data corresponding to the categorical feature is converted into a numerical vector via one-hot encoding module 507. Embedding can also be performed on categorical features depending on dataset size.);
performing, by the processor, a first encoding ([0080] The concatenation module 509 combines numerical vectors corresponding to all the features to form concatenated data. The concatenated data is provided to an auto-encoder module 511, where the auto-encoder module 511 encodes and decodes the concatenated data to reconstruct the input data) on a sequence of concatenated embedding vectors… ([0014] the anomaly detector decomposes each sequence of inputted sequential data into multiple fields based on multiple features of the sequential data. Further, data corresponding to each feature comprised in each field is vectorized and concatenated, and provided to the auto-encoder).
Henderson, Dubey, and Wang are considered analogous in the field of text processing. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Henderson and Dubey to combine the teachings of Wang because doing so would provide a better framework for analyzing an input sequence and a corresponding categorical sequence using concatenated data, improving loss determination (Wang [0080] The concatenated data is provided to an auto-encoder module 511, where the auto-encoder module 511 encodes and decodes the concatenated data to reconstruct the input data. The reconstructed input data comprises reconstruction loss, where the reconstruction loss is a difference between original input data and the reconstructed input data. The reconstructed input data comprises the plurality of features i.e., the embedded features, the categorical features, and the numerical features. In order to analyze data corresponding to each feature, the reconstructed input data is provided to the split module 515, where the split module 515 splits the reconstructed input data into a plurality of parts based on the plurality of features).
The combination of Henderson, Dubey, and Wang does not explicitly, but Vu teaches: obtained by concatenating the first embedding vector and the second embedding vector at each corresponding token position using a first encoder… ([0135] The encoded form of the utterance is generated using a sequence processing model, such as the CNN/BiLSTM model 4700 of FIG. 4B. In some examples, the concatenated and/or interpolated set of vectors and/or the first set of feature vectors is input into the CNN/BiLSTM model 4700. Based on the inputted vectors, the CNN of the CNN/BiLSTM model 4700 generates one or more character-level vector representations for each character of each word of the utterance. A BiLSTM model inherently uses token positions by processing sequences sequentially.);
wherein the first encoder and the second encoder are of different types ([0135] the concatenated and/or interpolated set of vectors and/or the first set of feature vectors is input into the CNN/BiLSTM model 4700. The first encoder is a BiLSTM model (as taught in Vu) and the second encoder is a feed-forward network (as taught in Henderson).).
Henderson, Dubey, Wang, and Vu are considered analogous in the field of text processing. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Henderson, Dubey, and Wang to combine the teachings of Vu because doing so would allow for the use of sequential encoders, such as a LSTM model, to help identify entities in a user input, improving slot tagging and user intent determination (Vu [0140] features of the present disclosure determine how contextually relevant one or more detected entities within a group of entities may be to a system query and/or a user's utterance(s), features of the present disclosure are able to correctly identify the intended referent for a particular named entity and improve the user's interaction with the digital assistant and/or chatbot system).
Regarding claim 2, the combination of Henderson, Dubey, Wang, and Vu teaches: the method of claim 1. Henderson further teaches: wherein the generating of the first input sequence comprises dividing the input sentence in units of tokens to generate the first input sequence… ([col 1, lines 59-61] FIG. 5(c) shows examples of tokenisation of input text performed by the tokenisation algorithm used in the method of FIG. 5(a)).
Dubey further teaches: and the generating of the second input sequence comprises generating the second input sequence based on whether each of a plurality of tokens included in the first input sequence matches the dictionary information included in the external dictionary ([0167] analysis module 634 can be configured to perform post-processing to determine whether a tag sentence is associated with one or more fundamental areas, such as reliability, security, usability, etc. In at least one example, classes or subclasses of the model, e.g., ontology 604, can respectively correspond to entries in the dictionary 606).
Regarding claim 8, the combination of Henderson, Dubey, Wang, and Vu teaches: the method of claim 1. Henderson further teaches: calculating a loss value for a result of performing the slot tagging ([col 24, lines 1-4] The loss is the negative log-likelihood, which is equal to the negative sum of the transition scores and unary potentials that correspond to the true tag labels, up to a normalization term), and adjusting weights of the slot tagging model based on the calculated loss value ([col 39-40, lines 55-4] The attention layers comprise scaled dot-product attention units, such that attention weights are calculated between every token simultaneously).
Regarding claim 9, Henderson teaches: a non-transitory computer-readable medium storing a program for implementing a method for training a slot tagging model ([col 5, lines 29-32] there is provided a non-transitory computer readable storage medium comprising program instructions stored thereon that are executable by a computer processor to perform). The rest of the claim recites similar limitations as claim 1 and therefore is rejected similarly.
Regarding claim 10, it recites similar limitations as claim 2 and therefore is rejected similarly.
Regarding claim 16, it recites similar limitations as claim 8 and therefore is rejected similarly.
Claims 6-7 and 14-15 are rejected under 35 U.S.C. 103 as being unpatentable over Henderson in view of Dubey, Wang, and Vu, as applied to claims 1-2, 8-10, and 16 above, and further in view of Liu et al. (US 20210216862 A1; hereinafter referred to as Liu).
