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 .
Specification
The disclosure is objected to because of the following informalities:
In paragraph 0054, line 10, “candidate output token 338 "July"” should read “candidate output token 338-4 "July"”.
In paragraph 0068, line 5, “ouptut token 308-1” should read “output token 308-1”.
Appropriate correction is required.
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 and 8 – 9 are rejected under 35 U.S.C. 103 as being unpatentable over Willmore et al. (US Patent No. 9,785,630), hereinafter Willmore, in view of Shazeer et al. (US Patent No. 11,556,786), hereinafter Shazeer.
Regarding claim 1, Willmore discloses a computer-implemented method, comprising:
processing an input token sequence with a language model to generate a plurality of raw scores for a plurality of candidate output tokens, each raw score representing a confidence that a candidate output token associated with the raw score represents a next token in the input token sequence (Column 6, lines 54-64, "Referring again to process 200 of FIG. 2, at block 204, a first probability of a predicted word can be determined based on a previously entered word using a word n-gram model (e.g., word n-gram language model 110 of FIG. 1, and as illustrated by word language model 320 of FIG. 3). As understood by one of ordinary skill in the art, a word n-gram model can compute the probability (or language score) of a current word wq given the available word history or linguistic context c available at a particular time, which can include observed words 324 in FIG. 3 (e.g., a sequence of previously entered words)."; An n-gram language model reads on a language model, the available word history reads on an input token sequence, and computing the probability or language score of a current word given the available word history reads on generating raw scores representing a confidence that a candidate output token associated with the raw score represents a next token in the input token sequence.);
applying a token domain weight to each respective raw score associated with each respective candidate output token of the plurality of candidate output tokens to generate a plurality of weighted raw scores for the plurality of candidate output tokens, wherein each token domain weight applied to each respective raw score is based on a frequency of an output token in a corpus of domain-specific data for a first domain (Column 7, lines 10-18, "Referring again to process 200 of FIG. 2, at block 206, a second probability of the predicted word can be determined based on a prefix character of a new word using a unigram model (e.g., unigram model 112 of FIG. 1, and as illustrated by unigram model 322 of FIG. 3). As understood by one of ordinary skill in the art, a unigram model can compute the probability (or word score) of a current word w in, for example, a particular language, domain, or the like. This word score can be represented by the probability P(w)."; Column 7, lines 29-34, "One of ordinary skill in the art will understand that the unigram model can be trained on relevant data sets (e.g., text relevant to a particular domain), structured in a variety of different ways (e.g., in a traversable trie), and modified for particular applications as desired."; Column 7, lines 35-50, "Referring again to process 200 of FIG. 2, at block 208, a combined probability of the predicted word can be determined based on the first probability from the word n-gram model at block 204 and the second probability from the unigram model at block 206. In one example, the product of the first and second probabilities for the same predicted word can be determined, such as by calculating P(w) · P(wq|c), where w and wq can be the same predicted word used in the unigram model and n-gram model, respectively. In another example, one or both of the first and second probabilities can be weighted or scored by a factor, modified by a logarithm, raised to a power (e.g., an empirically-derived value), or mathematically altered in other ways based, for example, on empirical testing to provide accurate and meaningful suggestions in a particular application, and to give appropriate weight to the sources of information."; The predicted word probability from the n-gram model reads on a raw score associated with each respective candidate output token, the predicted word probability from the unigram model reads on a token domain weight, a unigram model computing the probability or word score of a current word in a particular domain reads on determining a weight based on a frequency of an output token in a corpus of domain-specific data for a first domain, and determining a combined probability of the predicted word based on the first probability from the word n-gram model and the second probability from the unigram model by multiplying the probability from the word n-gram model and the probability from the unigram model reads on applying a token domain weight to each respective raw score associated with each respective candidate output token of the plurality of candidate output tokens to generate a plurality of weighted raw scores for the plurality of candidate output tokens.);
and generating, with the language model, an output token sequence based on the input token sequence and including a candidate output token associated with a weighted raw score of the plurality of weighted raw scores that is associated with a highest output probability (Column 3, lines 17-31, "Using the word n-gram language model, based on previously entered words, candidate words can be identified and a probability can be calculated for each candidate word. Such an n-gram probability can signify a likelihood that a candidate word corresponds to a word a user will enter. Using the unigram language model, based on one or more characters entered for a new word, candidate words beginning with the one or more entered characters can similarly be identified along with a probability for each candidate word. Such a unigram probability can, for example, signify a likelihood of a candidate word based on usage frequency in a particular language. The probabilities of the n-gram language model and unigram model can be combined, and the candidate word or words having the highest probability can be displayed for a user."; Determining a probability of candidate words based on previously entered words by combining the probabilities of the n-gram language model and unigram model and displaying the candidate word having the highest probability reads on generating an output token sequence based on the input token sequence and including a candidate output token associated with a weighted raw score of the plurality of weighted raw scores that is associated with a highest output probability.).
