DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
This non-final office action is in response to the applicant’s communication received on 04/21/2026 (“Amendment”).
Priority
Instant application claims priority to a provisional application 63/608,745 filed on 27-Dec-2023. Upon analysis of the provisional application, instant claims are not supported by the provisional application. For example, the provisional applicant describes at high level of using machine learning model in evaluation, however, does not describe in details of machine leaning model or the instructing a downstream system/causing the repair entity to initiate a repair process/processing the damage estimate using a downstream system in the independent claim(s). The examiner further finds that the provisional application does not provide support for the particular details in the dependent claim(s).
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 04/21/2026 has been entered.
Status of Claims
Claims 1, 9, and 17 have been amended.
Claims 7, 11, 12, and 16 had/have been canceled.
Claims 23-24 have been newly added.
Claims 1-6, 8-10, 13-15, and 17-24 are pending.
Specification/Claim Objection
The specification lacks antecedent basis for “embedding space”, “first token type” and “second token type”.
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 1-6, 8-10, 13-15, and 17-24 are rejected under 35 U.S.C. 101 because the claimed invention is directed to abstract idea without significantly more.
MPEP 2106 provides step(s) in determining eligibility under 35 U.S.C. § 101. Specifically, it must be determined whether the claim is directed to one of the four statutory categories of invention, i.e., process, machine, manufacture, or composition of matter. If the claim does fall within one of the statutory categories, it must then be determined whether the claim is directed to a judicial exception (i.e., law of nature, natural phenomenon, and abstract idea), and if so, it must additionally be determined whether the claim is a patent-eligible application of the exception. If an abstract idea is present in the claim, any additional elements in the claim must integrate the judicial exception into a practical application. If not, the inquiry continues to see whether any element or combination of elements in the claim must be sufficient to ensure that the claim amounts to significantly more than the abstract idea itself. Examples of abstract ideas include mathematical concepts, mental processes, and certain methods of organizing human activities.
Under Step 1, claims 1-6, 8, and 21-23 are directed to a processor-based system, claims 9, 10, 13-15, and 24 directed to a method (i.e. process), while claims 17-20 are directed to a non-transitory computer-readable storage medium. Thus, the claimed inventions are directed towards one of the four statutory categories under 35 USC § 101. Nevertheless, the claims also fall within the judicial exception of an abstract idea without significantly more.
Step 2A, 1st prong:
Claim 9 recites: A computer-implemented method, comprising:
a) receiving damage data associated with a vehicle, wherein the damage data comprises at least one of image data, telematics data, and a text-based description of damage to the vehicle;
b) generating, based on the damage data and using a first vehicle damage encoder, a first embedding token of a first token type in an embedding space representing the damage data;
c) receiving damage estimate data associated with the vehicle, for a damage estimate determined by a repair entity, wherein the damage estimate data comprises at least one of a listing of parts or a listing of services identified in the damage estimate associated with the vehicle;
d) generating, based on the damage estimate data and using a second damage estimate encoder, a second embedding token of a second token type in the embedding space representing the damage estimate data;
e) providing as input to a machine learning model: the first embedding token representing the damage data; and the second embedding token representing the damage estimate;
f) determining, based on an output of the machine learning model, a score associated with the damage estimate; and
g) based on the score associated with the damage estimate, transmitting instructions to cause the repair entity to initiate a repair process on the vehicle.
(Bold emphasis added on the additional element(s))
Under the broadest reasonable interpretation, the claim recites an operation that receives damage data (i.e., at least one of image data, telematic data, and a text-based description of damage to a vehicle) and damage estimate data (i.e., at least one of a listing of parts or a listing of services identified in the damage estimate for the vehicle and determined by a repair shop), both associated with a vehicle, as in steps a) and c); generating first and second token(s), i.e., information as in steps b) and d); e) providing the first and send token(s) to a model, f) determining based on an output of the model a score associated with the damage estimate; and g) transmitting instructions for the purpose of the repair shop (i.e., approval of the repair) to initiate a repair process on the vehicle based on the score. As such, the claim recites an abstract idea, i.e., certain methods of organizing human activities such as insurance and/or business relations/economic practices of evaluating repair estimates in order to process the repair. The process of determination also falls within the mental process that can be done with a pen and paper. In regard to the multimodal learning model using encoder that converts inputs into a fixed-length representation (i.e., tokens of different types) involves mathematical relations and calculations, including weighting functions, vector representation, etc., which constitute mathematical concepts. As such, the claim recites abstract idea.
