Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Arguments
The arguments filed 11/21/2025 have been entered. Claims 1-3, 5-6, 8, 10-17, 37-42 remain pending in the application.
Applicant’s argument, with respect to claim rejections of claim(s) 1-3, 5-6, 8, 10-17, 37-42 under 35 U.S.C 101 filed 08/25/2025 have been considered and they are persuasive. Therefore, the previous rejections as set forth in the previous office action will be withdraw.
Applicant’s argument, with respect to claim rejections of claim(s) 1-3, 5-6, 8, 10-17, 37-42 under 35 U.S.C 103 filed 08/25/2025 have been considered and they are not persuasive. Therefore, the previous rejections as set forth in the previous office action will be maintained.
The applicant argues that amended claim 1 recites usage of an “arrival ratio” defines as a proportion of a number of sub-samples actually detected in each sample to the total number of sub-samples included in the sample, and that this ratio is used to determine validify of the sample. Applicant contends that the cited references, particularly Yih and Oka, fail to discloses or suggest this feature. Applicant further asserts that Oka’s disclosed “accuracy” or “degree if satisfaction” relates to probability of assumed causes based on abnormality degree, rather than the claimed proportion of valid process data, and therefore is not equivalent to the claimed arrival ratio. Additionally, applicant maintains that the cited art does not disclose the claimed glass substrate/panel or filtering of samples based on arrival ratio. Accordingly, Applicant submits that the pending claims are patentable over the cited combination.
The examiner respectfully agrees that the cited prior arts does not teach the amended claim which recite an arrival ratio defines as a proportion of a number of sub-samples actually detected in each sample to the total number of sub-samples included in the sample, and that this ratio is used to determine validify of the sample, as well as the claimed glass substrate/panel or filtering of samples based on arrival ratio.
However, upon further consideration, new ground(s) of rejections have been raised (See Below.)
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 1-3, 5, 6, 12, 15-16, 37-42 are rejected under 35 U.S.C. 103 as being unpatentable over Yih et.al (US 20080319932 A1) in view of Oka et.al (US 20230213927 A1), further in view of Song et.al (US 20180238958 A1)
Regarding claim 1,
Yih teaches a part of the 1st limitation “obtaining sample data in response to a user’s input operation on a graphical interface, the sample data including characteristic data;” (paragraph 35 “When training data is selected that is characteristic of the data that is most important, then the classifier will be well optimized for this particular data.”, paragraph 67 “At 600, training data and threshold are received. The training data and/or threshold can, but is not limited to, being generated from an application, manually input”, and paragraph 90 “A user can enter commands and information into the computer 1002 through one or more wired/wireless input devices ... These and other input devices are often connected to the processing unit 1004 through an input device interface”. Yih discloses a system and method that facilitates and effectuates optimizing a classifier for greater performance. Within the disclosure, Yih discloses a user may provide training data through manually input on an interface of the application. The training data may include characteristic of the data that is most important such that the classifier may be optimized accordingly, wherein the characteristics of the training data may be obtained in view of the teaching by Song below.)
Yih teaches the 2nd limitation “displaying a sample distribution diagram on the graphical interface based on the sample data ...” (paragraph 40 “Turning to FIGS. 2A-2D, are illustrated an exemplary set of data for training a multi-stage classifier... different regions of the data have different relative distributions. ... FIG. 2A illustrates an exemplary training data set when viewed in a two-dimensional representation where the x-axis and y-axis represent the values of the two features in the examples”, and paragraph 91 “A monitor 1044 or other type of display device is also connected to the system bus 1008 via an interface”. Yih discloses an exemplary data visualization in FIGS. 2A-2D, which illustrate the distribution of training data, wherein the figure displays different data regions having different distributions, suggesting a sample distribution diagram because they graphically depict how the training data are distributed relative to classification boundaries. Furthermore, Yih discloses a monitor or a display device is connected to the interface of the system, which suggest that the monitor may display these data distribution that may be configured via the interface by a person ordinary skilled in the art.)
Yih teaches a part of the 3th limitation “obtaining a focus threshold used for classifying positive and negative samples; ...” (paragraph 37 “The first stage classifier component 110 is then used to classify the training data 102 with a defined threshold θ to produce first stage classified data 110. The threshold 0 is the accuracy and/or entropy that is desired, and can be user or system defined. Any appropriate threshold θ can be selected to coincide with the region of interest. For example, a false positive rate of 4% can be selected. Any instance of data within the training data 102 that is classified by the first stage classifier component 108 in the first stage classified data 110 as having a value less than the threshold θ of being positive is ignored”. Yih discloses the classifier applies a defined threshold to the training data in order to produce classified data. Yih further teaches that the threshold represents a confidence measure that can be user defined. For example, data instances with values below the threshold are ignored, while those meeting or exceeding the threshold are retained. Accordingly, a person ordinary skilled in the art would have been able to configure the threshold to determine whether training data are classified as positive or negative data. The threshold may be configured as an abnormality degree threshold based on the teaching combination with Oka below.)
Yih teaches the 4th limitation “displaying a mark of the focus threshold in the sample distribution diagram on the graphical interface” (paragraph 40 “FIG. 2B shows the first stage classifier that is learned. In the illustration, a threshold of 50% is used by the first stage classifier, but in practice, any threshold can be picked. Depending on the threshold picked, the classifier will optimize on a different type of data, and thus optimize for a different false positive rate.” Yih discloses Fig. 2B displays a mark of the threshold of the classifier model, wherein the Fig. 2B represent a data distribution diagram on the graphical interface as explained above.)
Yih teaches the 5th limitation “distinguishing data display effects of the positive and negative samples based on the focus threshold” (paragraph 40 “FIG. 2A illustrates an exemplary training data set when viewed in a two-dimensional representation where the x-axis and y-axis represent the values of the two features in the examples, a square □ is a positive example (spam) and an × is a negative example (not spam). Notice that the data is almost, but not quite linearly separable. FIG. 2B shows the first stage classifier that is learned. In the illustration, a threshold of 50% is used by the first stage classifier, but in practice, any threshold can be picked.” Yih discloses the example illustration in Fig. 2A, and Fig. 2B that represent training data distribution diagram, in which a threshold is picked and training data above and below the threshold may be identified. A person ordinary skilled in the art would have been able to distinguish the positive and negative data effects based on the threshold display, for example, according to Fig. 2B, a line of threshold is indicated, and any data points having display effect of being above the threshold may be considered as negative data, wherein data points below the threshold may be considered as positive data.)
