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 .
Claims 1-20 are pending. Claims 1-20 are rejected herein.
Priority
This application claims priority to provisional application 63/500,664 and provisional application 63/522,731. Therefore, this application has an effective filing date of 08 May 2023.
Claim Rejections - 35 USC § 101
Claims 11-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter.
Claim 11 is rejected because it does not sufficiently recite a non-transitory computer readable storage medium. The United States Patent and Trademark Office (USPTO) is obliged to give claims their broadest reasonable interpretation consistent with the specification during proceedings before the USPTO. See In re Zletz, 893 F.2d 319(Fed. Cir. 1989) (during patent examination the pending claims must be interpreted as broadly as their terms reasonably allow). The broadest reasonable interpretation of a claim drawn to a computer readable medium (also called machine readable medium and other such variations) typically covers forms of non-transitory tangible media and transitory propagating signals per se in view of the ordinary and customary meaning of computer readable media, particularly when the specification is silent. See MPEP 2111.01. When the broadest reasonable interpretation of a claim covers a signal per se, the claim must be rejected under 35 U.S.C. §101 as covering non-statutory subject matter. See In re Nuijten, 500 F.3d 1346, 1356-57 (Fed. Cir. 2007) (transitory embodiments are not directed to statutory subject matter) and Interim Examination Instructions for Evaluating Subject Matter Eligibility Under 35 U.S.C. §101, Aug. 24, 2009; p. 2.
The USPTO recognizes that applicants may have claims directed to computer readable media that cover signals per se, which the USPTO must reject under 35 U.S.C. §101 as covering both non-statutory subject matter and statutory subject matter. In an effort to assist the patent community in overcoming a rejection or potential rejection under 35 U.S.C. §101 in this situation, the USPTO suggests the following approach. A claim drawn to such a computer readable medium that covers both transitory and non-transitory embodiments may be amended to narrow the claim to cover only statutory embodiments to avoid a rejection under 35 U.S.C. §101 by adding the limitation "non-transitory" to the claim. Cf. Animals – Patentability, 1077 Off. Gaz. Pat. Office 24 (April 21, 1987) (suggesting that applicants add the limitation "non-human" to a claim covering a multi-cellular organism to avoid a rejection under 35 U.S.C. §101). Such an amendment would typically not raise the issue of new matter, even when the specification is silent because the broadest reasonable interpretation relies on the ordinary and customary meaning that includes signals per se. The limited situations in which such an amendment could raise issues of new matter occur, for example, when the specification does not support a non-transitory embodiment because a signal per se is the only viable embodiment such that the amended claim is impermissibly broadened beyond the supporting disclosure. See, e.g., Gentry Gallery, Inc. v. Berkline Corp., 134 F.3d 1473 (Fed. Cir. 1998).
Note that the Specification at para. [0087] teaches that as employed herein, the term “hardware processor subsystem” or “hardware processor” will refer to a processor, memory, software or combinations thereof that cooperate to perform one or more specific tasks. Note that the Specification at para. [0070] teaches that the memory 530 will be embodied as any type of volatile or non-volatile memory or data storage capable of performing the functions described herein. Therefore, an embodiment of the invention exists that has volatile or impermanent memory. Non-limiting examples of claims that are not directed to any of the categories include:
Products that do not have a physical or tangible form, such as information (often referred to as "data per se") or a computer program per se (often referred to as "software per se") when claimed as a product without any structural recitations;
Transitory forms of signal transmission (often referred to as "signals per se"), such as a propagating electrical or electromagnetic signal or carrier wave; and
Subject matter that the statute expressly prohibits from being patented, such as humans per se, which are excluded under The Leahy-Smith America Invents Act (AIA ), Public Law 112-29, sec. 33, 125 Stat. 284 (September 16, 2011).
The volatile memory is considered a transitory form of signal transmission (often referred to as “signals per se”) such as propagating electrical or electromagnetic signal or carrier wave.
Claims 12-20 depend from Claim 11 and are rejected for the reasons noted in the rejection(s) of Claim 11, above.
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 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
Step 1: The Statutory Categories
Claims 1 and 11 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claims recite a method and system for information extraction, both are within a statutory category for subject matter eligibility analysis purposes.
