DETAILED ACTION
Claims 1-20 are presented for examination.
This office action is in response to submission of application on 21-NOVEMBER-2025.
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 .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 23-JUNE-2022 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Response to Amendment
The amendment filed 21-NOVEMBER-2025. in response to the previous office action mailed 25-AUGUST-2025 has been entered. Claims 1-20 remain pending in the application.
With regards to the non-final office action’s rejection under 101, the amendment to the claims have overcome the previous action’s rejection.
With regards to the non-final office action’s rejections under 102 and 103, the amendments to the claims necessitated a new consideration of the art. After this consideration, the examiner respectfully disagrees with the applicant’s arguments that the art referenced in the previous office action does not teach the amendment claim limitations. A new 102 rejection over the prior art has been provided:
Furthermore, Arora discloses if it is determined that the location feature has a fixed vocabulary, retrieving, from a storage storing an embedding matrix with a fixed dictionary and size, a pre-trained embedding related to the location feature with distance-awareness thereby improving efficiency and accuracy of downstream usage, wherein the embedding for the location feature is learned based on distances between different pairs of locations:
Arora teaches that in the case where the location feature is a zip code, i.e. determines that the location feature has a fixed vocabulary, then Arora retrieves known embedding about the city or county that includes the zip code (Paragraph 67). This would be an embedding matrix with a fixed dictionary and size as the list of cities or counties is set and finite, while containing an embedded distance from other locations. The city or county that contains the zip code is therefore used a pre-trained embedding related to the location feature wherein the embedding may also rely on nearby locations, wherein using nearby locations in conjunction with the original city or county would be learning the embedding based on distances between different pair of locations (Paragraph 67).
Finally, Arora may then disclose if it is determined that the location feature has an open vocabulary, determining whether an embedding for the location feature exists, if an embedding exists, retrieving a pre-trained embedding, and if an embedding does not exist, learning an embedding related to the location feature via a multilayer neural network:
Arora teaches that for a previously unknown embedding, or an embedding that does not exist, an embedding is learned from previous knowledge via a machine learning model (Paragraph 38) wherein the machine learning model may be a multilayer neural network (Paragraph 13).
Furthermore, this demonstrates a determination of whether the embedding exists, as Arora does not demonstrate that the same process is performed for location that do have previous data (Paragraph 38). Likewise, the use of previous data demonstrates that when an embedding exists, the pre-trained embedding is retrieved, as the addition of future sites may use a pre-trained model trained on historical data (Paragraph 39) that would already have learned historical embeddings.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 1-2, 4, 8-9, 11, 15-16, 18 rejected under 35 U.S.C. 102(a)(2) as being anticipated by Arora et al. (Pub. No. US 20200210885 A1, filed December 31st 2018, hereinafter Arora).
Regarding claim 1:
Claim 1 recites:
A method implemented on at least one processor, a memory, and a communication platform for characterizing data, comprising: receiving a location feature; determining whether the location feature has a fixed vocabulary or an open vocabulary; if it is determined that the location feature has a fixed vocabulary, retrieving, from a storage storing an embedding matrix with a fixed dictionary and size, a pre-trained embedding related to the location feature with distance-awareness thereby improving efficiency and accuracy of downstream usage, wherein the embedding for the location feature is learned based on distances between different pairs of locations; if it is determined that the location feature has an open vocabulary, determining whether an embedding for the location feature exists, if an embedding exists, retrieving a pre-trained embedding, and if an embedding does not exist learning an embedding related to the location feature via a multilayer neural network; and generating a representation of the location feature based on the embedding related to the location feature, wherein the representation of the location feature using the embedding is to be used for location related predictions
Regarding the limitation a method implemented on at least one processor, a memory, and a communication platform for characterizing data, comprising:
Arora teaches the use of a computer which is known to encompass the use of a processor and memory (Paragraph 86). Furthermore, the data processing platform of Arora (Paragraph 86) would be a platform for characterizing data and hence would be analogous to the communication platform.
Regarding the limitation receiving a location feature:
Arora teaches receiving input include an indication of a location (Paragraph 22). This would be the location feature.
Regarding the limitation determining whether the location feature has a fixed vocabulary or an open vocabulary
Arora teaches that a particular site may be identified by either latitude and longitude, an example of an open vocabulary as the values may be continuous, or by a zip code, which would be fixed as there are a set number of zip codes (Paragraph 38). This would be an example of determining whether the location feature has a fixed or open vocabulary, wherein the location feature would be the geographical location of the site. Furthermore, Arora treats these two cases slightly differently – in the case where coordinates are used, if the site is previously unknown (i.e., after determining that the embedding does not exist) then predictions for the location, analogous to the embeddings, are made based on proximity and patterns of previously known locations (Paragraph 38).