Willmore does not specifically disclose: generating, using an activation function, an output probability for each weighted raw score of the plurality of weighted raw scores for the plurality of candidate output tokens.
Shazeer teaches:
generating, using an activation function, an output probability for each weighted raw score of the plurality of weighted raw scores for the plurality of candidate output tokens (Column 4, lines 33-36, "In particular, for a given output position, the decoder neural network 150 generates a time step output that defines a probability distribution over possible output tokens at the given output position."; Column 7, lines 17-24, "At each generation time step, the linear layer 180 applies a learned linear transformation to the output of the last decoder subnetwork 130 in order to project the output of the last decoder subnetwork 130 into the appropriate space for processing by the softmax layer 190. The softmax layer 190 then applies a softmax function over the outputs of the linear layer 180 to generate the probability distribution over the possible network outputs at the generation time step."; Applying a softmax function to generate the probability distribution over possible output tokens reads on generating an output probability for each weighted raw score of the plurality of weighted raw scores for the plurality of candidate output tokens using an activation function.).
Shazeer is considered to be analogous to the claimed invention because it is in the same field of natural language generation. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Willmore to incorporate the teachings of Shazeer to apply a softmax function to generate the probability distribution over possible output tokens. Doing so would allow for generating a sequence of words that form an answer to a question, where an input sequence is a sequence of words that form the question (Shazeer; Column 3, lines 25-35).
Regarding claim 2, Willmore in view of Shazeer discloses the computer-implemented method as claimed in claim 1.
Willmore further discloses:
applying the token domain weight to each respective raw score associated with each respective candidate output token of the plurality of candidate output tokens comprises applying a first token domain weight to a first raw score of the plurality of raw scores associated with a first candidate output token of the plurality of candidate output tokens, the first token domain weight is based on a first frequency of a first output token in the corpus of domain-specific data (Column 7, lines 10-18, "Referring again to process 200 of FIG. 2, at block 206, a second probability of the predicted word can be determined based on a prefix character of a new word using a unigram model (e.g., unigram model 112 of FIG. 1, and as illustrated by unigram model 322 of FIG. 3). As understood by one of ordinary skill in the art, a unigram model can compute the probability (or word score) of a current word w in, for example, a particular language, domain, or the like. This word score can be represented by the probability P(w)."; Column 7, lines 29-34, "One of ordinary skill in the art will understand that the unigram model can be trained on relevant data sets (e.g., text relevant to a particular domain), structured in a variety of different ways (e.g., in a traversable trie), and modified for particular applications as desired."; Column 7, lines 35-50, "Referring again to process 200 of FIG. 2, at block 208, a combined probability of the predicted word can be determined based on the first probability from the word n-gram model at block 204 and the second probability from the unigram model at block 206. In one example, the product of the first and second probabilities for the same predicted word can be determined, such as by calculating P(w) · P(wq|c), where w and wq can be the same predicted word used in the unigram model and n-gram model, respectively. In another example, one or both of the first and second probabilities can be weighted or scored by a factor, modified by a logarithm, raised to a power (e.g., an empirically-derived value), or mathematically altered in other ways based, for example, on empirical testing to provide accurate and meaningful suggestions in a particular application, and to give appropriate weight to the sources of information."; The predicted word probability from the n-gram model reads on a raw score associated with each respective candidate output token, the predicted word probability from the unigram model reads on a token domain weight, a unigram model computing the probability or word score of a current word in a particular domain reads on the token domain weight being based on a frequency of an output token in the corpus of domain-specific data, and determining a combined probability of the predicted word based on the first probability from the word n-gram model and the second probability from the unigram model by multiplying the probability from the word n-gram model and the probability from the unigram model reads on applying a token domain weight to a raw score associated with a candidate output token of the plurality of candidate output tokens to generate a plurality of weighted raw scores for the plurality of candidate output tokens.);