The other independent claims, i.e., claims 1 and 17, are significantly similar to claim 9 with exception to description of the machine learning model being a multimodal and that the repair entity is replaced by downstream system. As such, claims 1 and 17 also recite abstract idea.
Under the Step 2A (prong 2), this judicial exception is not integrated into a practical application. Specifically, the additional elements in the claim(s), i.e., a computer system comprising a processor and memory storing computer-executable instructions, encoder, downstream system, and multimodal/machine learning are recited at a high-level generality such that it amounts to no more than mere instructions to implement the abstract idea and/or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).
Under Step 2B, examiners should evaluate additional elements individually and in combination to determine whether they provide an inventive concept (i.e. whether the additional elements amount to significantly more than the exception itself). Here, the claim(s) do not include additional elements that are sufficient to amount to significantly more than the judicial exception. Specifically, the claim(s) as a whole, taken individually and in combination, do not provide an inventive concept. As explained above with respect to the integration of the abstract idea into a practical application, the additional elements used to perform the claimed judicial exception amount to no more than mere instructions to implement the abstract idea and/or merely uses a computer as a tool to perform an abstract idea. Mere instructions to implement the abstract idea on a computer, or merely using the computer as a tool to perform an abstract idea to apply the exception using a generic computer component cannot provide an inventive concept. Looking at the limitations as a combination adds nothing that is not already present when looking at the elements taken individually. There is no indication that the combination of the elements improves the functioning of the recited computer system or the system’s component(s) (i.e., a processor, memory storing instructions, encoder, and/or machine learning) that are responsible for performing the step(s).
The dependent claims 2-6, 8, 10, 13-15, and 18-24 expand further on the abstract idea related to certain method of organizing human activities, mental activities, and mathematical concept without reciting further additional elements as the claim(s) other than the additional element(s) that have been identified above.
For these reasons, the claims are rejected under 101.
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.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claim(s) 1-4, 8-10, 13, 17-18, 21, and 23-24 is/are rejected under 35 U.S.C. 103 as being unpatentable over US Patent No. 11,386,543 (“Ranca”) in view of US Patent Publication No. 20240370718 (“Panagopoulou”).
Per claim 1, Ranca teaches a system, comprising: a processor; and memory storing computer-executable instructions that, when executed by the processor, cause the computer system to perform operations (C5 L55-60, computer-implemented system, computer-programme product) comprising:
receiving vehicle damage data associated with a vehicle, the vehicle damage data including at least one of: image data of the vehicle; a text description of damage to the vehicle; or telematics data associated with the vehicle (C3 L24-27, receiving vehicle input data, comprising at least a plurality of images some comprising image data of damage to the vehicle; C8 L26-31, providing photos; C8 L28-42, injured person, crash details and photos of the damage are provided by the party and the computer system is configured to display a dynamic live estimate of the repair cost of the damage to the vehicle; C10 L1-45, use of photos and information to feed the trained models to predict damage; C11 L6-64, production of an output estimation of damage and the predicted repair cost for the vehicle; C17 L1-9; C24 L7-24; C24 L45-48, accident details and photographic images of any damage are provided by the party to the insurer);
receiving a vehicle damage estimate determined by a repair entity, the vehicle damage estimate associated with the vehicle, wherein the vehicle damage estimate comprises at least on of a listing of parts or a listing of services (C24 L7-24, the estimate from the repair shop is sent to the insurer for authorization by the insurer … estimated repair cost along with supporting details, repair labour proposed, parts required, materials required; C25 L12-17, shop prepares a repair estimate for the damage to the vehicle and reports this estimate to the insurer);
providing as input to multimodal machine learning model the vehicle damage data and the vehicle damage estimate as input to a multimodal machine learning model (Figure 24; C32 L51-52, image classifier; C33 L14-30, NLP model)(multimodal machine learning model is interpreted as model that processes information from multiple data sources, as in image and text; C10 L1-45, use of photos and information to feed the trained models to predict damage; C11/L6-L12/5, production of an output estimation of damage and the predicted repair cost for the vehicle using various inputs and models; C25 L41-48, comparing independent repair estimate to the provided repair estimate from the garage; C26 L16-34, inputs include the information provided to the insurer by the client and the repair estimate from the garage; C28 L33-37, the core assessment module is used to create an independent estimate of the repair work needed for the damaged vehicle based on the provided photograph that can then be used to verify the provided order sheet and body shop estimate; C29 L29-28, checks whether the body shop estimate list repair work that is excessive compared to the repair work predicted to be necessary from the input data in the generated independent estimate of the repair work needed for the damaged vehicle; C31 L9-25, image input and the vehicle repair estimate data is analysed using a decision classifier);
determining a score associated with the vehicle damage estimate, based at least in part on an output of the multimodal machine learning model (C9 L1-6, total loss calculation is made by a trained model; C21 L30-35; C22 LC30 L3-45; C28 L42-53, leakage check; C30 L11-15, level of leakage); and
instructing a downstream system to process the vehicle damage estimate, based on the damage estimate score associated with the vehicle damage estimate (C8 L62-67, total loss, repair or cash settlement; C9 L36-40, submit the information to a repair shop or body shop and to authorize the repair of the vehicle; C23 L42-54, determining whether claim input data can be approved or is anomalous or whether further assessment may be required … whether vehicle repair estimates are acceptable and legitimate; C30 L3-45).