Step 2A Prong One: The Abstract Idea
The limitations of (claim 1 being representative) configuring a language model with an information extraction instruction prompt and at least one labeled example prompt; validating configuration of the language model with an information extraction instruction prompt and at least one labeled example prompt; validating configuration of the language model using at least one validation prompt, the information extraction prompt, and the at least one labeled example prompt; correcting errors made by the language model in response to the at least one validation prompt using a correction prompt; performing information extraction on an unlabeled sentence using the language model to identify a relation from the unlabeled sentence, using the information extraction prompt, the at least one labeled example prompt, and the correction prompt; and performing an action responsive to the identified relation, as drafted, is a process that, under the broadest reasonable interpretation, covers performance of the limitation in the mind but for recitation of generic computer components. That is, other than reciting that the method is implemented by a computer consisting of a hardware processor and a memory that stores a computer program, nothing in the independent claims preclude the step from practically being performed in the mind. For example, but for the computer, hardware processor and memory that stores a computer program, this claim encompasses a person thinking about validating the language model, correcting errors made by the language model, performing information extraction on an unlabeled sentence and performing an action responsive to the identified relation in the manner described in the identified abstract idea, supra. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
Step 2A Prong Two: Practical Application
This judicial exception is not integrated into a practical application. In particular, the claim recites the additional element of a computer consisting of a hardware processor and memory that implements the identified abstract idea. The computer and various computer part(s) are not described by the applicant and is recited at a high-level of generality such that they amount to no more than mere instructions to apply the exception using a generic computer component. Accordingly, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea.
Step 2B: Significantly More
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the additional element of using a computer with a hardware processor and memory to perform the noted steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept (“significantly more”).
Dependent Claims and Additional Elements
Claims 2-10, and 12-20 are similarly rejected because they either further define/narrow the abstract idea and/or do not further limit the claim to a practical application or provide as inventive concept such that the claims are subject matter eligible even when considered individually or as an ordered combination. Claim(s) 2 and 12 merely describe(s) wherein the certain prompts include a textual description of an information extraction task, including a definition of a relation format. Claim(s) 3 and 13 merely describe wherein the at least one labeled example prompt is drawn form a set of training data that includes sentences and associated relations. Claim(s) 4 and 14 merely describe wherein the at least one validation prompt also comes from the training data set. Claim 5 and 15 merely describes wherein the correction prompt identifies a response to the at least validation prompt that does not match a label of the at least one validation prompt from the training data and provides supplies the language model with the label. Claim 6 and 16 merely describes the at least one labeled example prompt includes a confidence score and wherein inputting the test prompt to the language model further determines a confidence score associated with the relation. Claim 7 and 17 merely describe wherein the unlabeled sentence relates to a patient’s medical condition. Claim 8 and 18 merely describe wherein the performing the action includes automatically adjusting a patient’s treatment based on the identified relation. Claim 9 and 19 merely describe wherein the identified relation is stored in a medical history of the patient to aid in medical decision making by a healthcare professional. Claim 10 and 20 merely describe the language model is a certain type of model.
The dependent claims further recite the additional element of an explicitly pretrained large language model based on a machine learning model. MPEP 2106.05(f) indicates that a consideration when determining whether a claim integrates a judicial exception into a practical application in Step 2A Prong Two or recites significantly more than a judicial exception in Step 2B is whether the additional elements amount to more than a recitation of the words “apply it” (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer. The identified additional element is no more than mere recitation of the words “apply it” (or an equivalent) and/or are instructions to implement an abstract idea or other exception on a computer and therefore cannot provide a practical application or significantly more. Accordingly, even in combination, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea.
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.
Claim(s) 1,3,6,10,11,13,16 and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over US 11,983,488 B1 (hereafter Puri) in view of Raman (Planning with large language models via corrective re-prompting).