Yih does not teach a part of the 1st limitation “... detection data of samples”. However, Oka teaches this limitation (paragraph 9 “An abnormal irregularity cause identifying device includes ..., an abnormality determination unit that continuously calculates, for each of the plurality of sensors, an abnormality degree representing an extent of an irregularity of process data” Oka discloses an abnormal irregularity cause identifying method and device. Within the disclosure, Oka discloses the device comprises of method to calculate an abnormality degree representing an extent of an irregularity of process data. This abnormality degree suggests detection data of samples. A person ordinary skilled in the art would have been able to incorporate this abnormal irregularity cause identifying device into the system and device by Yih to calculate an abnormality degree associated with each training data by Yih.)
Yih does not teach a part of the 3rd limitation “... wherein the focus threshold is determined based on the detection data of the samples”. However, Oka teaches this part of the limitation (paragraph 52 “On the other hand, in the continuous processing, the timing may be defined by the sampling interval, or the like as illustrated in FIG. 7. In a row of the threshold value, a threshold value that is a criterion for determining an abnormality in each abnormality determination technique is registered.” Oka discloses detection data of the samples may be considered as abnormality degree as explained above. Oka further discloses a threshold value that is a criterion for determining an abnormality is registered, suggesting a threshold that may be configured based on the abnormality degree of training data. A person ordinary skilled in the art would have been able to configure the threshold of abnormality degree using the user’s defined system that let the user to define a threshold bi Yih based on the teaching combination below, such that the threshold can be used to classify positive and negative training data instances, wherein the training data instances may correspond with an abnormality degree.)
Yih does not teach the 6th limitation “determining a cause of abnormality of the samples based on the positive and negative samples”. However, Oka teaches this limitation (paragraph 38 “the abnormal irregularity cause identifying device 1 creates an abnormality detection model based on a knowledge base that stores a correspondence relationship between an assumed cause and an influence appearing as an abnormality, for example. For example, a model for identifying an abnormal irregularity, a sign thereof, and a cause thereof is created based on a technique, which is created based on the knowledge base, to detect a deviation from an acceptable range for change in process data”, and paragraph 59 “the abnormality degree with a positive or negative sign may be calculated. FIG. 12 is a diagram for explaining an example of calculating an abnormality degree by magnitudes of distances of pieces of time-series data from a criterion. FIG. 13 is a diagram for explaining an example of calculating an abnormality degree by distances of the same pieces of time-series data from a criterion in consideration of positive-negative directions”. Oka discloses the abnormal irregularity cause identifying device 1 utilize an abnormality detection model to identify a cause of the abnormal irregularity, wherein the abnormal irregularity may correspond with an abnormality degree, which may be calculated with a positive or negative sign as illustrated in Fig. 12 and Fig. 13. The method by Oka may be incorporated with the teaching by Yih to calculate an abnormality degree with a positive or negative sign for each training data, thereby classifying the training data into positive and negative data, which allow the device to derive the cause based on analyzing the abnormal irregularity as displayed on the distribution diagram of the training data. A person ordinary skilled in the art would have been able to configure the positive and negative training data of Yih into the method by Oka based on the teaching motivation below.)
Before the effective filing date, it would have been obvious for a person ordinary skilled in the art to incorporate the teaching of a system and method that facilitates and effectuates optimizing a classifier for greater performance by Yih with the teaching of an abnormal irregularity cause identifying method and device by Oka. The motivation to do so is referred to in Oka’s disclosure (paragraph 8 “In general, it is desirable to prevent abnormal irregularity to suppress an impact on safety, stability, quality of products, cost, and the like in a production facility. The present technology aims to improve the performance to identify a cause of abnormal irregularity in a production facility.”, paragraph 88 “the cause diagnosis unit 145 calculates the degree of satisfaction based on the abnormality degree calculated by the abnormality determination unit 144 for an event serving as the assumed cause of irregularity of the process data. The user can recognize a candidate of the cause of the irregularity and an accuracy thereof based on the magnitude of the degree of satisfaction and can easily identify the cause of irregularity.” Oka discloses the advantage of the invention, which aim to improve the performance to identify a cause of abnormal irregularity. Oka provides the calculation and processing steps, which eventually allow a user to recognize a candidate of the cause of the irregularity and an accuracy thereof. While Yih discloses a method of classification that has been optimized for a specific area of the classification region that is of interest, the method may be further improved upon integrating methods to determine abnormality in data according to the teaching by Oka, such that positive and negative training data may be further utilized in accordance with the method as taught by Oka, to detect abnormality in training data, thus allow a better understanding of training data, in which any person ordinary skilled in the art may benefit from using this understanding of training data for any machine learning purposes. For example, after the classification method by Yih classify positive and negative training data instances based on the threshold, a calculation with regard to the positive and negative sign may be performed utilizing the method by Oka to obtain an abnormality degree for each training data, the data may be displayed on the diagram with the abnormality degree and the threshold such that the user may analyze and identify a cause toward the abnormality of training data. This cause of abnormality in training data might help the user to evaluate the training data or update the training data accordingly, which improve the training data, thus improve the training of a machine learning model as understood by a person ordinary skilled in the art.)