Regarding claim 6, the combination of Henderson, Dubey, Wang, and Vu teaches: the method of claim 1. The combination of Henderson, Dubey, Wang, and Vu does not explicitly, but Liu teaches: wherein the merging comprises obtaining a third context vector by merging a first context vector obtained by the first encoding and a second context vector obtained by the second encoding ([0010] when the context includes the context text, embed sequential words of the context text into context text embeddings, perform a second neural network on the context text embeddings to obtain context feature vectors, encode the context meta info into context embedding vectors, and concatenate the context feature vectors and the context embedding vectors into concatenated context vectors).
Henderson, Dubey, Wang, Vu, and Liu are considered analogous in the field of text processing. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Henderson, Dubey, Wang, and Vu to combine the teachings of Liu because analyzing context associated with the text input leads to more accurate slot tagging (Liu [0204] attributes of the object are derived from task on the object, context related to the object, and external vocabulary, thus expanding range of attributes and improving accuracy of prediction).
Regarding claim 7, the combination of Henderson, Dubey, Wang, and Liu teaches: the method of claim 6. Henderson further teaches: wherein the merging comprises merging the first context vector and the second context vector using an addition method or an attention mechanism ([col 7, lines 9- 13] the combining layer may be an attention layer. The value extractor mode! predicts which tokens in the input sentence constitute the key phrase. At inference, the attention layer effectively acts like an additional self-attention layer).
Regarding claim 14, it recites similar limitations as claim 6 and therefore is rejected similarly.
Regarding claim 15, it recites similar limitations as claim 7 and therefore is rejected similarly.
Claims 17-20 are rejected under 35 U.S.C. 103 as being unpatentable over Henderson in view of Wang, Dubey, and Liu.
Regarding claim 17, Henderson teaches: a speech recognition apparatus, comprising: a communication module configured to receive a voice command of a user ([col 7-8, lines 65-1] The input module 101 receives a query through an input, which may be a receiver for receiving data from an external storage medium or a network, a microphone);
a language processing module configured to process the received voice command to classify an intent corresponding to the received voice command ([col 41, lines 58-63] Thus the first group of rules relate to blocks 1 and 2 in FIG. 11. If the proposed entity is not set (i.e. still do not know which restaurant is the intent of the user), the rule “Ask for restaurant name” will fire and one outcome will be the system act requesting the actual restaurant name (Block 1)) and perform slot tagging on the voice command ([col 2, lines 40-43] model is a trained model. In an example, the first tag is a slot value tag, and wherein the set of tags comprises a before tag, tagging a part before the slot value, and an after tag, tagging a part after the slot value);
and a control module configured to generate a signal, required to provide a function intended by the user, based on an output of the language processing module ([col 7-8, lines 65-5] The input module 101 receives a query through an input, which may be a receiver for receiving data from an external storage medium or a network, a microphone, screen or a keyboard for example. The output module 103 provides the response generated by the processor 105 to an output such as a speaker or screen, or a transmitter for transmitting data to an external storage medium or a network for example), wherein a slot tagging model used to perform the slot tagging in the language processing module comprises… ([col 2, lines 63-67] the first tag is a slot value tag, wherein the model is a second model corresponding to a first slot and wherein the first sequence of first representations and second sequence of second representations are generated using a first model);
a second encoding layer configured to perform a second encoding on the second embedding vector using a second encoder… ([col 14, lines 10-23] embeddings are taken as input to an input feed forward network 404 and also as input to a template feed forward network 406 in the third model 409. These project the embeddings down to lower dimensional embeddings, which will be referred to as d-dimensional embeddings. For example, where the output embeddings are 512-dimensional output representations, these may be projected down to a first set of 128-dimensional representations using the input feed-forward network (FFN) and projected down to a second set of 128-dimensional representations using the template feed-forward network (FFN). Each embedding output from the previous layer is taken as input to the input FFN 404 and the template FFN 404 in turn, and a sequence of reduced dimension embeddings is output from each of the input FFN. Multiple embeddings using a FFN are performed, which includes encoding.);
and an output layer configured to output a slot tagging result for the third context vector… ([col 25, lines 1-4] the value extractor model 109 learns an implicit universal space of slots and values, where slots are represented as the contexts in which a value might occur).
Henderson does not explicitly, but Wang discloses: an embedding layer configured to obtain a first embedding vector by embedding a first input sequence generated based on an input sentence ([0077] a part of the input data 501 corresponding to the embedded feature is provided to a word embedding module 505. The word embedding module 505 is configured to perform word embedding to produce a fixed dimensionality numerical vector representation… The vectorized embedded feature data is provided to a concatenation module 509), and a second embedding vector by embedding a second input sequence… ([0074, 0078] The set of possible words that can occur in domain names is an example of a categorical feature that would require embedding… the part of the input data corresponding to the categorical feature is converted into a numerical vector via one-hot encoding module 507. Embedding can also be performed on categorical features depending on dataset size.)
a first encoding layer implemented by the processor and configured to perform a first encoding on a sequence of concatenated embedding vectors… ([0080] The concatenation module 509 combines numerical vectors corresponding to all the features to form concatenated data. The concatenated data is provided to an auto-encoder module 511, where the auto-encoder module 511 encodes and decodes the concatenated data to reconstruct the input data).