Regarding Claim 1
Puri teaches:
A computer-implemented method for information extraction, comprising: configuring a language model with an information extraction instruction prompt and at least one labeled example prompt; [Puri teaches at col. 5 line 38-42 in some embodiments a language model application framework will correspond to at least one of generation, open Quester-Answer (QA), closed QA, brainstorming, chat, rewriting, summarization, classification, extraction, or other. Puri teaches at col. 23 line 16-22 featurization engine 1120 will include feature annotating & labeling engine 1112 (e.g., configured to annotate or label features from a model or data, which will be extracted by feature extraction engine 1114) feature extraction engine 1114 (e.g., configured to extract one or more features from a model or data), and/or feature scaling and selection engine 1116. Puri teaches at claim 1 “accessing a language model using the at least one processor based on the input text prompt, the set of model parameters, and the one or more user instructions;”…. The input text prompt is interpreted as the information extraction instruction prompt. Puri teaches at col. 18 line 14-15 in some embodiments, training data sets will comprise example user prompts. The example user prompts are interpreted to be labeled example prompts. Puri teaches at col. 5 line 57-63 in some embodiments the training dataset will also include annotated data, labeled data, or other types of enriched data. Collectively, the teachings of Puri encompass configuring a language model with an information extraction instruction prompt and at least one labeled example prompt.]
validating configuration of the language model using at least one validation prompt, the information extraction prompt, and the at least one labeled example prompt; [Puri teaches at col. 4 line 45 the model will also be used to insert text within text by providing a suffix prompt in addition to a prefix prompt, when writing long-form text, transitioning between paragraphs, following an outline, guiding the model towards an ending, or inserting code in the middle of a function or file. Puri teaches at col. 4 line 67 -col. 5 line 3 data input engine will obtain user instructions, comprising text data in the form of at least one of a sentence, a paragraph, or a user prompt. Puri teaches at col. 11 line 66- col. 12 line 4 LM optimization engine will perform optimization by aligning or fine-tuning a language model from language model access engine, based on one or more desired output behavior or user intent derived from prefix input data, suffix input data and/or output from content analysis engine. The data received from the LM optimization engine teaches, here, the validation prompt (prefix input data), the information extraction prompt (suffix input data), and the at least one labeled example prompt (output from content analysis engine). Note there is nothing in the broadest reasonable interpretation of the claim dictating the form of the prompts. Puri teaches at col. 14 line 56-59 in some embodiments, demonstration data will be used as validation model to validate a language model as part of a machine learning process (e.g., as discussed below with respect to step 817). Collectively, Puri teaches validating configuration of the language model using at least one validation prompt, the information extraction prompt, and the at least one labeled example prompt.]
performing information extraction on an unlabeled sentence using the language information to identify a relation from the unlabeled sentence, using the information extraction prompt, the at least one labeled example prompt, and the correction prompt; [The information extraction prompt, the at least one labeled example prompt, and the correction prompt have all been taught above by Puri. Puri teaches at col. 23 line 16-22 featurization engine 1120 will include feature annotating & labeling engine 1112 (e.g., configured to annotate or label features from a model or data, which will be extracted by feature extraction engine 1114) feature extraction engine 1114 (e.g., configured to extract one or more features from a model or data), and/or feature scaling and selection engine 1116. The labeling engine input is a unlabeled model or data. Puri teaches at col. 4 line 41-44 for instance, the model will be used to edit text given a prompt and an instruction from the user, thus providing a natural interface for translating and tweaking text, as well as for refactoring and working with code. Puri teaches at col. 4 line 45 the model will also be used to insert text within text by providing a suffix prompt in addition to a prefix prompt, when writing long-form text, transitioning between paragraphs, following an outline, guiding the model towards an ending, or inserting code in the middle of a function or file. Puri teaches at col. 5line 30-32 data normalization engine will also perform lemmatization, stemming and part of speech tagging of input data. Performing part of speech tagging of input data is performing information extraction on an unlabeled sentence using the language information to identify a relation from the unlabeled sentence. Regardless, collectively Puri teaches performing information extraction on an unlabeled sentence using the language information to identify a relation from the unlabeled sentence, using the information extraction prompt, the at least one labeled example prompt, and the correction prompt.]