Regarding the limitation if it is determined that the location feature has a fixed vocabulary, retrieving, from a storage storing an embedding matrix with a fixed dictionary and size, a pre-trained embedding related to the location feature with distance-awareness thereby improving efficiency and accuracy of downstream usage, wherein the embedding for the location feature is learned based on distances between different pairs of locations:
Arora teaches that in the case where the location feature is a zip code, i.e. determines that the location feature has a fixed vocabulary, then Arora retrieves known embedding about the city or county that includes the zip code (Paragraph 67). This would be an embedding matrix with a fixed dictionary and size as the list of cities or counties is set and finite, while containing an embedded distance from other locations. The city or county that contains the zip code is therefore used a pre-trained embedding related to the location feature wherein the embedding may also rely on nearby locations, wherein using nearby locations in conjunction with the original city or county would be learning the embedding based on distances between different pair of locations (Paragraph 67).
Regarding the limitation if it is determined that the location feature has an open vocabulary, determining whether an embedding for the location feature exists, if an embedding exists, retrieving a pre-trained embedding, and if an embedding does not exist learning an embedding related to the location feature via a multilayer neural network:
Arora teaches that for a previously unknown embedding, or an embedding that does not exist, an embedding is learned from previous knowledge via a machine learning model (Paragraph 38) wherein the machine learning model may be a deep neural network which would be a multilayer neural network (Paragraph 13).
Furthermore, this demonstrates a determination of whether the embedding exists, as Arora does not demonstrate that the same process is performed for location that do have previous data (Paragraph 38). Likewise, the use of previous data demonstrates that when an embedding exists, the pre-trained embedding is retrieved, as the addition of future sites may use a pre-trained model trained on historical data (Paragraph 39) that would already have learned historical embeddings.
Regarding the limitation generating a representation of the location feature based on the embedding relating to the location feature, wherein the representation of the location feature using the embedding is to be used for location related predictions
Arora teaches that encoded senses of distance can be used to provide prediction for locations for which no data in the training set exists (Paragraph 38). This would be a representation of a location feature based on the embedding, wherein the representation of the location feature using the embedding is to be used for location related predictions as the senses of distance between location are used in the prediction for a location.
Regarding claim 2, which depends upon claim 1:
Claim 2 recites:
The method of claim 1, wherein the location feature includes one of a zip code and an IP address.
Arora anticipates the method of claim 1 upon which claim 2 depends. Regarding the limitation of claim 2:
Arora teaches indicating locations to a model, wherein one form of the indicated location may be a zip code (Paragraph 7). This would encompass one of a zip code and an IP address.
Regarding claim 4, which depends upon claim 2:
Claim 4 recites:
The method of claim 2, wherein the location feature is linked to a coordinate of a center of a region associated with the location feature.
Arora anticipates the method of claim 2 upon which claim 4 depends. Regarding the limitation of claim 4:
Arora teaches a knowledge of a GPS location, wherein the location may be a central GPS location for a zip code. The zip code would be the location feature and the area of the zip code the region associated with the location feature, with the central GPS location being the center of a region.
Claims 8-9 and 11 recite a non-transitory computer readable storage medium that parallels the method of claims 1-2 and 4 respectively. Therefore, the analysis discussed above with respect to claims 1-2 and 4 also applies to claims 8-9 and 11 respectively. Accordingly, claims 8-9 and 11 are rejected based on substantially the same rationale as set forth above with respect to claims 1-2 and 4 respectively.
Claims 15-16 and 18 recite a non-transitory computer readable storage medium that parallels the method of claims 1-2 and 4 respectively. Therefore, the analysis discussed above with respect to claims 1-2 and 4 also applies to claims 15-16 and 18 respectively. Accordingly, claims 15-16 and 18 are rejected based on substantially the same rationale as set forth above with respect to claims 1-2 and 4 respectively.
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 3, 10, 17 rejected under 35 U.S.C. 103 as being unpatentable over Arora in view of Lumezanu et al. (Pub. No. US 20190098050 A1, filed August 13th 2018, hereinafter Lumezanu).
Regarding claim 3, which depends upon claim 2:
Claim 3 recites:
The method of claim 2, wherein the zip code has a fixed vocabulary; and the IP address has an open vocabulary.