Yih/Oka does not teach part of the 1st limitation “... wherein the characteristic data of the samples includes an arrival ratio and at least one of and at least one of a product model, a detection site, an abnormal type, a production equipment, an environmental parameter, detection time or generation time: the samples each include a plurality of sub-samples; the samples each include a plurality of sub-samples; the arrival ratio is used to indicate a proportion of a number of sub-samples actually detected in each sample to the total number of sub-samples included in the sample”. However, Song teaches this part of the limitation (paragraph 4 “Yield can be defined using different concepts including functional considerations (whether the device has the intended functionality, e.g., performs the functions it is designed to perform), and parametric considerations (whether the device performs within the intended operating range, e.g., speed, power, voltage, resistance, temperature, etc.); and yield can also consider production efficiency. Defects that reduce yield can occur because of faulty designs and/or faulty manufacturing”, paragraph 6 “Various methods herein identify failure types that caused defective items to fail testing, group the defective items by the failure types to produce failure-type groups, and determine failure-type limited yield within each of the failure-type groups. Such methods analyze the defective items in each failure-type group to identify defect types that caused the failure types, and also determine failure-type group-specific defect ratios based on proportions of the defect types within each of the failure-type groups”, and paragraph 9 “methods determine a second proportion or second ratio, that is the ratio of the number of defective items that have that type of defect to the total number of defective items analyzed in the failure-type group, to produce a failure-type group-specific defect ratio (that is specific to the failure-type group).” Song discloses automatically determining a second ratio of defective items within each failure type group relative to the total number of items in the failure-type group. Under the broadest reasonable interpretation, the claimed “arrival ratio” corresponds to the second ratio by Song because both represent a proportion of detected sub-units within a defined set of items, the total number of defective items corresponds to the claimed sample and the items that have that type of defect corresponds to the claimed sub-samples. Accordingly, Song’s second ratio teaches or at least suggest the claimed arrival ratio indicating a proportion of subsamples actually detected in each sample to the total number of sub-samples included in the sample. Furthermore, Song teaches evaluating yield and defect analysis may be performed based on parametric considerations such as speed, power, voltage, resistance. Under the broadest reasonable interpretation, such parametric operating conditions correspond to the claimed “environmental parameter” included in the characteristic data of the samples. Because the claimed require the characteristic data to include the arrival ratio and at least one of the listed parameters, Song’s disclosure of these parametric operating conditions and ratio teaches the claimed requirement. Since song’s ratio represents a quantitative characteristic of items within a dataset and therefore would have been recognized by a person of ordinary skill in the art as suitable characteristic data for inclusion in Yih’s training data, wherein Yih’s training data corresponds to the claimed sample data under the broadest reasonable interpretation because both represent dataset of items for classification and analysis, such that Yih’s training data corresponds to the claimed sample data and Song’s ratio represents a characteristic of items within such data that would have been included in Yih’s dataset.)
Yih/Oka does not teach the limitation “wherein the samples are glass substrates, the sub-samples are panels, each glass substrate is cut into a plurality of panels after the glass substrate has performed various processes, the arrival ratio of each sample is a proportion of a number of panels arriving at the detection site in the plurality of panels to a total number of the plurality of panels.” However, Song teaches this part of the limitation (paragraph 9 “methods determine a second proportion or second ratio, that is the ratio of the number of defective items that have that type of defect to the total number of defective items analyzed in the failure-type group, to produce a failure-type group-specific defect ratio (that is specific to the failure-type group)” Song discloses determining the second ratio of the number of defective items having a particular defective type to the total number of defective items analyzed within a failure-type group. Under the broadest reasonable interpretation, Song’s failure-type group corresponds to the claimed sample because it represents a defined set of items being analyzed, and the number of defective items that have that type of defect to the total number of defective items corresponds to the claimed sum-samples as they are individual member of that set. Accordingly, Song’s second ratio represents a proportion of sum-samples within each sample relative to the total sub-samples within that sample, which teaches or at least suggest the claimed arrival ratio indicating a proportion of panels within a plurality of panels. The claim does not require any glass-specific processing beyond the recited proportional relationship, and therefore Song’s proportional analysis of grouped manufactured items reasonably satisfies the claimed limitation under the broadest reasonable interpretation.)
Before the effective filing date, it would have been obvious for a person ordinary skilled in the art to incorporate the teaching of a system and method that facilitates and effectuates optimizing a classifier for greater performance by Yih, and the teaching of an abnormal irregularity cause identifying method and device by Oka, with the teaching of a ratio to represent proportion of a certain number of items under a condition within a group to detect abnormality in item data for better data analysis by Song. The motivation to do so is referred to in Song’s disclosure (paragraph 3 “As advanced semiconductor technology becomes more complicated, the types of defects are increasing and the root cause identification and resolution for each type of defect also becomes more challenging. In order to achieve fast yield improvement under this situation with so many types of defects and limited resources, it is important to deploy the limited resources to focus on the defects with the greatest yield impact.”, and paragraph 8 “These methods analyze (electrically and physically) the defective items for physical defects, so as to identify defect types that caused such failures. For each failure-type group, these methods determining the defect-types include failure analysis... Each of the defect types can cause multiple failure types, and each of the failure types can result from multiple defect types. Further, the defect types include characteristics of the manufactured items that diverge from a design of the manufactured items.”, paragraph 12 “These methods rank the defect types, in inverse order to the failure-type influenced defect-type total limited yield. With this ranking, these methods can produce the most yield improvement by applying defect solutions to resolve the defect type, in order of the ranking of the defect types”, and paragraph 16 “the processor is capable of automatically weighting each failure-type group-specific defect ratio using the failure-type limited yield of the failure-type group to produce a weighted failure-type group-specific defect limited yield for each of the physical defect types within the failure-type group” Song discloses grouping defective manufactured items into failure-type groups and determining a ratio representing a proportion of items having a particular defect within the group, and further teaches evaluating defect behavior using parametric operating conditions (e.g., speed, power, voltage). These disclosures provide quantitative characteristic information about items that may constitute a dataset. It would have been obvious to a person ordinary skill in the art to incorporate such ratio-based and parametric characteristic information into the classification framework of Yih, which expressly seeks to improve classifier performance by appropriately characterizing and weighting training data (e.g., based oin false positive/false negative considerations). A person ordinary skilled in the art would have recognized the including additional quantitative characteristics of the items that may that may constitute a dataset, as taught by Song, would have predictably improved the effectiveness and accuracy of Yih’s classifier.)