[…]
and performing an action responsive to the identified relation. [Puri teaches at col. 11 line 66- col. 12 line 4 LM optimization engine will perform optimization by aligning or fine-tuning a language model from language model access engine, based on one or more desired output behavior or user intent derived from prefix input data, suffix input data and/or output from content analysis engine. This is teaching and performing an action responsive to the identified relation. The optimization is the action. The difference aligning or fine tuning the model and the desired output behavior or user intent derived from prefix input data, suffix data, and output from content analysis engine is performing an action responsive to the identified relation.]
Puri may not explicitly teach:
correcting errors made by the language model in response to the at least one validation prompt using a correction prompt;
Raman teaches:
correcting errors made by the language model in response to the at least one validation prompt using a correction prompt; [Raman teaches at the Abstract when an agent is unable to execute an action, our approach re-prompts the LLM with precondition error information to extract an executable corrective action to achieve the intended goal in the current context. The original prompt is the validation prompt here and the re-prompt is the correction prompt.]
Therefore, it would have been prima facie obvious to one of ordinary skill in the art of healthcare, at the time of filing, to modify the systems and methods for language model-based text editing of Puri to planning with large language models via corrective re-prompting of Raman with the motivation of extracting the common sense knowledge present in Large Language Models (LLMs) offers a path to designing intelligent, embodied agents (Raman at the Abstract).
Regarding Claim 11
Due to its similarity to Claim 1, Claim 11 is similarly analyzed and rejected in a manner consistent with the rejection of Claim 1.
Regarding Claim 3
Puri/Raman teach the method of claim 1. Puri/Raman further teach:
wherein the at least one labeled example prompt is drawn from a set of training data that includes sentences and associated relations. [Puri teaches col. 4 line 14-17 teaches while a model will have an initial configuration, this configuration will change over time as the model learns from input data (e.g., training input data), which allows the model to improve its abilities. Puri teaches at col. 18 line 14-15 in some embodiments, training data sets will comprise example user prompts. Puri teaches at col. 5 line 57-63 in some embodiments the training dataset will also include annotated data, labeled data, or other types of enriched data. Puri teaches at col. 14 line 11-15 for example, input data will include one or more of a user-written or machine-written prompt, a user-written or machine-written instruction, web-crawled text, or any other text data (E.g., one or more words, phrases, sentences, or paragraphs). Puri, for example, the at least one processor will determine that the selected language model matches a label included with the input data and/or may determine a semantic similarity between at least a portion of the input data and the selected language model (e.g., by computing distance between a text embedding associated with sample data input and a text embedding associated with one or more language models from which the selection is made). Puri teaches at col. 15 line 24-26 teaches at step 809, the at least one processor will perform context analysis (e.g., in response to and/or based on the input data). Collectively, this teaches wherein the at least one labeled example prompt is drawn from a set of training data that includes sentences and associated relations.]
Regarding Claim 13
Due to its similarity to Claim 3, Claim 13 is similarly analyzed and rejected in a manner consistent with the rejection of Claim 3.
Regarding Claim 6
Puri/Raman teach the method of claim 1. Puri/Raman further teach:
wherein the at least one labeled example prompt includes a confidence score and wherein inputting the test prompt to the language model further determines confidence score associated with the relation. [Puri teaches col. 4 line 14-17 teaches while a model will have an initial configuration, this configuration will change over time as the model learns from input data (e.g., training input data), which allows the model to improve its abilities. Puri teaches at col. 18 line 14-15 in some embodiments, training data sets will comprise example user prompts. Puri teaches at col. 5 line 57-63 in some embodiments the training dataset will also include annotated data, labeled data, or other types of enriched data. Puri teaches at col. 17 line 14-19 in some embodiments, language model access engine 106 will align a language model to maximize the proximity of its outputs to the one or more desired behavior outputs (e.g., maximize a quality score or a numerical similarity score of output data). The numerical similarity score is interpreted as the confidence score. Collectively, Puri teaches wherein the at least one labeled example prompt includes a confidence score and wherein inputting the test prompt to the language model further determines confidence score associated with the relation.]
Regarding Claim 16
Due to its similarity to Claim 6, Claim 16 is similarly analyzed and rejected in a manner consistent with the rejection of Claim 6.