and the computer-implemented method further comprises determining the first candidate output token corresponds to the first output token in the corpus of domain-specific data (Column 7, lines 10-18, "Referring again to process 200 of FIG. 2, at block 206, a second probability of the predicted word can be determined based on a prefix character of a new word using a unigram model (e.g., unigram model 112 of FIG. 1, and as illustrated by unigram model 322 of FIG. 3). As understood by one of ordinary skill in the art, a unigram model can compute the probability (or word score) of a current word w in, for example, a particular language, domain, or the like. This word score can be represented by the probability P(w)."; A unigram model computing the probability or word score of a current word in a particular domain reads on determining the candidate output token corresponds to the output token in the corpus of domain-specific data.).
Regarding claim 8, arguments analogous to claim 1 are applicable. In addition, Willmore discloses a processing system, comprising:
a memory comprising computer-executable instructions (Column 5, lines 3-6, "It should be appreciated that system 100 can include instructions stored in a non-transitory computer-readable storage medium, such as memory 108 or storage device 106, and executed by processor 104.");
and a processor configured to execute the computer-executable instructions (Column 4, lines 41-42, "In one example, processor 104 can be configured to perform any of the methods discussed herein") and cause the processing system to perform the steps of claim 1.
Regarding claim 9, arguments analogous to claim 2 are applicable.
Claims 3 and 10 are rejected under 35 U.S.C. 103 as being unpatentable over Willmore in view of Shazeer, and further in view of Wu et al. (US Patent No. 8,826,226), hereinafter Wu.
Regarding claim 3, Willmore in view of Shazeer discloses the computer-implemented method as claimed in claim 2.
Willmore further discloses:
generating, via a scoring method, a first frequency score indicative of the first frequency of the first output token in the corpus of domain-specific data (Column 7, lines 10-18, "Referring again to process 200 of FIG. 2, at block 206, a second probability of the predicted word can be determined based on a prefix character of a new word using a unigram model (e.g., unigram model 112 of FIG. 1, and as illustrated by unigram model 322 of FIG. 3). As understood by one of ordinary skill in the art, a unigram model can compute the probability (or word score) of a current word w in, for example, a particular language, domain, or the like. This word score can be represented by the probability P(w)."; A unigram model computing the probability or word score of a current word in a particular domain reads on generating a first frequency score indicative of the first frequency of the first output token in the corpus of domain-specific data.);
and assigning the first token domain weight to the first output token based on the first frequency score (Column 7, lines 35-50, "Referring again to process 200 of FIG. 2, at block 208, a combined probability of the predicted word can be determined based on the first probability from the word n-gram model at block 204 and the second probability from the unigram model at block 206. In one example, the product of the first and second probabilities for the same predicted word can be determined, such as by calculating P(w) · P(wq|c), where w and wq can be the same predicted word used in the unigram model and n-gram model, respectively. In another example, one or both of the first and second probabilities can be weighted or scored by a factor, modified by a logarithm, raised to a power (e.g., an empirically-derived value), or mathematically altered in other ways based, for example, on empirical testing to provide accurate and meaningful suggestions in a particular application, and to give appropriate weight to the sources of information."; Determining a combined probability of the predicted word based on the first probability from the word n-gram model and the second probability from the unigram model by multiplying the probability from the word n-gram model and the probability from the unigram model reads on assigning the first token domain weight to the first output token based on the first frequency score.).