Ranca does not particularly teach generating, based on the vehicle damage data and using a first vehicle damage encoder, a first embedding token of a first token type in an embedding space representing the vehicle damage data; generating, based on the vehicle damage estimate and using a second damage estimate encoder, a second embedding token of a second token type in the embedding space representing the vehicle damage estimate; and that the first embedding token and the second embedding token are provided as input to the model.
Panagopoulou, however, teaches generating, based on the first data and using a first vehicle damage encoder, a first embedding token of a first token type in an embedding space representing the first data; generating, based on the second data and using a second damage estimate encoder, a second embedding token of a second token type in the embedding space representing the second data; and that the first embedding token and the second embedding token are provided as input to a model (Abstract: encoding by a first multimodal encoder adapted for the first modality, the first input of the first modality into a first encoded representation, encoding, by a second multimodal encoder adapted for the second modality, the second input of the second modality into a second encoded representation, and generating by a neural network output based on the inputs).
As Ranca teaches receiving vehicle damage data and vehicle damage estimate and providing the two as input to a machine learning model used in analysis of the estimate as described above, it would have been obvious to one of ordinary skill in the art before the effective filing of instant claim to combine the technique of utilizing encoders as taught by Panagopoulou and apply the technique to the vehicle damage data and the vehicle damage estimate in order for multi-modal models for responding to instructions ([0004]).
As per claim 2, Ranca/Panagopoulou further teaches: providing the image data as input to an image-based damage prediction model; and determining, based on an output of the image-based damage prediction model, at least one of: a part identifier associated with a potentially damaged part on the vehicle; a severity value of a potentially damaged part on the vehicle; or a confidence value associated with a potentially damaged part damaged part on the vehicle (Ranca: C4 L2-16, use of classifier; C33 L62, to generate a detailed list of parts and labour operations based on the independent generated damage estimate output from the image classifier in order to classify whether the claim input in whole or in part is correct; C34 L60-65, specifying the exact part number).
As per claim 3, Ranca/Panagopoulou further teaches wherein the input to the multimodal machine learning model further comprises at least one of: a listing of parts associated with the vehicle damage estimate; and a listing of expenses associated with the vehicle damage estimate (Ranca: C2 L28-46, specific parts required; C26 L19-34, a list of the parts needing replacement … the prices for any materials listed and the prices for any labour listed; C33 L4-15, list of parts and labour operations and provided to a model). The applicant is reminded that the description of data, i.e., description of input, is non-functional descriptive material.
As per claim 4, Ranca/Panagopoulou further teaches wherein the vehicle damage estimate comprises a current version of the vehicle damage estimate determined by the repair entity, and wherein the input to the multimodal machine learning model is indicative of at least one of: an entity attribute associated with the repair entity; a version number associated with the current version of the vehicle damage estimate; or a previous vehicle damage estimate from a previous version of the damage estimate determined by the repair entity (Ranca: C27 L27-38, updated claim; C31 L51-60, change their propose repair estimate; C32 L30-33, resubmission). The applicant is reminded that the description of data, i.e., description of input and description of damage estimate, is non-functional descriptive material.