Regarding Claim 10
Puri/Raman teach the method of claim 1. Puri/Raman further teach:
wherein the language model is a pretrained large language model based on a machine learning model. [Puri teaches at col. 23 line 63- col. 24 line 8 a machine learning model will be or include, without limitation, one or more of (E.g., such as in the case of a metamodel) a statistical model, an algorithm, a neural network (NN), a convolutional neural network (CNN), a generative neural network (GNN), a Word2Vec model, a bag of words model, a term frequency-inverse document frequency (tf-idf) model, a GTP (Generative Pre-trained Transformer) model (or other autoregressive model), a Proximal Policy Optimization (PPO) model, a nearest neighbor model, a linear regression model, a Temporal Difference (TB) model, a Deep Adversarial Network mode, or any other type of model described further herein. Puri teaches at col. 4lin 37-41 teaches that LMs (Puri defines LMs at col. 1 line 13 to be large language models) of various capabilities, described herein, will be utilized to improve the versatility and robustness of Application Programing Interfaces (APIs) to perform a multitude of tasks involving understanding or generating natural language or code.]
Regarding Claim 20
Due to its similarity to Claim 10, Claim 20 is similarly analyzed and rejected in a manner consistent with the rejection of Claim 10.
Claim(s) 2,4,5,12,14 and 15 is/are rejected under 35 U.S.C. 103 as being unpatentable over US 11,983,488 B1 (hereafter Puri) in view of Raman(Planning with Large Language Models via Corrective Re-prompting) in view of Agrawal (Large Language Models are Few-Shot Clinical Information Extractors).
Regarding Claim 2
Puri/Raman teach the method of claim 1. Puri/Raman may not explicitly teach:
wherein the information extraction instruction prompt and the at least one labeled example prompt include a textual description of an information extraction task, including a definition of a relation format.
Agrawal teaches
wherein the information extraction instruction prompt and the at least one labeled example prompt include a textual description of an information extraction task, including a definition of a relation format. [Agrawal teaches at Figure 1 a Prompt: create a list of medications. This teaches the information extraction instruction prompt and the at least one labeled example prompt, which are interpreted to be one and the same, there being no indication in the claim that they cannot be the same. Agrawal teaches at Figure 1 the prompt includes a textual description of the information extraction task: create a list of medications. Agrawal teaches at Table 1 a medication attribute extraction format under the Answer column: aspirin: {dose 325 mg, freq: per day, duration: three years, reason: TIA}. The definition of a relation format is interpreted to be dose, frequency, duration. Collectively, Agrawal teaches wherein the information extraction instruction prompt and the at least one labeled example prompt include a textual description of an information extraction task, including a definition of a relation format.]
Therefore, it would have been prima facie obvious to one of ordinary skill in the art of healthcare, at the time of filing, to modify the systems and methods for language model-based text editing of Puri to the planning with large language models via corrective re-prompting of Raman to the Large Language Models are Few-Shot Clinical Information Extractors of Agrawal with the motivation of addressing a long-running goal of the clinical NLP community, the extraction of important variables trapped in clinical notes.
Regarding Claim 12
Due to its similarity to Claim 2, Claim 12 is similarly analyzed and rejected in a manner consistent with the rejection of Claim 2.
Regarding Claim 4
Puri/Raman teach the method of claim 3. Puri/Raman may not explicitly teach:
wherein the at least one validation prompt is also drawn from the set of training data.
Agrawal teaches:
wherein the at least one validation prompt is also drawn from the set of training data. [Agrawal teaches at pg. 3 prompt-based learning requires the specification of a prompt template to be applied on the input. Agrawal teaches at pg. 3 in this work, we handcraft our prompt templates using a set of 5 validation examples per task. Collectively, this teaches wherein the at least one validation prompt is also drawn from the set of training data.]
Therefore, it would have been prima facie obvious to one of ordinary skill in the art of healthcare, at the time of filing, to modify the systems and methods for language model-based text editing of Puri to the planning with large language models via corrective re-prompting of Raman to the Large Language Models are Few-Shot Clinical Information Extractors of Agrawal with the motivation of addressing a long-running goal of the clinical NLP community, the extraction of important variables trapped in clinical notes.
