DETAILED ACTION
This non-final office action is responsive to application 17/745,899 as submitted 17 May 2022.
Claim status is currently pending and under examination for claims 1-15 of which independent claims are 1, 8 and 15.
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 .
Priority
Receipt is acknowledged of certified copies of papers required by 37 CFR 1.55.
The later-filed application must be an application for a patent for an invention which is also disclosed in the prior application (the parent or original nonprovisional application or provisional application). The disclosure of the invention in the parent application and in the later-filed application must be sufficient to comply with the requirements of 35 U.S.C. 112(a) or the first paragraph of pre-AIA 35 U.S.C. 112, except for the best mode requirement. See Transco Products, Inc. v. Performance Contracting, Inc., 38 F.3d 551, 32 USPQ2d 1077 (Fed. Cir. 1994).
The disclosure of the prior-filed application, Application No. IN202041014479, fails to provide adequate written description in the manner provided by 35 U.S.C. 112(a) for one or more claims of this application. Specifically, the Foreign Priority document is silent as to “token” and “sentence” as required by the independent claims. Foreign Priority document is [022] paragraphs and 3 Figures as opposed to instant application at [091] paragraphs and 8 Figures. The abbreviated priority document is not found to provide sufficient written description to support the claimed invention and therefore the application effective filing date is 12/31/2020 corresponding to PCT WO2021/198766A1. If applicant believes adequate support is provided, they must specifically point out in reply where full disclosure is made.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-15 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. In determining whether the claims are subject matter eligible, the examiner applies the guidance set forth under MPEP 2106.
Step 1: Is the claim to a process, machine, manufacture, or composition of matter? Yes—all claims fall within one of the four statutory categories: claims 1-7 are a method/process, claims 8-14 are a system/machine, and claim 15 is a non-transitory computer-readable medium/article of manufacture. Thus, the analysis should proceed per MPEP 2106.03
Step 2A, prong one: Does the claim recite an abstract idea, law of nature or natural phenomenon? Yes—the claims, under the broadest reasonable interpretation, recites an abstract idea. In this case, the claims fall within the enumerated grouping of abstract idea being “Mental Processes.” More particularly, claims recite:
“validating a report generated by a report generating model” (Mental Process, verify report produced by mental model/hypothesis)
“an actual inference corresponding to the anomaly, wherein the actual inference comprises at least one textual sentence” (Mental estimation/inference to observe an abnormality e.g. [0047] “the actual inference, in some embodiments, may be provided by a medical practitioner/doctor”)
“tokenizing: the output to generate a plurality of output tokens; and the actual inference to generate a plurality of inference tokens” (Mental Process, e.g. parsing or segmenting data)
“classifying: the plurality of output tokens into one or more predetermined categories to generate one or more sets of output tokens corresponding to the one or more predetermined categories, wherein each of the one or more sets of output tokens comprises at least one textual element” and “the plurality of inference tokens into one or more predetermined categories to generate one or more sets of inference tokens corresponding to the one or more predetermined categories, wherein each of the one or more sets of inference tokens comprises at least one textual element” (Mental Process, categorization e.g. [0046] “assigning a label (or class label)”)
“comparing each set of the one or more sets of output tokens with a corresponding set of the one or more sets of inference tokens” (Mental Process, judgment)
“assigning a match score to each set of the one or more sets of output tokens based on the comparison, wherein the match score is indicative of the degree of match between an associated set of output tokens and a corresponding set of inference tokens” (Mental evaluation e.g. [0079] “score 710 is calculated as a sum”)
“determining a combined score for the output based on the match score assigned to each set of the one or more sets of output tokens” (Mental determination, evaluation)
“validating the output based on the combined score” (Mental Process, evaluation)
Focus of the claim concerns validating reports of a model together with anomaly inference, tokenization, classification, comparison and scoring for determination. These functions do not preclude mental performance and thus may be performed in the mind of a human. For example, a radiologist ensures outlier data of a category meets certain criteria of an analysis/evaluation. In view of the foregoing, claims are considered drawn to the abstract idea of mental processes under MPEP 2106.04
Step 2A, prong two: Does the claim recite additional elements that integrate the judicial exception into a practical application? No—a practical application is not integrated by the judicial exception because the additional elements are as follows:
“receiving: an output from the report generating model, wherein the output corresponds to an anomaly detected in an image, and wherein the output comprises at least one textual sentence” MPEP 2106.05(g) adding insignificant extra-solution activity to the judicial exception, e.g. mere data gathering or selecting a particular type of data to be manipulated
The balance of the claim concerns receiving report data of image and sentence as a pre-solution activity for gathering data. This does not serve to meaningfully limit the claim or convey a concrete real-world application to raise eligibility of the claim that principally remains drawn to abstract idea. Accordingly, these additional elements do not integrate the abstract idea into a practical application.