As per claim 8, Ranca/Panagopoulou further teaches wherein instructing the downstream system comprises at least one of: providing the vehicle damage estimate to an estimate analysis process for processing; transmitting a request for an update to the vehicle damage estimate to a system associated with the repair entity; or initiating an automated process to instruct the repair entity to repair the vehicle, based on the damage estimate score associated with the vehicle damage estimate (Ranca: C23 L55-60, the repair operations can be approved on behalf of the insurer; C30 L 30-37, resubmit).
Per claim 9, Ranca teaches A computer-implemented method, comprising:
receiving damage data associated with a vehicle, wherein the damage data comprises at least one of image data, telematics data, and a text-based description of damage to the vehicle (C3 L24-27, receiving vehicle input data, comprising at least a plurality of images some comprising image data of damage to the vehicle; C8 L26-31, providing photos; C8 L28-42, injured person, crash details and photos of the damage are provided by the party and the computer system is configured to display a dynamic live estimate of the repair cost of the damage to the vehicle; C10 L1-45, use of photos and information to feed the trained models to predict damage; C11 L6-64, production of an output estimation of damage and the predicted repair cost for the vehicle; C17 L1-9; C24 L7-24; C24 L45-48, accident details and photographic images of any damage are provided by the party to the insurer);
receiving damage estimate data associated with the vehicle, for a damage estimate determined by a repair entity, wherein the damage estimate data comprises at least one of a listing of parts or a listing of services identified in the damage estimate associated with the vehicle (C24 L7-24, the estimate from the repair shop is sent to the insurer for authorization by the insurer … estimated repair cost along with supporting details, repair labour proposed, parts required, materials required; C25 L12-17, shop prepares a repair estimate for the damage to the vehicle and reports this estimate to the insurer);
providing the damage data and the damage estimate data as input to a machine learning model (Figure 24; C10 L1-45, use of photos and information to feed the trained models to predict damage; C11/L6-L12/5, production of an output estimation of damage and the predicted repair cost for the vehicle using various inputs and models; C25 L41-48, comparing independent repair estimate to the provided repair estimate from the garage; C26 L16-34, inputs include the information provided to the insurer by the client and the repair estimate from the garage; C28 L33-37, the core assessment module is used to create an independent estimate of the repair work needed for the damaged vehicle based on the provided photograph that can then be used to verify the provided order sheet and body shop estimate; C29 L29-28, checks whether the body shop estimate list repair work that is excessive compared to the repair work predicted to be necessary from the input data in the generated independent estimate of the repair work needed for the damaged vehicle; C31 L9-25, image input and the vehicle repair estimate data is analysed using a decision classifier);
determining, based on an output of the machine learning model, a score associated with the damage estimate (C10 L1-45, use of photos and information to feed the trained models to predict damage; C11/L6-L12/5, production of an output estimation of damage and the predicted repair cost for the vehicle using various inputs and models; C25 L41-48, comparing independent repair estimate to the provided repair estimate from the garage; C26 L16-34, inputs include the information provided to the insurer by the client and the repair estimate from the garage; C28 L33-37, the core assessment module is used to create an independent estimate of the repair work needed for the damaged vehicle based on the provided photograph that can then be used to verify the provided order sheet and body shop estimate; C29 L29-28, checks whether the body shop estimate list repair work that is excessive compared to the repair work predicted to be necessary from the input data in the generated independent estimate of the repair work needed for the damaged vehicle; C31 L9-25, image input and the vehicle repair estimate data is analysed using a decision classifier; C9 L1-6, total loss calculation is made by a trained model; C21 L30-35; C30 L3-45; C28 L42-53, leakage check; C30 L11-15, level of leakage); and
based on the score associated with the damage estimate, transmitting instructions to cause the repair entity to initiate a repair process on the vehicle (C8 L62-67, total loss, repair or cash settlement; C9 L36-40, submit the information to a repair shop or body shop and to authorize the repair of the vehicle; C23 L42-54, determining whether claim input data can be approved or is anomalous or whether further assessment may be required … whether vehicle repair estimates are acceptable and legitimate; C30 L3-45).
Ranca does not particularly teach generating, based on the vehicle damage data and using a first vehicle damage encoder, a first embedding token of a first token type in an embedding space representing the vehicle damage data; generating, based on the vehicle damage estimate and using a second damage estimate encoder, a second embedding token of a second token type in the embedding space representing the vehicle damage estimate; and that the first embedding token and the second embedding token are provided as input to the model.