Regarding Claim 14
Due to its similarity to Claim 4, Claim 14 is similarly analyzed and rejected in a manner consistent with the rejection of Claim 4.
Regarding Claim 5
Puri/Raman/Agrawal teach the method of claim 4. Puri/Raman/Agrawal further teach:
wherein the correction prompt identifies a response to the at least one validation prompt that does not match a label of the at least one validation prompt from the training data and provides supplies the language model with the label. [Puri teaches at col. 17 line 42-43 in some embodiments, the model output will include edited text based on an input text prompt. Puri teaches at col. 17 line 44-45 in some embodiments, the model output will include inserted text based on an input text prompt. Puri teaches at col. 17 line 45-49 in some embodiments, a user device will provide evaluation data (e.g., one or more indications of accuracy of model generations), which the model will use for subsequent training, to further improve the accuracy of the model.]
Regarding Claim 15
Due to its similarity to Claim 5, Claim 15 is similarly analyzed and rejected in a manner consistent with the rejection of Claim 5.
Claim(s) 7-9,17-19 is/are rejected under 35 U.S.C. 103 as being unpatentable over US 11,983,488 B1 (hereafter Puri) in view of Raman(Planning with Large Language Models via Corrective Re-Prompting) in view of US 20240029901 A1 (hereafter Ezhov).
Regarding Claim 7
Puri/Raman teach the method of claim 1. Puri/Raman may not explicitly teach:
wherein the unlabeled sentence relates to a patient’s medical condition.
Ezhov teaches:
wherein the unlabeled sentence relates to a patient’s medical condition. [Ezhov teaches at para. [0092] advantageously, in some embodiments, to accumulate positive cases fasters, a weak model could be trained and ran for all of the unlabeled data. Ezhov teaches at para. [0029] in yet another embodiment of the invention, a method to generate a personalized medical summary (PMS) from practitioner-patient conversation comprising capturing a conversation between the practitioner and the patient, transcribing the conversation between the practitioner and the patient and generating the PMS for the patient based on the transcribed conversation, wherein the PMS is generated by a dental diagnosis module (DAIM) using diagnosis-AI module (DAIM). Ezhov teaches at para. [0217] it accepts practitioner-patient conversations, similar to the input sentences a language model would take, then applies self-attention and feedforward layers to generate relevant code or scripts. Ezhov teaches at para. [0217] in this context, self-attention layers help the DAIM module 3403b identify the relative importance of each word or phrase in the conversation. Ezhov teaches at para. [0217] this allows the module to understand what features of the input should be given higher priority while generating the code. Collectively, this teaches, wherein the unlabeled sentence relates to a patient’s medical condition.]
Therefore, it would have been prima facie obvious to one of ordinary skill in the art of healthcare, at the time of filing, to modify the systems and methods for language model-based text editing of Puri to planning with large language models via corrective re-prompting of Raman to the systems and methods to generate a personalized medical summary (PMS) from a practitioner-patient conversations of Ezhov with the motivation of addressing problems with CBCT, which includes one or more limitations such as time consumption and complexity for personnel to become fully acquainted with the imaging software and correctly using digital imaging and communications in medicine (DICOM) data (Ezhov at para. [0003]).
Regarding Claim 17
Due to its similarity to Claim 7, Claim 17 is similarly analyzed and rejected in a manner consistent with the rejection of Claim 7.
Regarding Claim 8
Puri/Raman/Ezhov teach the method of claim 7. Puri/Raman/Ezhov further teach:
wherein performing the action includes automatically adjusting a patient’s treatment based on the identified relation. [Ezhov teaches at para. [0217] in this context, self-attention layers help the DAIM module 3403b identify the relative importance of each word or phrase in the conversation. This teaches the identified relation. Ezhov teaches at para. [0215] the DAIM is rewarded for completing tasks correctly and penalized for completing tasks incorrectly. Ezhov teaches at para. [0215] further in an embodiment for the invention, once the DAIM has been optimized and validate, it is deployed on new patient data or practitioner diagnosis/treatment to improve its performance. Collectively, the Ezhov teaches wherein performing the action includes automatically adjusting a patient’s treatment based on the identified relation.]