Step 2B: Does the claim recite additional elements that amount to significantly more than the judicial exception? No—the claims do not include additional elements that amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea in to a practical application, the additional elements are identified with respect to MPEP 2106.05 and do not reveal an inventive concept. In particular, the additional elements are as follows:
“receiving: an output from the report generating model, wherein the output corresponds to an anomaly detected in an image, and wherein the output comprises at least one textual sentence” MPEP 2106.05(g) adding insignificant extra-solution activity to the judicial exception, e.g. mere data gathering or selecting a particular type of data to be manipulated. Particularly, receiving data is a well-understood, routine and conventional (WURC) activity specifically identified under MPEP 2106.05(d)(II)(i).
Significantly more is not demonstrated by the balance of the claim. Receiving report information of image and sentence data is a conventional activity carried out in the ordinary capacity of the skilled artisan. Nothing in the additional elements sheds light on watershed break-through or breathes life into the claim elevating it to patent eligibility. If the claim language provides only a result-oriented solution, with insufficient detail for how a computer accomplishes it, then the claims do contain an inventive concept. Taken alone, their additional elements do not amount to significantly more than the above-identified judicial exception (the abstract idea). Looking at the limitations as an ordered combination adds nothing that is not already present when looking at the elements taken individually. There is no indication that the combination of elements improves the functioning of a computer or improves any other technology. In view of the above, claim 1 is not patent eligible.
Independent claims 8 and 15 are rejected for substantively same rationale as claim 1, further comprising additional elements of generic computer components. This includes claim 8 “system comprising: a processor, and a memory communicatively coupled to the processor, wherein the memory stores processor-executable instructions” and claim 15 “non-transitory computer-readable medium storing computer-executable instructions.” These additional elements are recited at a high level of generality and fall under MPEP 2106.05(g) mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. More particularly, a general purpose computer does not qualify as a particular machine under MPEP 2106.05(b). Therefore, the additional elements fail to integrate the judicial exception into a practical application or amount to significantly more.
Dependent claims when analyzed as a whole are held to be patent ineligible under 35 U.S.C. 101 because the additional recited limitations fail to establish that the claims are not directed to an abstract idea, as they recite further embellishment of the judicial exception, or that they include additional elements integrating the judicial exception into a practical application or amount to significantly more.
Dependent claims 2 and 9 disclose wherein the image is an X-ray or MRI image. The limitation is considered as additional elements further embellishing the received data. Mere data gathering or selecting the type of data to be manipulated falls under insignificant extra-solution activity MPEP 2106.05(g) as it merely provides a known data environment for the same process. This can be further considered generally linking the use of the judicial exception to a particular field of use per MPEP 2106.05(g). The limitation does not satisfy the test of particular transformation or detail technical steps of the solution. Accordingly, the additional elements are insufficient to integrate the abstract idea into a practical application or amount to significantly more.
Dependent claims 3 and 10 disclose assigning labels to features and extracting features anomalous from an image using a CNN. The assigning labels limitation is considered part of the abstract idea being a mental process of judgment or opinion. The CNN is an additional element which amounts to adding insignificant extra-solution activity to the judicial exception under MPEP 2106.05(g). Particularly, said extra-solution activity is a well-understood, routine and conventional activity as is evidenced by Wang et al., “TieNet” arXiv: 1801.04334v1 at Fig 2 “Common CNN.” Therefore, additional element is not found to provide inventive concept and does not integrate the judicial exception into a practical application or amount to significantly more.