Arora anticipates the method of claim 2 upon which claim 3 depends. Regarding the limitation wherein the zip code has a fixed vocabulary:
Arora teaches that there may be no training data for certain zip code (Paragraph 7) which shows that zip codes are not learned from the data be preset beforehand, such that they are present in the process even if they are not in the data. This would constitute a fixed vocabulary of zip codes.
However, Arora does not teach the IP address has an open vocabulary:
Lumezanu in the same field of endeavor of information management of network connections teaches a sparse set of known distances between target and source IP addresses (Paragraph 6). This demonstrates that not all relevant IP addresses are known in full to Lumezanu, which would constitute an open vocabulary.
Arora, Lumezanu, and the present application are all in the same field of endeavor of information management of network connections.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to implement a method that utilized the teachings of Arora and the teachings of Lumezanu. This would have granted the advantage of not needing the target to know the distances of every possible source IP address (Lumezanu, Paragraph 5).
Claim 10 recites a non-transitory computer readable storage medium that parallels the method of claim 3. Therefore, the analysis discussed above with respect to claim 3 also applies to claim 10. Accordingly, claim 10 is rejected based on substantially the same rationale as set forth above with respect to claim 3.
Claim 17 recites a system that parallels the method of claim 3. Therefore, the analysis discussed above with respect to claim 3 also applies to claim 17. Accordingly, claim 17 is rejected based on substantially the same rationale as set forth above with respect to claim 3.
Claims 5-6, 12-13, and 19 rejected under 35 U.S.C. 103 as being unpatentable over Arora in view of Segev (Pub. No. US 10692004 B1, filed June 30th 2019, hereinafter Segev).
Regarding claim 5, which depends upon claim4:
Claim 5 recites:
The method of claim 4, wherein the embedding for the location feature is trained via machine learning based on a training data batch having a plurality of location features, wherein the machine learning comprises: determining the coordinate associated with each of the location features included in the training data batch; determining a pair-wise distance of each of pairs of location features in the training data batch based on the coordinates of the location features of the pair to generate a distance matrix; initializing an embedding for each of the location features in the training data batch; estimating a similarity between each of pairs of embeddings of the location features in the training data batch; computing a loss based on the pair-wise distances of pairs of location features in the distance matrix and the similarities of embeddings of corresponding pairs of location features; adjusting values of the embeddings of the location features by minimizing the loss; and repeating steps of estimating, computing, and adjusting until a pre-determined condition with respect to the loss is met.
Arora anticipates the method of claim 4 upon which claim 5 depends. Regarding the limitation wherein the embedding for the location feature is trained via machine learning based on a training data batch having a plurality of location features, wherein the machine learning comprises: determining the coordinate associated with each of the location features included in the training data batch:
Arora teaches using a machine learning model to encode a sense of distance between locations indicated in the training data (Paragraph 38). The locations indicated in the training data would be the plurality of location features in the training data batch while the embedding would be the sense of distance. Furthermore, these locations may be associated with coordinates (Paragraph 38).
Regarding the limitation initializing an embedding for each of the location features in the training data batch:
Arora teaches determining a sense of distance between locations in the training data (Paragraph 38). This would be initialized the embedding for the location features in the training data batch.
Regarding the limitation adjusting values of the embeddings of the location features by minimizing the loss; and repeating steps of estimating, computing, and adjusting until a pre-determined condition with respect to the loss is met:
Arora teaches combining multiple models in order to provide the smallest prediction error on test data (Paragraph 81). Providing the smallest prediction error would be minimizing the loss and likewise the multiple machine learning models that are generated in order to be combined would be repeating the steps until the smallest prediction error is reached, which would be a pre-determined condition.
However, Arora does not teach determining a pair-wise distance of each of pairs of location features in the training data batch based on the coordinates of the location features of the pair to generate a distance matrix:
Segev in the same field of endeavor of information management teaches that pair-wise distances between the entries of each vector in an input matrix are computed to produce a similarity matrix, which is then processed to compute a distance matrix (Paragraph 87). This would include cases where the values are coordinates of the location features as taught by Arora as Segev determines pair-wise distances of each pair of the vectors to generate a distance matrix
Arora, Segev, and the present application are all in the same field of endeavor of information management.
Furthermore, Arora does not teach estimating a similarity between each of pairs of embeddings of the location features in the training data batch:
Segev teaches generating a similarity matrix by computing pair-wise distances between entries of each vector is a matrix (Paragraph 81). A similarity matrix would be a similarity between each pair.