Willmore in view of Shazeer does not specifically disclose: tokenizing the corpus of domain-specific data to obtain at least the first output token.
Wu teaches:
tokenizing the corpus of domain-specific data to obtain at least the first output token (Column 1, lines 62 - Column 2, line 1, "These and other embodiments can optionally include one or more of the following features. Generating a language model includes identifying a collection of n-grams from the documents from the one or more selected clusters. Each n-gram has a corresponding relative frequency of occurring in the documents from the one or more selected clusters. Each cluster is associated with a topic or domain."; Identifying a collection of n-grams from documents from a cluster, where each n-gram has a corresponding relative frequency of occurring in the documents from the cluster and the cluster is associated with a topic domain, reads on tokenizing the corpus of domain-specific data to obtain at least the first output token.).
Wu is considered to be analogous to the claimed invention because it is in the same field of natural language generation. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Willmore in view of Shazeer to incorporate the teachings of Wu to identify a collection of n-grams from documents from a cluster, where each n-gram has a corresponding relative frequency of occurring in the documents from the cluster and the cluster is associated with a topic domain. Doing so would allow for generating a custom language model that can be topic-dependent and adaptive to a topic of a user's current input (Wu; Column 3, lines 5-16).
Regarding claim 10, arguments analogous to claim 3 are applicable.
Claims 4 – 5 and 11 – 12 are rejected under 35 U.S.C. 103 as being unpatentable over Willmore in view of Shazeer and Wu, and further in view of Subrahmanian et al. (US Patent No. 12,417,506), hereinafter Subrahmanian.
Regarding claim 4, Willmore in view of Shazeer and Wu discloses the computer-implemented method as claimed in claim 3.
Willmore further discloses:
generating, via the scoring method, a second frequency score for the new output token (Column 7, lines 10-18, "Referring again to process 200 of FIG. 2, at block 206, a second probability of the predicted word can be determined based on a prefix character of a new word using a unigram model (e.g., unigram model 112 of FIG. 1, and as illustrated by unigram model 322 of FIG. 3). As understood by one of ordinary skill in the art, a unigram model can compute the probability (or word score) of a current word w in, for example, a particular language, domain, or the like. This word score can be represented by the probability P(w)."; A unigram model computing the probability or word score of a current word in a particular domain reads on generating a frequency score for the new output token.);
and assigning a second token domain weight to the new output token based on the second frequency score (Column 7, lines 35-50, "Referring again to process 200 of FIG. 2, at block 208, a combined probability of the predicted word can be determined based on the first probability from the word n-gram model at block 204 and the second probability from the unigram model at block 206. In one example, the product of the first and second probabilities for the same predicted word can be determined, such as by calculating P(w) · P(wq|c), where w and wq can be the same predicted word used in the unigram model and n-gram model, respectively. In another example, one or both of the first and second probabilities can be weighted or scored by a factor, modified by a logarithm, raised to a power (e.g., an empirically-derived value), or mathematically altered in other ways based, for example, on empirical testing to provide accurate and meaningful suggestions in a particular application, and to give appropriate weight to the sources of information."; Determining a combined probability of the predicted word based on the first probability from the word n-gram model and the second probability from the unigram model by multiplying the probability from the word n-gram model and the probability from the unigram model reads on assigning a token domain weight to the new output token based on the frequency score.).
Willmore in view of Shazeer and Wu does not specifically disclose: further comprising: based on the first frequency score associated with the first output token satisfying a frequency score threshold, generating, via an embedding component, a vector embedding for the first output token; generating a new output token based on the vector embedding.