Dependent claims 4 and 11 disclose LSTM for generating text sentence. The LSTM is considered an additional element which amounts to adding insignificant extra-solution activity to the judicial exception under MPEP 2106.05(g). Particularly, said extra-solution activity is a well-understood, routine and conventional activity as is evidenced by Wang as in claim 3, where Fig 2 shows CNN-LSTM together, in combination, as an established neural network architecture for similar functionality. Therefore, the additional elements do not demonstrate inventive concept and fail to integrate the judicial exception into a practical application or amount to significantly more.
Dependent claims 5 and 12 disclose preprocessing tokens to extract relevant and inference tokens. The limitation is considered part of the abstract idea being mental process of evaluating. For example, sorting information to be used for estimation and deciding what is relevant. This could be choosing a dimension of data represented in the form of token words. There are no additional elements.
Dependent claims 6 and 13 disclose wherein preprocessing tokens of a category is performed based on a historical database. The database is considered an additional element which amounts to adding insignificant extras-solution activity to the judicial exception. Particularly, said extra-solution activity is a well-understood, routine and conventional activity under MPEP 2106.05(d)(II)(iv). As such, mere inclusion of database does not elevate the claim to eligibility demonstrating inventive concept. The additional element does not integrate the judicial exception into a practical application or amount to significantly more.
Dependent claims 7 and 14 disclose comparing relevant tokens from output tokens with relevant tokens from inference tokens. The limitation is considered part of the abstract idea being judgment. Mental comparison may be performed by a human to readily identify information. There are no additional elements.
Taken alone, their additional elements do not amount to significantly more than the above-identified judicial exception (the abstract idea). Looking at the limitations as an ordered combination adds nothing that is not already present when looking at the elements taken individually. Their collective functions merely provide conventional computer implementation.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-2, 8-9 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over:
Harzig et al., US PG Pub No 2020/0294654A1 hereinafter Harzig, in view of
Li et al., “Hybrid Retrieval-Generation Reinforced Agent for Medical Image Report Generation” hereinafter Li (arXiv: 1805.08298v2), in view of
Bustos et al., “PadChest: A Large Chest X-ray Image Dataset with Multi-label Annotated Reports” hereinafter Bustos (arXiv: 1901.07441v2) in view of
Zhang et al., “BERTscore: Evaluating Text Generation with BERT” hereinafter Zhang (arXiv: 1904.09675v3, ~9,400 citations)
With respect to claim 1, Liu teaches:
A method of validating a report generated by a report generating model {Harzig discloses [0003] “method may include… neural network generates the written report based on a sentence annotation model” and includes “verification” [0040] and/or “physician review” [0031]}, the method comprising:
receiving: an output from the report generating model, wherein the output corresponds to an anomaly detected in an image, and wherein the output comprises at least one textual sentence; and an actual inference corresponding to the anomaly, wherein the actual inference comprises at least one textual sentence {Harzig [0055] “receive, by API… written report generation unit 785 may be exported via the output” thus [0003] “providing the written reports to a treating physician”. Figs 2-3 show initial repository of images and reports as well as final generated report based on sentence annotation and abnormality, the sentences describe x-rays e.g. Figs 1, 5-6. An inference by example is [0037-38] “text report generated in the test… predict whether an input sentence is normal or abnormal” and image-based anomaly detection is [0025] “object recognition to detect shapes, textures, or other aspects of biological structures and distinguish normal structure from abnormal structures… apparent to a person of ordinary skill in the art” employing [0041] “VGG-Net… LSTM” VGG-Net is a known CNN};
However, Harzig only nominally mentions tokens [0041] and does not fairly disclose the following limitation which is disclosed by Li:
tokenizing: the output to generate a plurality of output tokens; and the actual inference to generate a plurality of inference tokens {Li [P.6 ¶3-4] “tokenizing” with [P.3 Sect.3 ¶1] “output tokens” for medical report generation, inference over tokens regards words of sentence decoder shown Fig 2 and described [P.4 ¶1,3] generative RNN};
Li is directed to automatic report generation with CNN-RNN to model image and sentence data thus being analogous. A person having ordinary skill in the art would have considered it obvious prior to the effective filing date to perform tokenization per Li in combination as applying known techniques to known methods ready for improvement to yield a predictable results and/or for a motivation of “preprocessing the reports by tokenizing… filtering tokens” [P.6 ¶3] which may further be helpful to “select sentences in the training set” [P.6 Last¶].