Panagopoulou, however, teaches generating, based on the first data and using a first vehicle damage encoder, a first embedding token of a first token type in an embedding space representing the first data; generating, based on the second data and using a second damage estimate encoder, a second embedding token of a second token type in the embedding space representing the second data; and that the first embedding token and the second embedding token are provided as input to a model (Abstract: encoding by a first multimodal encoder adapted for the first modality, the first input of the first modality into a first encoded representation, encoding, by a second multimodal encoder adapted for the second modality, the second input of the second modality into a second encoded representation, and generating by a neural network output based on the inputs).
As Ranca teaches receiving vehicle damage data and vehicle damage estimate and providing the two as input to a machine learning model used in analysis of the estimate as described above, it would have been obvious to one of ordinary skill in the art before the effective filing of instant claim to combine the technique of utilizing encoders as taught by Panagopoulou and apply the technique to the vehicle damage data and the vehicle damage estimate in order for multi-modal models for responding to instructions ([0004]).
As per claim 10, Ranca/Panagopoulou further teaches wherein the damage data includes image data of the vehicle, and wherein the method further comprises: providing the image data as input to a second machine learning model; determining, based on an output of the second machine learning model, a part identifier associated with a potentially damaged part on the vehicle, a severity value of the potentially damaged part, and a confidence value of the potentially damaged part; and providing the part identifier, the severity value, and the confidence value as input to the machine learning model (Ranca: C4 L2-16, use of classifier; C10 L21-42, confidence value; C15 L11-23; C17 L19-25, associated confidence value … the output is fed into each of a set of further assessment models; C32 L13-17, severity of the damage; C33 L62, to generate a detailed list of parts and labour operations based on the independent generated damage estimate output from the image classifier in order to classify whether the claim input in whole or in part is correct; C34 L30-35, severity damage to a part; C34 L60-65, specifying the exact part number).
As per claim 13, Ranca/Panagopoulou further teaches wherein the damage estimate comprises a current version of the damage estimate determined by the repair entity, and wherein the input to the machine learning model comprises at least one of: an input based on an entity attribute associated with the repair entity; an input based on a version number associated with the current version of the damage estimate; or an input based on previous damage estimate data from a previous version of the damage estimate determined by the repair entity (Ranka: C4 L2-16, use of classifier; C10 L21-42, confidence value; C15 L11-23; C17 L19-25, associated confidence value … the output is fed into each of a set of further assessment models; C32 L13-17, severity of the damage; C33 L62, to generate a detailed list of parts and labour operations based on the independent generated damage estimate output from the image classifier in order to classify whether the claim input in whole or in part is correct; C34 L30-35, severity damage to a part; C34 L60-65, specifying the exact part number). The applicant is reminded that the description of data, i.e., description of input and description of damage estimate, is non-functional descriptive material.
Per claim 17, Ranca teaches one or more non-transitory computer-readable media storing instructions executable by a processor, wherein the instructions, when executed by the processor, cause the processor to perform operations comprising:
receiving damage data associated with a vehicle, wherein the damage data comprises at least one of image data, telematics data, and a text-based description of damage to the vehicle (C3 L24-27, receiving vehicle input data, comprising at least a plurality of images some comprising image data of damage to the vehicle; C8 L26-31, providing photos; C8 L28-42, injured person, crash details and photos of the damage are provided by the party and the computer system is configured to display a dynamic live estimate of the repair cost of the damage to the vehicle; C10 L1-45, use of photos and information to feed the trained models to predict damage; C11 L6-64, production of an output estimation of damage and the predicted repair cost for the vehicle; C17 L1-9; C24 L7-24; C24 L45-48, accident details and photographic images of any damage are provided by the party to the insurer);
receiving damage estimate data associated with the vehicle, for a damage estimate determined by a repair entity, wherein the damage estimate data comprises at least one of a listing of parts or a listing of services identified in the damage estimate associated with the vehicle (C24 L7-24, the estimate from the repair shop is sent to the insurer for authorization by the insurer … estimated repair cost along with supporting details, repair labour proposed, parts required, materials required; C25 L12-17, shop prepares a repair estimate for the damage to the vehicle and reports this estimate to the insurer);