Subrahmanian teaches:
based on the first frequency score associated with the first output token satisfying a frequency score threshold, generating, via an embedding component, a vector embedding for the first output token (Column 5, line 64 - Column 6, line 1, "In one embodiment, pre-processor 201 is further configured to identify the word embeddings for each word in the documents of the built corpus as embedding vectors, such as via Word2Vec. A word embedding associates a numeric vector with each word."; Column 11, lines 54-58, "In step 404, pre-processor 201 of document generator 101 computes the concept importance value by, for example, computing the inverse document frequency value for the concept or word from the corpus of domain specific documents that was not eliminated in step 403 for each concept"; Column 14, line 62 - Column 15, line 8, "The following discusses the constraints of the optimization problem in connection with the explicit joint concept replacement method (see optimization problem (1)). Constraint (i) indicates that a concept in the selected set B of bins should be replaced at least μ times to ensure that fake documents are sufficiently different from d. The selection of a set of bins suggests that the replacement concepts c′ for c should not be “too close” to the original c nor should they be “too far” away. In one embodiment, this constraint specifies a range of concept importance values. For example, if n=10, and B={B8, B9}, the concept importance values (e.g., TFIDF scores) are within the 70%˜ 90% th percentile of all concept importance values."; Selecting concepts for replacement based on an inverse document frequency value for the concept or word from the corpus of domain specific documents, where embedding vectors are determined for the words, reads on generating a vector embedding for the first output token based on the first frequency score associated with the first output token satisfying a frequency score threshold.);
generating a new output token based on the vector embedding (Column 4, lines 61-63, "In particular, embodiments of the present invention merge word embeddings and clustering in order to identify potential replacements for concepts."; Column 14, lines 9-12, "In step 505, optimizer 205 of document generator 101 replaces the identified concepts in document 103 with the corresponding identified replacement concepts as disused further below."; Determining a concept replacement by merging and clustering word embeddings reads on generating a new output token based on the vector embedding.).
Subrahmanian is considered to be analogous to the claimed invention because it is in the same field of natural language generation. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Willmore in view of Shazeer and Wu to incorporate the teachings of Subrahmanian to select concepts for replacement based on an inverse document frequency value for the concept or word from the corpus of domain specific documents, where embedding vectors are determined for the words, and determine a concept replacement by merging and clustering word embeddings. Doing so would allow for improving protection of data by automatically generating fake versions of a given document similar enough to the original document to be credible (Subrahmanian; Column 4, lines 52-60).
Regarding claim 5, Willmore in view of Shazeer, Wu, and Subrahmanian discloses the computer-implemented method as claimed in claim 4.
Subrahmanian further teaches:
wherein: the scoring method comprises a term frequency-inverse document frequency (TF-IDF) method (Column 11, lines 54-58, "In step 404, pre-processor 201 of document generator 101 computes the concept importance value by, for example, computing the inverse document frequency value for the concept or word from the corpus of domain specific documents that was not eliminated in step 403 for each concept");
and the embedding component comprises a Word2Vec system of models (Column 5, line 64 - Column 6, line 1, "In one embodiment, pre-processor 201 is further configured to identify the word embeddings for each word in the documents of the built corpus as embedding vectors, such as via Word2Vec. A word embedding associates a numeric vector with each word.").
Subrahmanian is considered to be analogous to the claimed invention because it is in the same field of natural language generation. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Willmore in view of Shazeer, Wu, and Subrahmanian to further incorporate the teachings of Subrahmanian to determine a frequency score using a term frequency-inverse document frequency method and generate word vectors using Word2Vec. Doing so would allow for improving protection of data by automatically generating fake versions of a given document similar enough to the original document to be credible (Subrahmanian; Column 4, lines 52-60).
Regarding claim 11, arguments analogous to claim 4 are applicable.
Regarding claim 12, arguments analogous to claim 5 are applicable.
Claims 6 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Willmore in view of Shazeer, and further in view of Singh et al. (US Patent No. 12,394,411), hereinafter Singh.
Regarding claim 6, Willmore in view of Shazeer discloses the computer-implemented method as claimed in claim 1, does not specifically disclose: wherein the language model is: fine-tuned for the first domain; or fine-tuned for a second domain, different from the first domain.