Regarding claim 2 depends on claim 1, thus the rejection of claim 1 is incorporated.
Yih teaches the limitation “The method according to claim 1, wherein the focus threshold includes at least one first focus threshold” (paragraph 67 “At 600, training data and threshold are received. The training data and/or threshold can, but is not limited to, being generated from an application, manually input, loaded from a data store, and/or received from another system locally or remote” Yih discloses obtaining the threshold, which may be manually input by a user, wherein a person ordinary skilled in the art can manually input one or more threshold.)
Yih teaches the limitation “receiving a user’s setting operation of the at least one first focus threshold” (paragraph 67 “At 600, training data and threshold are received. The training data and/or threshold can, but is not limited to, being generated from an application, manually input, loaded from a data store, and/or received from another system locally or remote” Yih discloses training data and threshold may be manually input by a user, suggest a user’s setting operation of the threshold as the user manually input the threshold.)
Yih teaches the limitation “displaying at least one mark of the at least one first focus threshold in the sample distribution diagram on the graphical interface” (paragraph 40 “FIG. 2B shows the first stage classifier that is learned. In the illustration, a threshold of 50% is used by the first stage classifier, but in practice, any threshold can be picked. Depending on the threshold picked, the classifier will optimize on a different type of data, and thus optimize for a different false positive rate.” Yih discloses Fig. 2B displays a mark of the threshold of the classifier model, wherein the Fig. 2B represent a display of data distribution diagram on the graphical interface as explained above.)
Yih teaches the limitation “distinguishing the data display effects of the positive and negative samples based on the at least on first focus threshold” (paragraph 40 “FIG. 2A illustrates an exemplary training data set when viewed in a two-dimensional representation where the x-axis and y-axis represent the values of the two features in the examples, a square □ is a positive example (spam) and an × is a negative example (not spam). Notice that the data is almost, but not quite linearly separable. FIG. 2B shows the first stage classifier that is learned. In the illustration, a threshold of 50% is used by the first stage classifier, but in practice, any threshold can be picked.” Yih discloses the example illustration in Fig. 2A, and Fig. 2B that represent training data distribution diagram, in which a threshold is picked and training data above and below the threshold may be identified. A person ordinary skilled in the art would have been able to distinguish the positive and negative data effects based on the threshold display, for example, according to Fig. 2B, a line of threshold is indicated, and any data points having display effect of being above the threshold may be considered as negative data, wherein data points below the threshold may be considered as positive data.)
Regarding claim 3 depends on claim 2, thus the rejection of claim 2 is incorporated.
Yih teaches a part of the limitation “The method according to claim 2, wherein the at least one first focus threshold includes a first value; ...” (paragraph 37 “The threshold 0 is the accuracy and/or entropy that is desired, and can be user or system defined. Any appropriate threshold θ can be selected to coincide with the region of interest.” Yih discloses a threshold can be user defined, which may be represented as a value such as the threshold θ, suggesting the first value which represent the threshold within the claim.)
Oka teaches a part of the limitation “... distinguishing the data display effects of the positive and negative samples based on the at least one first focus threshold includes: distinguishing the data display effects of the positive and negative samples based on a relationship between magnitudes of the detection data of the samples and a magnitude of the first value” (paragraph 59 “For divergence from a criterion such as an average, the abnormality degree with a positive or negative sign may be calculated. FIG. 12 is a diagram for explaining an example of calculating an abnormality degree by magnitudes of distances of pieces of time-series data from a criterion. FIG. 13 is a diagram for explaining an example of calculating an abnormality degree by distances of the same pieces of time-series data from a criterion in consideration of positive-negative directions” Oka discloses calculating an abnormality degree by magnitudes of distances of time-series data from a criterion. The abnormality degree may be assigned with a positive or negative sign, which may correspond with the positive and negative training data based on the combination with the teaching by Yih. The calculation represents a relationship between the magnitude of the abnormality degree and the criterion, wherein the criterion may be the threshold with value by Yih as configured by a person ordinary skilled in the art. It would have been obvious to the person of ordinary skill to use Yih’s threshold as Oka’s criterion, such that the abnormality degree may comprise a positive or negative sign corresponding with the positive and negative training data, and the display of the data depends on whether the magnitude of the distance of each training data point is above or below the threshold.)
Regarding claim 5 depends on claim 1, thus the rejection of claim 1 is incorporated.
Yih teaches the limitation “screening the sample data based on a user’s filtering operation of at least one filtering threshold” (paragraph 35 “Typically, probabilistic classifiers attempt to maximize the probability across all of the training data. However, for e-mail spam filtering, it is important to produce a low false positive rate for classifying good e-mails as spam e-mails. One way to focus on this low false positive region is to explicitly select data for training that is in the region of interest”. Yih discloses a user operation of email spam filtering, thus suggest a threshold correspond to the filtering may be obtained, in which data is determined based on the threshold corresponding to the filtering process.)
Yih teaches the limitation “displaying a distribution diagram of screened samples on the graphical interface” (paragraph 40 “Turning to FIGS. 2A-2D, are illustrated an exemplary set of data for training a multi-stage classifier... different regions of the data have different relative distributions. ... FIG. 2A illustrates an exemplary training data set when viewed in a two-dimensional representation where the x-axis and y-axis represent the values of the two features in the examples”, and paragraph 91 “A monitor 1044 or other type of display device is also connected to the system bus 1008 via an interface”. Yih discloses an exemplary data visualization in FIGS. 2A-2D, which illustrate the distribution of training data, wherein the figure displays different data regions having different distributions, suggesting a sample distribution diagram because they graphically depict how the training data are distributed relative to classification boundaries. Furthermore, Yih discloses a monitor or a display device is connected to the interface of the system, which suggest that the monitor may display these data distribution that may be configured via the interface by a person ordinary skilled in the art.)
Regarding claim 6 depends on claim 5, thus the rejection of claim 5 is incorporated.