However, Li only nominally mentions classification [P.12 ¶] and does not fairly teach or suggest the following limitation which is met by Bustos:
classifying: the plurality of output tokens into one or more predetermined categories to generate one or more sets of output tokens corresponding to the one or more predetermined categories, wherein each of the one or more sets of output tokens comprises at least one textual element; and the plurality of inference tokens into the one or more predetermined categories to generate one or more sets of inference tokens corresponding to the one or more predetermined categories, wherein each of the one or more sets of inference tokens comprises at least one textual element {Bustos Fig 1 “multi-label text classifier” further shown Figs 6-7 emphasis multi-label, described [P.11 ¶1,3] “We define the task of annotating the medical entities as a multi-label text classification problem in which, for each sentence i, the goal is to predict… radiographic findings” and “each input instance corresponds to a sentence… for each of its N tokens” thus each token representing sentences are classified [P.10 ¶4] “classes to be inferred” output of multi-label classification again shown Figs 6-7. See also [P.20 ¶1] “categorize radiographic findings” defined labels at [P.26-29] App. 1.1.1 - 1.1.2}; and
Bustos is directed to validated classifier and recurrent models for sentence and image data relevant to reports thus being analogous. A person having ordinary skill in the art would have considered it obvious prior to the effective filing date to perform multi-label classifying per Bustos in combination for a motivation “The goal was to produce a large dataset of annotated reports to be used as output of an image classifier trained with the x-rays” [P.10 Last¶] and/or “To help advance automatic clinical text annotations… help automatically label other large-scale x-ray repositories” [P.3 Last2¶].
However, Bustos in combination does not teach or suggest the following limitations which are met by Zhang:
comparing each set of the one or more sets of output tokens with a corresponding set of the one or more sets of inference tokens {Zhang [P.4 ¶2] “cosine similarity of a reference token xi and a candidate token x’j” illustrated Fig 1 pairwise cosine similarity, similarity is comparing the paired tokens};
assigning a match score to each set of the one or more sets of output tokens based on the comparison, wherein the match score is indicative of the degree of match between an associated set of output tokens and a corresponding set of inference tokens {Zhang discloses [P.4 ¶3] “BERTscore: The complete score matches each token in x to a token in x’ …We use greedy matching to maximize the matching similarity score, where each token is matched to the most similar token in the other sentence” shown Fig 1, see also [P.16,15] “BERTscore assigns high similarity… BERTscore assigns low similarity”};
determining a combined score for the output based on the match score assigned to each set of the one or more sets of output tokens {Zhang discloses [P.1 ¶2] “BERTscore computes the similarity of two sentences as a sum of cosine similarities” sum is sigma denoted ∑ importance weighting [P.4 ¶4] and/or shown as + in Fig 1 right. Limitation read in light of instant spec [079] “combined match score 710 is calculated as a sum of individual match scores”}; and
validating the output based on the combined score {Zhang [P.8 ¶2] “BERTscore is a good fit for using during validation” with “commonly used test and validations sets” e.g. image captioning [P.6 Last¶] and/or [P.23 ¶2] outputs used for validation}.
Zhang is directed to models for text generation evaluation thus being analogous. A person having ordinary skill in the art would have considered it obvious prior to the effective filing date to employ BERTscore per Zhang in combination to arrive at the invention as claimed for a motivation “BERTscore is purposely designed to be simple, task agnostic, and easy to use” [P.9 ¶2] and because “BERTscore correlates better human judgments and provides stronger model selection performance than existing metrics” such that “BERTscore addresses two common pitfalls in n-gram-based metrics. First, such methods often fail to robustly match paraphrases… Second, n-gram models fail to capture distant dependencies and penalize semantically-critical ordering changes” [P.1 ¶1,4]. Also, [P.8 ¶2] “BERTscore is relatively fast” thus reducing time of processing.