Singh teaches:
wherein the language model is: fine-tuned for the first domain; or fine-tuned for a second domain, different from the first domain (Column 9, lines 40-48, "For example, response system 244 can attempt to interpret the requests according to the most likely intent by invoking a sentence generation model (e.g., NLG model) fine-tuned for a particular domain or application (e.g., restaurant domain or application), generate one or more sentences with a high probability of having the same intent, return those sentences as a response, output, or revised query, and fulfill the specific action defined by the intent."; Column 18, lines 23-31, "As shown in FIG. 5, the general-purpose NLG model 502 can be fine-tuned with training data 503 (e.g., domain data maintained in data store 322), or mass linguistic data that is specific for a domain or an application. Because the general pattern of a language can be different from the specific language used in a particular domain or application, the general-purpose NLG model 502 can be fine-tuned for its own domain and target purpose, such as a restaurant domain, or a specific restaurant."; Fine-tuning a sentence generation model for a particular domain reads on the language model being fine-tuned for the first domain.).
Singh is considered to be analogous to the claimed invention because it is in the same field of natural language generation. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Willmore in view of Shazeer to incorporate the teachings of Singh to fine-tune a sentence generation model for a particular domain. Doing so would allow for improving the accuracy and effectiveness of a virtual assistant in understanding a user's spoken query (Singh; Column 3, line 47 - Column 4, line 3).
Regarding claim 13, arguments analogous to claim 6 are applicable.
Claims 7 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Willmore in view of Shazeer, and further in view of Cohen et al. (US Patent Application Publication No. 2025/0335816), hereinafter Cohen.
Regarding claim 7, Willmore in view of Shazeer discloses the computer-implemented method as claimed in claim 1, does not specifically disclose: wherein: the first domain comprises a tax domain; the input token sequence is associated with a user query for the tax domain; and the output token sequence represents a response to the user query for the tax domain.
Cohen teaches:
the first domain comprises a tax domain; the input token sequence is associated with a user query for the tax domain; and the output token sequence represents a response to the user query for the tax domain (Paragraph 0031, lines 1-10, "A dataset 105 may be provided to a validated language processing machine learning model 110. The dataset may be a dataset associated with a domain for which user may ask questions that are to be answered by a target language model. For example, the domain may be income tax filing, the dataset 105 may comprise tax filing instructions, and users may ask questions related to income tax filing. Thus, the goal of training a target language model 160 in this example may be to allow the target language model 160 to answer questions related to income tax filing."; A domain for which a user may ask questions that are to be answered by a target language model, where the domain is income tax filing, reads on the first domain comprising a tax domain, users asking questions related to income tax filing reads on the input token sequence being associated with a user query for the tax domain, and the target language model answering questions related to income tax filing reads on the output token sequence representing a response to the user query for the tax domain.).
Cohen is considered to be analogous to the claimed invention because it is in the same field of natural language generation. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Willmore in view of Shazeer to incorporate the teachings of Cohen to train a target language model in an income tax filing domain. Doing so would allow for a target language model answering questions related to income tax filing (Cohen; Paragraph 0031, lines 1-10).
Regarding claim 14, arguments analogous to claim 7 are applicable.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Houghton et al. (US Patent No. 11,557,289) teaches a method for automatically generating a topic specific language model and recognizing words in a speech signal using the generated model.
Biadsy et al. (US Patent No. 11,557,289) teaches a method for language models using domain-specific model components.
Li et al. ("N-gram Is Back: Residual Learning of Neural Text Generation with n-gram Language Model") teaches a method for combining a neural language model with an n-gram model.
Diao et al. ("Taming Pre-trained Language Models with N-gram Representations for Low-Resource Domain Adaptation") teaches a method for incorporating information of unseen and domain-specific words via the adaptation of word-based n-grams into a generic pretrained model.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to James Boggs whose telephone number is (571)272-2968. The examiner can normally be reached M-F 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 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.
/JAMES BOGGS/Examiner, Art Unit 2657