Oka teaches the limitation “The method according to claim 5, wherein the at least one filtering threshold includes at least one of an abnormal ratio threshold, an arrival ratio threshold, a production equipment threshold, an environmental parameter threshold, a detection time threshold or a generation time threshold” (paragraph 87 “Furthermore, the output control unit 146 outputs the abnormality degrees calculated ... and issues an alarm when any of the abnormality degrees exceeds a predetermined threshold value”. Oka discloses comparing the abnormality degrees with a predetermined threshold value, suggesting an abnormality degree threshold, which teach the abnormal ratio threshold within the claim.)
Regarding claim 8 depends on claim 1, thus the rejection of claim 1 is incorporated.
Song teaches the limitation “The method according to claim 1, wherein the detection data of the samples includes at least one of an abnormal ratio or a measurement parameter; the abnormal ratio is used to indicate a proportion of a number of abnormal sub-samples in each sample to a total number of sub-samples included in the sample.” (paragraph 9 “methods determine a second proportion or second ratio, that is the ratio of the number of defective items that have that type of defect to the total number of defective items analyzed in the failure-type group, to produce a failure-type group-specific defect ratio (that is specific to the failure-type group)” Song discloses automatically determining a second ratio of defective items within each failure type group relative to the total number of items in the failure-type group. Under the broadest reasonable interpretation, the claimed “arrival ratio” corresponds to the second ratio by Song because both represent a proportion of detected sub-units within a defined set of items, the total number of defective items corresponds to the claimed sample and the items that have that type of defect corresponds to the claimed sub-samples. Accordingly, Song’s second ratio teaches or at least suggest the claimed arrival ratio indicating a proportion of subsamples actually detected in each sample to the total number of sub-samples included in the sample.)
Regarding claim 12, the applicant is directed to the rejection of claim 1 above, because the claim recites similar limitation and processing steps, thus the claim is rejected under the same rationale.
Regarding claim 15, depends on claim 12 thus the rejection of claim 12 is incorporated. The applicant is directed to the rejection of claim 5 above, because the claim recites similar limitation and processing steps, thus the claim is rejected under the same rationale.
Regarding claim 16, depends on claim 15 thus the rejection of claim 15 is incorporated. The applicant is directed to the rejection of claim 6 above, because the claim recites similar limitation and processing steps, thus the claim is rejected under the same rationale.
Regarding claim 17, depends on claim 12 thus the rejection of claim 12 is incorporated. The applicant is directed to the rejection of claim 8 above, because the claim recites similar limitation and processing steps, thus the claim is rejected under the same rationale.
Regarding claim 37 depends on claim 1, thus the rejection of claim 1 is incorporated.
Yih teaches the limitation “A data processing apparatus, comprising a memory and a processor; the memory being coupled to the processor; the memory being used to store computer program codes, and the computer program codes including computer instructions; wherein when executing the computer instructions, the processor causes the data processing apparatus to perform the data processing method according to claim 1” (paragraph 66 “The claimed subject matter can be described in the general context of computer-executable instructions, such as program modules, executed by one or more components.”, paragraph 78 “Furthermore, all or portions of the claimed subject matter can be implemented as a system, method, apparatus, or article of manufacture using standard programming and/or engineering techniques to produce software, firmware, hardware or any combination thereof to control a computer to implement the disclosed subject matter.”, and paragraph 85 “With reference again to FIG. 10, the exemplary environment 1000 for implementing various aspects includes a computer 1002, the computer 1002 including a processing unit 1004, a system memory 1006 and a system bus 1008. The system bus 1008 couples system components including, but not limited to, the system memory 1006 to the processing unit 1004. The processing unit 1004 can be any of various commercially available processors.”, and paragraph 89 “A number of program modules can be stored in the drives and RAM 1012, including an operating system 1030, one or more application programs 1032, other program modules 1034 and program data 1036.” Yih discloses the embodiment of the subject matter can be implemented as a system, method, apparatus which comprises of computer including one or more processor couple with the system memory, wherein the memory is used to store an operating system comprising of computer-executable instructions implemented by code to perform the teaching method, wherein a person ordinary skilled in the art may configure the teaching combination to be performed using the above configuration.)
Regarding claim 38 depends on claim 1, thus the rejection of claim 1 is incorporated.
Yih teaches the limitation “A non-transitory computer-readable storage medium having stored thereon computer program instructions, wherein when run on a data processing apparatus, the computer program instructions cause the data processing apparatus to perform the data processing method according to claim 1” (paragraph 84 “Computer-readable media can be any available media that can be accessed by the computer and includes both volatile and non-volatile media, removable and non-removable media. By way of example, and not limitation, computer-readable media can comprise computer storage media and communication media. Computer storage media includes both volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital video disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can be accessed by the computer.” Yih discloses computer-readable media such as non-volatile media, which suggest non-transitory computer-readable storage medium, wherein the computer-readable media may store computer program instructions to be processed by the processing unit and method as taught in claim 1.)
Regarding claim 39 depends on claim 1, thus the rejection of claim 1 is incorporated.
Yih teaches the limitation “A computer program product, comprising computer program instructions, wherein when executed on a data processing apparatus, the computer program instructions cause the data processing apparatus to perform the data processing method according to claim 1” (paragraph 66 “The claimed subject matter can be described in the general context of computer-executable instructions, such as program modules, executed by one or more components.”, paragraph 78 “Furthermore, all or portions of the claimed subject matter can be implemented as a system, method, apparatus, or article of manufacture using standard programming and/or engineering techniques to produce software, firmware, hardware or any combination thereof to control a computer to implement the disclosed subject matter.”, and paragraph 85 “With reference again to FIG. 10, the exemplary environment 1000 for implementing various aspects includes a computer 1002, the computer 1002 including a processing unit 1004, a system memory 1006 and a system bus 1008. The system bus 1008 couples system components including, but not limited to, the system memory 1006 to the processing unit 1004. The processing unit 1004 can be any of various commercially available processors”, and paragraph 89 “A number of program modules can be stored in the drives and RAM 1012, including an operating system 1030, one or more application programs 1032, other program modules 1034 and program data 1036.” Yih discloses the embodiment of the subject matter can be implemented as a system, method, apparatus which comprises of computer including one or more processor couple with the system memory, wherein the computer suggest a computer program product, and wherein the memory is used to store an operating system comprising of computer-executable instructions implemented by code to perform the teaching method, wherein a person ordinary skilled in the art may configure the teaching combination to be performed using the above configuration.)