With respect to claim 2, the combination of Harzig, Li, Bustos and Zhang teaches the method as claimed in claim 1, wherein
the image is one of an X-ray or a Magnetic Resonance Image (MRI) scan image {Harzig [0040] “dataset comprises 112,120 chest X-ray images” Figs 5-6 illustrate}.
With respect to claim 8, the rejection of claim 1 is incorporated. The difference in scope being a system comprising processor and memory for performing limitations of method claim 1. Harzig discloses [0001] “systems and methods for computer-aided generation of descriptions of abnormalities in new medical images” and [0005] “processor communicatively coupled to the memory” shown Fig 7. The remainder of this claim is rejected for the same rationale as claim 1.
With respect to claim 9, the combination of Harzig, Li, Bustos and Zhang teaches the system as claimed in claim 8, and further teaches the limitation of claim 2. Therefore, the rejection of claim 2 is applied to claim 9.
With respect to claim 15, the rejection of claim 1 is incorporated. The difference in scope being a non-transitory computer-readable medium storing computer-executable instruction for generating recommendation for a user configured to perform limitations of method claim 1. Harzig discloses [0004] “non-transitory computer readable medium encoded with instructions” similar at [0051-52] as well as providing suggesting [0020,22] or recommending [0031]. The remainder of this claim is rejected for the same rationale as claim 1.
Claims 3-4 and 10-11 are rejected under 35 U.S.C. 103 as being unpatentable over Harzig, Li, Bustos and Zhang in view of McKinney et al., US PG Pub No 2021/0065859A1 hereinafter McKinney.
With respect to claim 3, the combination of Harzig, Li, Bustos and Zhang teaches the method as claimed in claim 1, further comprising:
extracting one or more features indicative of one or more anomalies from the image using a first Al model, wherein the first Al model is a Convolutional Neural Network (CNN) {Harzig Fig 2:240 [0025] “image feature extraction model 240 on a neural network to train the image feature extraction model 240 to extract feature representations of image content… distinguish normal structures from abnormal structures” particularly [0041] “image feature learning was based on VGG-Net” VGG-Net is a known CNN}; and
However, Harzig does not explicitly disclose assigning labels which is met by McKinney:
assigning a label to each of the one or more features {McKinney discloses per [0019,26] “assign a structured label to a further input medical image, e.g. a chest X-ray” illustrated Fig 9 CNN feature extractor for x-ray image, similar at [0051] “images which have assigned structured labels automatically” Fig 8}.
McKinney is directed to CNN models for computer vision based reports thus being analogous. A person having ordinary skill in the art would have considered it obvious prior to the effective filing date to assign labels per McKinney in combination to arrive at the invention as claimed for a motivation to “generate the structured labels for medical images automatically and without requiring vast amounts of time from trained radiologists” [0051,62].
With respect to claim 4, the combination of Harzig, Li, Bustos, Zhang and McKinney teaches the method as claimed in claim 3, further comprising:
generating at least one textual sentence based on the label assigned to each of the one or more features using a second Al model, whereon the second Al model is a Long Short-Term Memory (LSTM) model {Harzig [0028] “generated sentence” where [0041] “The text generation was based on a Hierarchical LSTM and further integrated with sentence annotations” annotation is assigned label e.g. [0038] “automatically label sentences”}.
With respect to claim 10, the combination of Harzig, Li, Bustos and Zhang teaches the system as claimed in claim 8, and further combination with McKinney teaches the limitation of claim 3. Therefore, the rejection of claim 3 is applied to claim 10.
With respect to claim 11, the combination of Harzig, Li, Bustos, Zhang and McKinney teaches the system as claimed in claim 10, and further teaches the limitation of claim 4. Therefore, the rejection of claim 4 is applied to claim 11.
Claims 5-6 and 12-13 are rejected under 35 U.S.C. 103 as being unpatentable over Harzig, Li, Bustos and Zhang in view of Sun et al., “VideoBERT: A Joint Model for Video and Language Representation Learning” hereinafter Sun (arXiv: 1904.01766v2).