providing the damage data and the damage estimate data as input to a multimodal machine learning model (C32 L51-52, image classifier; C33 L14-30, NLP model) (C10 L1-45, use of photos and information to feed the trained models to predict damage; C11/L6-L12/5, production of an output estimation of damage and the predicted repair cost for the vehicle using various inputs and models; C25 L41-48, comparing independent repair estimate to the provided repair estimate from the garage; C26 L16-34, inputs include the information provided to the insurer by the client and the repair estimate from the garage; C28 L33-37, the core assessment module is used to create an independent estimate of the repair work needed for the damaged vehicle based on the provided photograph that can then be used to verify the provided order sheet and body shop estimate; C29 L29-28, checks whether the body shop estimate list repair work that is excessive compared to the repair work predicted to be necessary from the input data in the generated independent estimate of the repair work needed for the damaged vehicle; C31 L9-25, image input and the vehicle repair estimate data is analysed using a decision classifier);
determining, based on an output of the multimodal machine learning model, a damage estimate score associated with the damage estimate (C9 L1-6, total loss calculation is made by a trained model; C21 L30-35; C30 L3-45); and
processing the damage estimate using a downstream system, based on the damage estimate score (C8 L62-67, total loss, repair or cash settlement; C9 L36-40, submit the information to a repair shop or body shop and to authorize the repair of the vehicle; C23 L42-54, determining whether claim input data can be approved or is anomalous or whether further assessment may be required … whether vehicle repair estimates are acceptable and legitimate; C30 L3-45).
Ranca does not particularly teach generating, based on the vehicle damage data and using a first vehicle damage encoder, a first embedding token of a first token type in an embedding space representing the vehicle damage data; generating, based on the vehicle damage estimate and using a second damage estimate encoder, a second embedding token of a second token type in the embedding space representing the vehicle damage estimate; and that the first embedding token and the second embedding token are provided as input to the model.
Panagopoulou, however, teaches generating, based on the first data and using a first vehicle damage encoder, a first embedding token of a first token type in an embedding space representing the first data; generating, based on the second data and using a second damage estimate encoder, a second embedding token of a second token type in the embedding space representing the second data; and that the first embedding token and the second embedding token are provided as input to a model (Abstract: encoding by a first multimodal encoder adapted for the first modality, the first input of the first modality into a first encoded representation, encoding, by a second multimodal encoder adapted for the second modality, the second input of the second modality into a second encoded representation, and generating by a neural network output based on the inputs).
As Ranca teaches receiving vehicle damage data and vehicle damage estimate and providing the two as input to a machine learning model used in analysis of the estimate as described above, it would have been obvious to one of ordinary skill in the art before the effective filing of instant claim to combine the technique of utilizing encoders as taught by Panagopoulou and apply the technique to the vehicle damage data and the vehicle damage estimate in order for multi-modal models for responding to instructions ([0004]).
As per claim 18, Ranca/Panagopoulou further teaches wherein the damage data includes image data of the vehicle, and wherein the operations further comprise: providing the image data as input to an image-based damage prediction model; determining, based on an output of the image-based damage prediction model, a part identifier associated with a potentially damaged part on the vehicle, a severity value of the potentially damaged part, and a confidence value of the potentially damaged part; and providing the part identifier, the severity value, and the confidence value as input to the multimodal machine learning model (Ranca: C4 L2-16, use of classifier; C10 L21-42, confidence value; C15 L11-23; C17 L19-25, associated confidence value … the output is fed into each of a set of further assessment models; C32 L13-17, severity of the damage; C33 L62, to generate a detailed list of parts and labour operations based on the independent generated damage estimate output from the image classifier in order to classify whether the claim input in whole or in part is correct; C34 L30-35, severity damage to a part; C34 L60-65, specifying the exact part number).
As per claim 21, Ranca/Panagopoulou further teaches wherein the multimodal machine learning model is trained to receive as input: a first fixed-size encoding, at a first predetermined location in the input, representing the vehicle damage data; a second fixed-size encoding, at a second predetermined location in the input, representing the vehicle damage estimate; and a third fixed-size encoding, at a third predetermined location in the input, representing a repair entity profile (by Panagopoulou disclosing a multi-modal machine learning process, i.e., model, that is configured to receive encoded data, Hyatt discloses multi-modal machine learning model that is capable of receiving any data that has been encoded). Furthermore, the description of multimodel machine learning model does not move to distinguish over prior art as the description does not affect the positively recited functions of the system.