Henderson and Wang are considered analogous in the field of text processing. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Henderson to combine the teachings of Wang because doing so would provide a better framework for analyzing an input sequence and a corresponding categorical sequence using concatenated data, improving loss determination (Wang [0080] The concatenated data is provided to an auto-encoder module 511, where the auto-encoder module 511 encodes and decodes the concatenated data to reconstruct the input data. The reconstructed input data comprises reconstruction loss, where the reconstruction loss is a difference between original input data and the reconstructed input data. The reconstructed input data comprises the plurality of features i.e., the embedded features, the categorical features, and the numerical features. In order to analyze data corresponding to each feature, the reconstructed input data is provided to the split module 515, where the split module 515 splits the reconstructed input data into a plurality of parts based on the plurality of features).
Henderson in view of Wang does not explicitly, but Dubey teaches: wherein the second input sequence is generated based on an external dictionary and represents a predefined category corresponding to a token of the first input sequence without including a specific value associated with the token… ([0150] the mapping module 632 can determine a collection of class(es) of the model associated with individual word(s) or phrase(s) of the free-form user text based at least in part on the association determined by the classification module 630… the term “tag sentence” refers to a sequence of tags corresponding to some or all of the free-form user text, arranged in the same order in which the corresponding words or phrases are presented in the user text).
Henderson, Wang, and Dubey are considered analogous in the field of text processing. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Henderson and Wang to combine the teachings of Dubey because doing so would allow for the use of an external dictionary to track and determine categories of input text for more accurate text analysis (Dubey [0004] The knowledge base can include a dictionary, which can be associated with a plurality of entries. The plurality of entries can respectively correspond with individual classes of a plurality of classes of a model, e.g., an ontology. A first text segment, e.g., a word or a phrase, that is not found in a dictionary can be received. A set of features associated with the first text segment can be determined and can be provided to a classifier).
The combination of Henderson, Wang, and Dubey does not explicitly, but Liu teaches: obtained by concatenating the first embedding vector and the second embedding vector at each corresponding token position using a first encoder… ([0119] The BiLSTM module 123 is configured to, upon receiving the sequential embedding vectors {e.sub.1, e.sub.2, . . . , e.sub.i, . . . , e.sub.T}, feed the embedding vectors to the BiLSTM, perform BiLSTM on the embedding vectors to obtain task feature vectors {u.sub.1, u.sub.2, . . . , u.sub.i, . . . , u.sub.T}, and send the task feature vectors to the task concatenating module 125. A BiLSTM model inherently uses token positions by processing sequences sequentially.);
wherein the first encoder and the second encoder are of different types ([0121] The task concatenating module 125 is configured to, upon receiving the task feature vectors {u.sub.1, u.sub.2, . . . , u.sub.i, . . . , u.sub.T} (encoding of the task description) from the BiLSTM module 123 and the task embedding vectors from the task meta info module 124, concatenate the task feature vectors and the task embedding vectors to obtain concatenated task vectors… The first encoder is a BiLSTM model (as taught in Liu) and the second encoder is a feed-forward network (as taught in Henderson).).
Henderson, Wang, Dubey, and Liu are considered analogous in the field of language processing. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Henderson, Wang, and Dubey to combine the teachings of Liu because analyzing context associated with the text input leads to more accurate slot tagging and using LSTM models for sequential processing incorporates token positions for better contextual understanding (Liu [0204] attributes of the object are derived from task on the object, context related to the object, and external vocabulary, thus expanding range of attributes and improving accuracy of prediction).
Regarding claim 18, the combination of Henderson, Wang, Dubey, and Liu teaches: the speech recognition apparatus of claim 17. Dubey further teaches: a memory configured to store the external dictionary ([0048] Computer-readable media 118 can store, for example, executable instructions of an operating system 122, an inference engine 124, a training engine 126, and other modules, programs, or applications that are loadable and executable by processing units 116. Computer-readable media can also store, for example, a knowledge base 128), wherein the external dictionary stored in the memory is configured to be updated with new data added ([0141] the classification module 630 can be configured to update the dictionary 606, e.g., to associate, in the dictionary 606, the particular text segment with a particular class of the model based at least in part on the one or more attribute(s) of the particular text segment and one or more of the attribute(s) of individual one(s) of the words or phrases in the dictionary 606).
Regarding claim 19, it recites similar limitations as claim 17 and therefore is rejected similarly.
Regarding claim 20, it recites similar limitations as claim 18 and therefore is rejected similarly.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Nathan Tengbumroong whose telephone number is (703)756-1725. The examiner can normally be reached Monday - Friday, 11:30 am - 8:00 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, Hai Phan can be reached at 571-272-6338. 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.
/NATHAN TENGBUMROONG/Examiner, Art Unit 2654
/HAI PHAN/Supervisory Patent Examiner, Art Unit 2654
Prosecution Insights