Finally, Arora does not teach computing a loss based on the pair-wise distances of pairs of location features in the distance matrix and the similarities of embeddings of corresponding pairs of location features:
Segev teaches computing a cost function, which is a loss function, based on the parameters of the neural network (Paragraph 81) which include the matrices described above.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to implement a method that utilized the teachings of Arora and the teachings of Segev. This would have granted the advantage of requiring less computational effort as well faster processing (Segev, Paragraph 15).
Regarding claim 6, which depends upon claim 5:
Claim 6 recites:
The method of claim 5, wherein when the location feature corresponds to a zip code, the embedding is retrieved from a storage for previously trained embeddings for zip codes; and the step of initializing is performed by assigning random numbers as values of each of the embeddings for the location features in the training data batch.
Arora in view of Segev teaches the method of claim 5 upon which claim 6 depends. Regarding the limitation wherein when the location feature corresponds to a zip code, the embedding is retrieved from a storage for previously trained embeddings for zip codes:
Arora teaches that models can be trained using pre-qualification results collected over time for different locations (Paragraph 3). The pre-qualification results would be the embeddings retrieved from storage for previously trained embeddings for zip codes, as the location of Arora may be a zip code.
However, Arora does not teach the step of initializing is performed by assigning random numbers as values of each of the embeddings for the location features in the training data batch:
Segev teaches that data is prepared for processing, or initialized, by application of random projections to a given input matrix (Paragraph 40). The random projections would be random numbers assigned to the embeddings of the input matrix, which would be the training data.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to implement a method that utilized the teachings of Arora and the teachings of Segev. This would have granted the advantage of requiring less computational effort as well faster processing (Segev, Paragraph 15).
Claims 12-13 recite a non-transitory computer readable storage medium that parallels the method of claims 5-6 respectively. Therefore, the analysis discussed above with respect to claims 5-6 also applies to claims 12-13 respectively. Accordingly, claims 12-13 are rejected based on substantially the same rationale as set forth above with respect to claims 5-6 respectively.
Claim 19 recites a system that parallels the method of claim 5. Therefore, the analysis discussed above with respect to claim 5 also applies to claim 19. Accordingly, claim 19 is rejected based on substantially the same rationale as set forth above with respect to claim 5.
Claims 7, 14, 20 rejected under 35 U.S.C. 103 as being unpatentable over Arora in view of Segev further in view of Lumezanu.
Regarding claim 7, which depends upon claim 5:
Claim 7 recites:
The method of claim 5, wherein when the location feature corresponds to an IP address having a plurality of digits, the step of initializing comprises: devising a one hot vector for each of the plurality of digits to generate a corresponding plurality of one hot vectors; feeding each of the plurality of one hot vectors to a corresponding layer of a multilayer neural network, respectively, wherein each of the plurality of layers generates, as an output, a linear combination of an input along with an activation function, the input is a concatenation of a one hot vector of a corresponding digit with an output from a previous layer; outputting, at the last layer of the multilayer neural network, an output vector as the embedding for the location feature.
Arora in view of Segev teaches the method of claim 5 upon which claim 7 depends. Regarding the limitation outputting, at the last layer of the multilayer neural network, an output vector as the embedding for the location feature:
Arora teaches outputting an output vector from a model that represents a corresponding service level available at a location (Paragraph 50). This would be an embedding for the location feature.
However, Arora in view of Segev does not teach wherein when the location feature corresponds to an IP address having a plurality of digits, the step of initializing comprises: devising a one hot vector for each of the plurality of digits to generate a corresponding plurality of one hot vectors:
Lumezanu teaches that each byte of an IP address can be expressed as a one-hot vector (Paragraph 30). This would be analogous to devising a one-hot vector for each of the digits as the byte of the IP address merely changes the way the digits of the IP address are expressed.
Furthermore, Arora in view of Segev does not teach feeding each of the plurality of one hot vectors to a corresponding layer of a multilayer neural network, respectively, wherein each of the plurality of layers generates, as an output, a linear combination of an input along with an activation function, the input is a concatenation of a one hot vector of a corresponding digit with an output from a previous layer:
Lumezanu teaches that bytes of the IP address, which may be represented as one-hot vectors as per above, are input into a neural network wherein for subsequent input layers the output of the previous layer is concatenated with a new input byte before being based to the activation function (Paragraph 38). This would be analogous to the layers generating an input that is a concatenation of a one-hot vector, as the byte may be a one-hot vector, and the previous layer’s output, and the input is combined with the activation function.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to implement a method that utilized the teachings of Arora and the teachings of Segev and the teachings of Lumezanu. This would have granted the advantage of not needing the target to know the distances of every possible source IP address (Lumezanu, Paragraph 5).