With respect to claim 5, the combination of Harzig, Li, Bustos and Zhang teaches the method as claimed in claim 1. Sun teaches further comprising
preprocessing: the plurality of output tokens to extract relevant tokens from each of the one or more sets of output tokens; and the plurality of inference tokens to extract relevant tokens from each of the one or more sets of inference tokens {Sun [P.5 Sect4.2] “preprocessing steps from BERT and tokenize the text” more particularly [P.7 Sect4.6] “VideoBERT when used as a feature extractor… extract the features for the video tokens and the masked out of text tokens” illustratively Fig 3 token masking is the technique resolving relevance of tokens for inference, e.g. [P.6 ¶1] “extract the verb and noun labels from the tokens predicted in the first and second masked slots”}.
Sun is directed to text sentence captioning for image/video with generative modeling thus being analogous. A person having ordinary skill in the art would have considered it obvious prior to the effective filing date to preprocess per Sun in combination to arrive at the invention as claimed for a motivation “our main contribution in this paper is a simple way to learn high level video representations that capture semantically meaningful and temporally long-range structure” [P.2 ¶4] and which “extends naturally to sequences of linguistic and visual tokens, applying a next sentence prediction” [P.4 ¶3].
With respect to claim 6, the combination of Harzig, Li, Bustos and Zhang teaches the method as claimed in claim 5, wherein
the preprocessing of a set of output tokens or a set of inference tokens associated with a category is performed based on a historical database corresponding to the category {Sun [P.3 Last2¶] “classification token [CLS]” classification class is category associated with tokens Fig 3, the historical database may include e.g. Wikipedia [P.5 ¶1]}. It would have been obvious to employ known databases for categories of token data as obvious to try in choosing from finite predictable solutions with a reasonable expectation of success.
With respect to claim 12, the combination of Harzig, Li, Bustos and Zhang teaches the system as claimed in claim 8, and further combination with Sun teaches the limitation of claim 5. Therefore, the rejection of claim 5 is applied to claim 12.
With respect to claim 13, the combination of Harzig, Li, Bustos, Zhang and Sun teaches the system as claimed in claim 12, and further teaches the limitation of claim 6. Therefore, the rejection of claim 6 is applied to claim 13.
Claims 7 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Harzig, Li, Bustos, Zhang and Sun in view of Clark et al., “What Does BERT Look At? An Analysis of BERT’s Attention” hereinafter Clark (arXiv: 1906.04341v1).
With respect to claim 7, the combination of Harzig, Li, Bustos, Zhang and Sun teaches the method as claimed in claim 5, wherein comparing each set of the one or more sets of output tokens with a corresponding set of the one or more sets of inference tokens {Zhang, claim 1}. Clark teaches comprises:
comparing the relevant tokens from each of the one or more sets of output tokens with the relevant tokens from a corresponding set of the one or more sets of inference tokens {Clark discloses [P.8 Sect.6] Equation is comparing tokens by js-divergence measure for attention heads of BERT}.
Clark is directed to tokenization for modeling sentences thus being analogous. A person having ordinary skill in the art would have considered it obvious prior to the effective filing date employ the token comparison of Clark in combination to arrive at the invention as claimed as applying known techniques to know methods to yield predictable results as it describes how attention mechanisms function with further insight into similarities among attention heads in the same layer and further notes a motivation “our motivation for looking at attention is… we are seeking to understand information learned by the models… important finding from an analysis perspective” [P.9 ¶6].
With respect to claim 14, the combination of Harzig, Li, Bustos, Zhang and Sun teaches the system as claimed in claim 12, and further combination with Clark teaches the limitation of claim 7. Therefore, the rejection of claim 7 is applied to claim 14.
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Liu et al., “Clinically Accurate Chest X-ray Report Generation” arXiv: 1904.02633v2 Fig 2
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Chase P Hinckley whose telephone number is (571)272-7935. The examiner can normally be reached M-F 9:00 - 5:00.
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, Miranda M. Huang can be reached at 571-270-7092. 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.
/CHASE P. HINCKLEY/Examiner, Art Unit 2124