Regarding claim 40, depends on claim 12 thus the rejection of claim 12 is incorporated. The applicant is directed to the rejection of claim 37 above, because the claim recites similar limitation and processing steps, thus the claim is rejected under the same rationale.
Regarding claim 41, depends on claim 12 thus the rejection of claim 12 is incorporated. The applicant is directed to the rejection of claim 38 above, because the claim recites similar limitation and processing steps, thus the claim is rejected under the same rationale.
Regarding claim 42, depends on claim 12 thus the rejection of claim 12 is incorporated. The applicant is directed to the rejection of claim 39 above, because the claim recites similar limitation and processing steps, thus the claim is rejected under the same rationale.
Claims 10, 13 are rejected under 35 U.S.C. 103 as being unpatentable over Yih et.al (US 20080319932 A1) in view of Oka et.al (US 20230213927 A1), further in view of Song et.al (US 20180238958 A1), further in view of Kimura et.al (US 20200226048 A1)
Regarding claim 10 depends on claim 1, thus the rejection of claim 1 is incorporated.
Yih teaches the limitation “The method according to claim 1, wherein the focus threshold includes a second focus threshold, and a number of the samples is N” (paragraph 38 “Optionally, more than one region can be of interest when there are more than two classes. In this case, a threshold can be defined for each class. Training data can then be selected for use in training the second stage classifier by employing all of the thresholds”, and paragraph 102 “For the experiments in this discussion, the training data are messages received between Jul. 1, 2005 and Nov. 30, 2005. We randomly picked 5,000 messages from each day and the total number of messages for training is 765,000”. Yih discloses there can be more than one threshold determined, suggesting a person ordinary skilled in the art can determine a second threshold, and also provide an example of training data with a predetermined range of training data, suggesting N number of samples within the claim.)
Yih teaches the limitation “displaying a mark of the second focus threshold in the sample distribution diagram on the graphical interface” (paragraph 40 “FIG. 2B shows the first stage classifier that is learned. In the illustration, a threshold of 50% is used by the first stage classifier, but in practice, any threshold can be picked. Depending on the threshold picked, the classifier will optimize on a different type of data, and thus optimize for a different false positive rate.” Yih discloses Fig. 2B displays a mark of the threshold of the classifier model, wherein the Fig. 2B represent a data distribution diagram on the graphical interface as explained above. A person ordinary skilled in the art can configure and program the display of the second threshold.)
Yih teaches the limitation “distinguishing the data display effects of the positive and negative samples based on the second focus threshold” (paragraph 40 “FIG. 2A illustrates an exemplary training data set when viewed in a two-dimensional representation where the x-axis and y-axis represent the values of the two features in the examples, a square □ is a positive example (spam) and an × is a negative example (not spam). Notice that the data is almost, but not quite linearly separable. FIG. 2B shows the first stage classifier that is learned. In the illustration, a threshold of 50% is used by the first stage classifier, but in practice, any threshold can be picked.” Yih discloses the example illustration in Fig. 2A, and Fig. 2B that represent training data distribution diagram, in which a threshold is picked and training data above and below the threshold may be identified. A person ordinary skilled in the art would have been able to distinguish the positive and negative data effects based on the threshold display, for example, according to Fig. 2B, a line of threshold is indicated, and any data points having display effect of being above the threshold may be considered as negative data, wherein data points below the threshold may be considered as positive data. A person ordinary skilled in the art can configure and program the display of the second threshold.)
Song teaches the limitation “arranging detection data of N samples in an ascending order;” (paragraph 12 “These methods rank the defect types, in inverse order to the failure-type influenced defect-type total limited yield. With this ranking, these methods can produce the most yield improvement by applying defect solutions to resolve the defect type, in order of the ranking of the defect types.” Song discloses the ranking of detected defect types in inverse order such that each group of defective items detected is ranked and place in the inverse order. Under the broadest reasonable interpretation, ranking defect-related data constitutes arranging the corresponding detection data of multiple samples in an ordered sequence. Although Song describes an inverse order, a person ordinary skilled in the art would have understood that inverse order and ascending order are merely opposite directional implementations of the same ordering operation. According, Song teaches or at least suggests arranging detection data of N samples in an ordered manner as claimed.)
Yih/Oka/Song does not teach “using a median or a mean of the detection data of the N samples as a reference focus value;”. However, Kimura teaches this limitation (paragraph 29 “For example, the storage 221 stores reference information that is referred to at the time of the determination unit 203, described later, determining an abnormality in a learning model. The reference information is at least one of statistical information (first statistical information) of a plurality of pieces of output data (first output data) that are obtained by inputting input data (first input data) that is input in the reference period in a learning model that is learned.”, and paragraph 35 “An example of the statistical information will now be described. When a plurality of pieces of output data are given as (v1, v2, . . . , vn) (n is an integer of two or more), the statistical information is an arithmetic mean u that is calculated by Formula (1)”. Kimura discloses formula to calculate an arithmetic mean based on the statistical information, wherein the statistical information can be reference information. One of ordinary skilled in the art would have been able to calculate a reference mean value based on reference information, wherein the reference information may be the abnormality degree of each training data as disclosed by Yih/Oka based on the motivation to combine the teaching below.)