As per claims 23 and 24, Ranca/Panagopoulou further teaches wherein the output of the multimodal machine learning model comprises: a first accuracy score associated with the vehicle damage estimate; and a second competitiveness score associated with the vehicle damage estimate (Ranca: C20 L60-67, predicted repair cost; C21 L30-42; C26 L34-40, repairs are unnecessary, incorrectly priced or even fraudulent; C28 L42-53, leakage check; C30 L11-15, level of leakage)(by disclosing incorrectly price teaches competitiveness score).
Claim(s) 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Ranca/Panagopoulou in view of US Patent No. 10,922,726 (“Nelson”).
Per claim 19, while Ranca/Panagopoulou discloses wherein the input to the multimodal machine learning model comprises: a first input based on an image of vehicle damage within the damage data and a third input based on a vehicle specification of the vehicle (Ranca: C3 L24-27, receiving vehicle input data, comprising at least a plurality of images some comprising image data of damage to the vehicle; C8 L26-31, providing photos; C8 L28-42, injured person, crash details and photos of the damage are provided by the party and the computer system is configured to display a dynamic live estimate of the repair cost of the damage to the vehicle; C10 L1-45, use of photos and information to feed the trained models to predict damage; C11/L6-L12/5; C17 L1-9; C18 L15-17, vehicle specification; C22 L5-16, vehicle information include model information, specifics of the vehicle; C24 L7-24; C24 L45-48, accident details and photographic images of any damage are provided by the party to the insurer)(telematics data is optional in the claim), Ranca/Panagopoulou does not specifically teach a second input based on telematics data from the vehicle.
Nelson, however, teaches wherein a second input is based on telematics data from a vehicle (C5 L49-58, telematic data, sensor data, diagnostics data).
It would have been obvious to one of ordinary skill in the art before the effective filing of instant claim to include the any relevant information, i.e., known information such as telematic data, as taught by Nelson as data collected and input to the multimodal machine model in Ranca/Panagopoulou for the purpose of producing accurate prediction and estimation.
Claim(s) 5, 6, 14, 15, and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Ranca/Panagopoulou as applied in claims 1, 9, and 17 as above, in view of US Patent Publication No. 20230334587 (“Van Oosterom”).
Per claims 5 and 14, Ranca/Panagopoulou discloses claims 1 and 14 above. While Ranca/Panagopoulou discloses evaluation of the repair entity’s estimate to check incorrectly priced or even fraudulent in order to determine whether the claim input data can be approved or is anomalous or whether further assessment may be required (Ranca: C23 L 35-46), Ranca/Panagopoulou does not teach determining the entity score.
Van Oosterom, however, teaches determining repair shop or body shop score based on parameters ([0022], i.e., expense-related parameters, improper repair operations, etc.). Hence, as Ranca teaches evaluation of the repair entity’s estimate, it would have been obvious to include the technique of scoring the repair entity to Ranca/Panagopoulou as for the purpose of finding the right balance at scale between operational costs, customer experience and ensuring consistency in repairs ([0003]).
As per claims 6, 15, and 20, Ranca/Panagopoulou claims 1 and 14 above. While Ranca/Panagopoulou discloses evaluation of the repair entity’s estimate to check incorrectly priced or even fraudulent in order to determine whether the claim input data can be approved or is anomalous or whether further assessment may be required (Ranca: C23 L 35-46), claims 1 and 14 above, Ranca/Panagopoulou does not teach that the claim input data can be approved is based on an entity score associated with the repair entity.
Van Oosterom, however, teaches determining repair shop or body shop score based on parameters and use of the score in making decision ([0022], i.e., expense-related parameters, improper repair operations, etc.; [0059]).
Hence, as Ranca teaches approval of the claim based on evaluation of the repair entity’s estimate, it would have been obvious to include the technique of scoring the repair entity to Ranca/Panagopoulou as for the purpose of finding the right balance at scale between operational costs, customer experience and ensuring consistency in repairs ([0003]).
Response to the Argument(s)
101
The applicant asserts that methods for training and/or executing multimodal machine learning models to score vehicle estimates, based on multimodal data including the damage estimates to be scored as well as additional data such as vehicle damage data, image data, telematics data, vehicle specifications, and/or location data cannot be viewed as falling into the group of certain methods of organizing human activity, mental processes, and mathematical formulae, calculations, or relationships (see pages 10-11).