Claim 14 recites a non-transitory computer readable storage medium that parallels the method of claim 7. Therefore, the analysis discussed above with respect to claim 7 also applies to claim 14. Accordingly, claim 14 is rejected based on substantially the same rationale as set forth above with respect to claim 7.
Claim 20 recites a system that parallels the method of claim 7. Therefore, the analysis discussed above with respect to claim 7 also applies to claim 20. Accordingly, claim 20 is rejected based on substantially the same rationale as set forth above with respect to claim 7.
Response to Arguments
Applicant’s arguments filed 21-NOVEMBER-2025 have been fully considered, but the examiner believes that not all are fully persuasive.
Regarding the applicant’s remarks on the non-final office action’s 102 and 103 rejection of the claims, the applicant argues that none of Arora, Lumezanu, or Segev does not teach the amended limitations of these claims. As such, the applicant argues that all claims dependent on the above would additionally not be anticipated under 102 or obvious under 103. However, the examiner believes that Arora does teach the amended limitations and respectfully requests applicant’s consideration of the following:
The applicant claims that none of the art referenced in the previous action teaches the amended claim limitations. However, the examiner believes that Arora teaches the amended and original limitations of claim 1, 8, and 15.
Regarding the applicant’s claim that Arora does not disclose determining whether the location feature has a fixed or an open vocabulary and if it is determine that the location feature has an open vocabulary, determining whether an embedding for the location feature that has an open vocabulary exists:
Arora teaches that a particular site may be identified by either latitude and longitude, an example of an open vocabulary as the values may be continuous, or by a zip code, which would be fixed as there are a set number of zip codes (Paragraph 38). This would be an example of determining whether the location feature has a fixed or open vocabulary, wherein the location feature would be the geographical location of the site. Furthermore, Arora treats these two cases slightly differently – in the case where coordinates are used, if the site is previously unknown (i.e., after determining that the embedding does not exist) then predictions for the location, analogous to the embeddings, are made based on proximity and patterns of previously known locations (Paragraph 38).
Furthermore, Arora discloses if it is determined that the location feature has a fixed vocabulary, retrieving, from a storage storing an embedding matrix with a fixed dictionary and size, a pre-trained embedding related to the location feature with distance-awareness thereby improving efficiency and accuracy of downstream usage, wherein the embedding for the location feature is learned based on distances between different pairs of locations:
Arora teaches that in the case where the location feature is a zip code, i.e. determines that the location feature has a fixed vocabulary, then Arora retrieves known embedding about the city or county that includes the zip code (Paragraph 67). This would be an embedding matrix with a fixed dictionary and size as the list of cities or counties is set and finite, while containing an embedded distance from other locations. The city or county that contains the zip code is therefore used a pre-trained embedding related to the location feature wherein the embedding may also rely on nearby locations, wherein using nearby locations in conjunction with the original city or county would be learning the embedding based on distances between different pair of locations (Paragraph 67).
Finally, Arora may then disclose if it is determined that the location feature has an open vocabulary, determining whether an embedding for the location feature exists, if an embedding exists, retrieving a pre-trained embedding, and if an embedding does not exist, learning an embedding related to the location feature via a multilayer neural network:
Arora teaches that for a previously unknown embedding, or an embedding that does not exist, an embedding is learned from previous knowledge via a machine learning model (Paragraph 38) wherein the machine learning model may be a multilayer neural network (Paragraph 13).
Furthermore, this demonstrates a determination of whether the embedding exists, as Arora does not demonstrate that the same process is performed for location that do have previous data (Paragraph 38). Likewise, the use of previous data demonstrates that when an embedding exists, the pre-trained embedding is retrieved, as the addition of future sites may use a pre-trained model trained on historical data (Paragraph 39) that would already have learned historical embeddings.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ALEXANDRIA JOSEPHINE MILLER whose telephone number is (703)756-5684. The examiner can normally be reached Monday-Thursday: 7:30 - 5:00 pm, every other Friday 7:30 - 4:00.
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, Mariela Reyes can be reached on (571) 270-1006. 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.
/A.J.M./Examiner, Art Unit 2142
/Mariela Reyes/Supervisory Patent Examiner, Art Unit 2142