Yih/Oka/Song does not teach “determining the second focus threshold based on the reference focus value and the detection data of the N samples;”. However, Kimura teaches this limitation (paragraph 60 “Furthermore, the monitoring apparatus 200 stores the threshold to be used by the determination unit 203 to determine an abnormality, in the storage 221, for example (step S104). The threshold to be stored may be a value specified by a user who uses the monitoring apparatus 200”, paragraph 63 “The calculator 202 calculates the degree of abnormality by Formula (4) described above, for example, by using the calculated statistical information (such as the arithmetic mean m) and the statistical information”, and paragraph 64 “The determination unit 203 determines whether or not the calculated degree of abnormality exceeds the threshold (step S204).” Kimura discloses the determination unit determines whether the calculated degree of abnormality exceeds the threshold, wherein the degree of abnormality is calculated based on the arithmetic mean value of reference information, wherein the reference information may comprise the abnormality degree of each training data as disclosed by Yih/Oka, in which a person ordinary skilled in the art can configure this arithmetic mean as a second threshold value to further determine abnormality in data.)
Before the effective filing date, it would have been obvious for a person ordinary skilled in the art to incorporate the teaching of a system and method that facilitates and effectuates optimizing a classifier for greater performance by Yih, the teaching of an abnormal irregularity cause identifying method and device by Oka, and the teaching of a ratio to represent proportion of a certain number of items under a condition within a group to detect abnormality in item data for better data analysis by Song, with the teaching of the monitoring system to determine abnormality and a degree of abnormality between data with the mean and abnormality degree formula by Kimura. The motivation to do so is referred to in Kimura’s disclosure (paragraph 18 “A monitoring system according to a first embodiment detects an abnormality in a learning model by storing output data of a learning model corresponding to input data that is input in a reference period or statistical information of the output data (statistics), and by comparing statistical information of output data of the learning model corresponding to later input data and the statistical information that is obtained from stored data. This allows a change in a trend of the input data to be grasped and an abnormality in the learning model to be detected with high accuracy by a simple configuration.”, and paragraph 93 “As described above, with the monitoring system according to the second embodiment, in the case where occurrence of an abnormality is determined, information indicating the cause of the abnormality and the like may be output.” Kimura discloses the benefit of the invention, which provide a monitoring system to detect abnormality in data of a learning model based on data that is stored as reference information. The system provides a mean to detect abnormality in data with high accuracy by a simple configuration. The system also provides formula to calculate mean value as well as display screen to display any information regarding abnormality. The teaching combination by Yih/Oka/Song may incorporate the teaching by Kimura for the above benefit and further improvement on abnormal data determination.)
Regarding claim 13 depends on claim 12 thus the rejection of claim 12 is incorporated. The applicant is directed to the rejection of claim 10 above, because the claim recites similar limitation and processing steps, thus the claim is rejected under the same rationale
Claims 11, 14 are rejected under 35 U.S.C. 103 as being unpatentable over Yih et.al (US 20080319932 A1) in view of Oka et.al (US 20230213927 A1), further in view of Song et.al (US 20180238958 A1), further in view of Kimura et.al (US 20200226048 A1), further in view of Velasco et.al (NPL: Thresholding using the Isodata clustering algorithm), further in view of Leys et.al (NPL: Detecting outliers: Do not use standard deviation around the mean, use absolute deviation around the median),
Regarding claim 11 depends on claim 10, thus the rejection of claim 10 is incorporated.
Kimura teaches a part of the 2nd limitation “... comparing each element in the second difference one by one, and determining a number k of which li is less than ui (li < ui), where I = 1,2,3,...,N; and updating a reference focus index to k, and updating the reference focus value to a value of k-th detection data in the detection data of the N samples arranged in sequence” (paragraph 43 “Next, the calculator 202 compares the calculated statistical information and the statistical information that is obtained from the reference information that is stored in the storage 221, and calculates a degree of abnormality indicating a degree of temporal change in the output data of the learning model”. Kimura discloses degree of abnormality/change calculated, wherein this calculation corresponds to updating the reference focus value to a value of k-th abnormal ratio data of the N samples within the claim. One of ordinary skilled in the art could assign an index to this degree using conventional coding method. A person ordinary skilled in the art would have been able to configure an index through conventional coding using program such as Python or Java, wherein the index can indicate the comparison instance between the calculated statistical information (e.g., mean of the absolute value of each difference of each training data points above/below the threshold) and the calculated statistical information of the reference information (e.g, threshold Ti, threshold Ti+1 as the method is performed repetitively), wherein these statistical information can be configured in view of the teaching combination with Velasco and Leys below.)
Kimura teaches a part of the 3rd limitation “... determining the second focus threshold based on detection data corresponding to the reference focus index in the detection data of the N samples arranged in sequence” (paragraph 60 “Furthermore, the monitoring apparatus 200 stores the threshold to be used by the determination unit 203 to determine an abnormality, in the storage 221, for example (step S104). The threshold to be stored may be a value specified by a user who uses the monitoring apparatus 200”, paragraph 63 “The calculator 202 calculates the degree of abnormality by Formula (4) described above, for example, by using the calculated statistical information (such as the arithmetic mean m) and the statistical information”, and paragraph 64 “The determination unit 203 determines whether or not the calculated degree of abnormality exceeds the threshold (step S204).” Kimura discloses the determination unit determines whether the calculated degree of abnormality exceeds the threshold, wherein the degree of abnormality is calculated based on the arithmetic mean value of reference information, wherein the reference information may comprise the abnormality degree of each training data. One of ordinary skilled in the art may configure an index corresponding to each reference information using conventional programming language such as Python or Java as described above.)
Yih/Oka/song/Kimura does not teach the 1st limitation “in step a, averaging detection data, less than or equal to the reference focus value, of the detection data of the N samples to obtain a first Mean; and averaging detection data, greater than the reference focus value, of the detection data of the N samples to obtain a second Mean”. However, Velasco teaches this limitation (Page 1 section 1 “given a threshold Ti the next threshold Ti+1 is the average of Vabove and Vbelow, where Vabove is obtained by integrating all points above Ti and Vbelow by integrating all points below Ti.”. Velasco discloses an iterative method for threshold selection with the ISODATA algorithm. The method teaches a calculation to obtain a first average of Vbelow and a second average of Vabove which indicate a first Mean and a second Mean within the claim, and a threshold Ti, wherein one of ordinary skilled in the art may configure Ti to be a threshold of arithmetic mean value with regard to the reference information as disclosed by Kimura based on the teaching combination below, such that any data point above this Ti may be integrated and average, and similarly for any data point below this mean. This configuration of above and below Ti indicate a comparison with the reference focus value similar to the claim, and the process may be performed repetitively with Ti+1.)