The examiner respectfully disagrees. Gathering information, i.e., 1) at least one of image data, telematic data, or text-based description of damage to a vehicle and 2) at least one of listing of parts or a listing of services identified and determined by a repair shop as damage estimate, formatting the gathered information, scoring of the damage estimate (i.e., accuracy/competitiveness of the damage estimate), and transmitting instructions (i.e., whether to approve or deny the repair according to the damage estimate) based on the score is certainly a certain method of organizing human activities (i.e., commercial interactions in business relations, managing relationships or interactions between people).
The steps related to providing of the formatted information to the multimodal machine model for the purpose of output a predicted repair cost, an estimate accuracy score, and/or an estimate competitiveness score (see [0059] specification), under the broadest reasonable interpretation to be a mental activities and/or mathematical concept (multimodal). Even when machine learning (architecture) is considered as additional element(s), the additional element represents mere instructions or uses computer component to implement the abstract idea.
The applicant asserts that the claims provide technical improvements with respect to training and/or using machine learning models, based on multimodal data using multiple input tokens of different types and points to paragraphs 23, 37, 58, and 107 of the specification to various technical improvements including providing more efficient operations and obtaining more accurate vehicle damage estimates more quickly, as well as automatically flagging and/or analyzing outlier estimates, and/or automatically initiating vehicle repairs (see pages 11-12).
In response, the examiner would like to point out that the paragraphs 23, 37, 58, and 107 do not describe why the multiple input tokens of different types improves upon the training or the algorithm of the model. The sections merely discuss the types of information that are used in the determining of the accuracy/validity/etc. of the received damage estimate which in itself is an abstract idea.
The additional element(s) of using machine, i.e., processor-based machine(s), are used to perform the judicial exception to result in automation and speed of the judicial exception. In other words, the implementing of the judicial exception by the machine does not improve upon the machine itself rather implementing of the judicial exception on generic computer.
The applicant asserts that the claim is significantly more as the Office has not shown that the use of multimodal machine leaning models to generate scores associated with vehicle damage estimates was “well-understood, routine, conventional activity in the field”. In response, the examiner would like to point out that “well-understood, routing, conventional activity in the field” was not the basis of the rejection. Furthermore, the examiner would like to point out that lack of novelty under 35 U.S.C. 102 or obviousness under 35 U.S.C. 103 of a claimed invention does not necessarily indicate that additional elements are well-understood, routine, conventional elements.
103
The applicant asserts that Ranca does not teach or suggest a machine learning model that receives damage estimates as input. The examiner respectfully disagrees as Ranca clearly teaches a machine learning model that receives damage estimates as input (see at least Figure 24).
The other arguments are moot in light of the reference that has been introduced to address the newly added limitations.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
US 20220058579 A1 discloses use of NLP machine learning model that inputs vehicle repair estimate as a third input and the output of the NLP machine learning model that comprises an accuracy score for the estimate and/or a vehicle repair procedure. The NLP machine learning also is fed with first input, i.e., one or more images of a damaged vehicle, for output respresenting a plurality of vehicle repair operations.
US 10535103 B1 discloses analyzing video or image data before and after collision for estimating an extent of damage to a vehicle. The damage may also be estimated by analyzing any telematic data of the vehicle before and during the collision.
US 20230289744 A1 discloses use of self-learning artificial intelligence that can evaluate a geometric scan of a vehicle, images of the vehicle, and the like collectively presented as damaged data and compare the information to items and data in a corresponding damage estimate to predict the severity and location of the damage and the length of time it will take for a particular shop to perform the repair.
US 10360601 B1 discloses automatic generation of repair estimates based on a plurality of data-points via predictive analytics of databases of prior repair estimates and images or simulated structured estimate data in order to identify, source, price and procure all the necessary parts.
The prior art of record does not particularly teach that claim 21 wherein the operation further comprises: determining that the repair entity profile is not available for a repair entity associated with the vehicle damage estimate; and providing, as the input to the multimodal machine learning model, a filler token at the third predetermined location in the input, based at least in part on determining that the repair entity profile is not available.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to STEVEN S KIM whose telephone number is (571)270-5287. The examiner can normally be reached Monday -Friday: 7:00 - 3:30.
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, Patrick McAtee can be reached on 571-272-7575. 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.
/STEVEN S KIM/Primary Examiner, Art Unit 3698
Prosecution Insights