Yih/Oka/song/Kimura does not teach a part of the 3rd limitation “in step c, repeating the step a and the step b until a value of the reference focus index does not change before and after an update; ...” However, Velasco teaches this limitation (Page 4 section 2 “3. Recompute the means as the averages of the samples in each class, 4. If any mean has changed value, go to loop; otherwise stop.” Velasco discloses repeating the calculation of mean and if any mean has changed value, repeat the process, otherwise, stop the process, which suggest the recalculation of the threshold Ti+1 until the value does not change Ti+1 = Ti (i.e., no change occurs before and after the update). This corresponds to the claim limitation of repeating until a value does not change before and after an update, wherein one of ordinary skilled in the art can configure an index for each instance of comparison between data using conventional programming method as described above.)
Before the effective filing date, it would have been obvious for a person ordinary skilled in the art to incorporate the teaching of a system and method that facilitates and effectuates optimizing a classifier for greater performance by Yih, the teaching of an abnormal irregularity cause identifying method and device by Oka, the teaching of a ratio to represent proportion of a certain number of items under a condition within a group to detect abnormality in item data for better data analysis by Song, and the teaching of the monitoring system to determine abnormality and a degree of abnormality between data with the mean and abnormality degree formula by Kimura, with the teaching of an iterative method for threshold selection with the ISODATA algorithm by Velasco. The motivation to do so is referred to in Velasco’s disclosure (page 3 section 1 “After an initial guess is made, at each iteration we get a new threshold in the following way: ... Ti+1 is, hopefully , a better threshold than Ti for object—background discrimination”. Velasco teaches a method to provide a better threshold for classifying data, wherein the teaching combination by Yih/Oka/Kimura relies on threshold to determine abnormality degree. Therefore, the teaching combination may be further improved by incorporating the teaching by Velasco in determining a better threshold as well as the ISODATA algorithm.)
Yih/Oka/Song/Kimura/Velasco does not teach a part of the 2nd limitation “in step b, making a difference between each of the detection data of the N samples that are arranged in sequence and the first mean Meani one by one, and taking an absolute value of each difference to obtain a first mean difference DiffLowerMean, DiffLowerMean = [l1, l2, l3 ... li ..., ln]; making a difference between each of the detection data of the N samples that are arranged in sequence and the second mean Meanu one by one, and taking an absolute value of each difference to obtain a second mean difference DiffUpperMean, DiffUpperMean = [u1, u2, u3 ... ui ..., un]; ...”. However, Leys teaches this limitation (page 2 “To calculate the median, observations have to be sorted in ascending order to identify the mean rank of the statistical series and to determine the value associated with that rank ... Concretely, calculating the MAD implies the following steps: (a) the series in which the median is subtracted of each observation becomes the series of absolute values (1–7), (3–7), (3–7), (6–7), (8–7), (10–7), (10–7), and (1000–7), that is, 6, 4, 4, 1, 1, 3, 3, and 993; (b) when ranked, we obtain: 1, 1, 3, 3, 4, 4, 6, and 993; (c) and (d) the median equals 3.5 ...”. Leys discloses a MAD calculating method with an example of a series of values that demonstrate a process in which each data point (observation) is sorted in ascending order, then a difference by subtraction between the median of all data points is calculated, and an absolute value of each subtraction (difference) is then taken, and finally a mean of these absolute value is calculated. A person ordinary skilled in the art would have been able to configure the MAD calculating method with the Vbelow and Vabove and any data points within the Vbelow and Vabove as taught by Velasco based on the teaching combination and motivation below to obtain an average distance between each data value that are above and below the mean value of reference information to determine if data is widely scattered or close together, which help the user in determining if the data is abnormal or not.)
Before the effective filing date, it would have been obvious for a person ordinary skilled in the art to incorporate the teaching of a system and method that facilitates and effectuates optimizing a classifier for greater performance by Yih, the teaching of an abnormal irregularity cause identifying method and device by Oka, the teaching of a ratio to represent proportion of a certain number of items under a condition within a group to detect abnormality in item data for better data analysis by Song, the teaching of the monitoring system to determine abnormality and a degree of abnormality between data with the mean and abnormality degree formula by Kimura, and the teaching of an iterative method for threshold selection with the ISODATA algorithm by Velasco, with the teaching of absolute deviation around the median calculating method by Leys. The motivation to do so is referred to in Leys’ disclosure (page 1 “we describe a robust and easy to conduct method, for detecting outlying values in univariate statistic the Median Absolute Deviation.”, and page 2 “The median (M) is, like the mean, a measure of central tendency but offers the advantage of being very insensitive to the presence of outliers ... Moreover, the MAD is totally immune to the sample size. These two properties have led Huber (1981) to describe the MAD as the “single most useful ancillary estimate of scale” Leys discloses the benefit of the Median Absolute Deviation (MAD) method as a way of dealing with the problem of outliers, in which the calculation method provide a robust and easy calculation for detecting outlying values and offers the advantage of being very insensitive to the presence of outliers as well as immunity to the sample size. An outlier is a data point that is significantly different from other data points in a dataset, wherein detecting this outlier may benefit in detecting abnormality degree in data analyzing using the techniques within the teaching combination, thus the teaching combination may be improved upon incorporating the MAD calculation method to determine outlier of each training data with abnormality degree with regard to one or more threshold.)
Regarding claim 14, depends on claim 13 thus the rejection of claim 13 is incorporated. The applicant is directed to the rejection of claim 11 above, because the claim recites similar limitation and processing steps, thus the claim is rejected under the same rationale.
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 DUY TU DIEP whose telephone number is (703)756-1738. The examiner can normally be reached M-F 8-4:30.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Alexey Shmatov can be reached at (571) 270-3428. 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.
/DUY T DIEP/Examiner, Art Unit 2123 /ALEXEY SHMATOV/Supervisory Patent Examiner, Art Unit 2123