Regarding Claim 18
Due to its similarity to Claim 8, Claim 18 is similarly analyzed and rejected in a manner consistent with the rejection of Claim 8.
Regarding Claim 9
Puri/Raman/Ezhov teach the method of claim 7. Puri/Raman/Ezhov further teach:
wherein the identified relation is stored in a medical history of the patient to aid in medical decision making by a healthcare professional. [Ezhov teaches at para. [0029] in yet another embodiment of the invention, a method to generate a personalized medical summary (PMS) from practitioner-patient conversation comprising capturing a conversation between the practitioner and the patient, transcribing the conversation between the practitioner and the patient and generating the PMS for the patient based on the transcribed conversation, wherein the PMS is generated by a dental diagnosis module (DAIM) using diagnosis-AI module (DAIM). The stored PMS generated by a dental diagnosis module (DAIM) from the information derived from the practitioner-patient conversation is interpreted as the identified relation is stored in a medical history of the patient. The identified relation is interpreted as medical data associated with or derived from the information extraction step, there being no description of what the identified relation must be in the claim. Ezhov teaches at para. [0033] additionally in further embodiments of the invention, the method commences by receiving data about the patient and a prompt from the practitioner. Ezhov teaches at para. [0033] this data could include information about the patient’s medical history, current symptoms, and any relevant test results, while the prompt could be a question or suggestion from the practitioner about how to proceed with the diagnosis or treatment. These last teachings teach that the stored info is used by the practitioner to aid in the decision making by a healthcare professional. Collectively, Ezhov teaches wherein the identified relation is stored in a medical history of the patient to aid in medical decision making by a healthcare professional.]
Regarding Claim 19
Due to its similarity to Claim 9, Claim 19 is similarly analyzed and rejected in a manner consistent with the rejection of Claim 9.
Response to Arguments
35 U.S.C. 101 Rejection Arguments and Responses
Applicant argues that the rejections are directed to abstract ideas with significantly more. Applicant argues that the claims are directed to an improvement to a technology. MPEP 2106.04, Applicant argues, sets out two requirements to establish that claims integrate an abstract idea with an improvement to a technology. First, the Specification needs to describes the improvement, and second the claims need to reflect that improvement.
The Examiner disagrees.
The way in which the additional elements use or interact with the exception may integrate it into a practical application. Respectfully, the claims present mere instructions to apply an exception on a computer with the additional elements. For example, beyond indicating that the claims are computer implemented the entire abstract idea could be completed mentally: “…configuring a language model with an information extraction instruction prompt and at least one labeled example prompt; validating configuration of the language model with an information extraction instruction prompt and at least one labeled example prompt; validating configuration of the language model using at least one validation prompt, the information extraction prompt, and the at least one labeled example prompt; correcting errors made by the language model in response to the at least one validation prompt using a correction prompt; performing information extraction on an unlabeled sentence using the language model to identify a relation from the unlabeled sentence, using the information extraction prompt, the at least one labeled example prompt, and the correction prompt; and performing an action responsive to the identified relation.” Note that the abstract idea has been rewritten to include the prompts because even the prompts themselves aren’t meaningfully limited to the realm of the computer improvement, thereby raising questions about the scope of the claimed improvement aligning with features, steps and/or limitations recited in the claim. The prompts themselves, given broadest reasonable interpretation could be other than computer prompts.
Second, it is important to keep in mind that an improvement in the abstract idea itself (e.g. a recited fundamental economic concept) is not an improvement in technology. In order to demonstrate an improvement in this way a technical solution to a technical problem must be demonstrated in the claims. The improvement discussed re in context learning is an improvement imparted by the judicial exception itself (it is an improved abstract idea) it is not yet a technical solution. It is an improved abstract idea that Applicant is running on the computer, but that improvement would be expected to improve in context learning even if executing the abstract idea mentally, therefore, the claims cannot be said to be an improvement to technology. For example, Applicant states, “To that end, specific-domain open information extraction using a large language model will be enhanced by in-context learning.” Indeed, Applicant does not describe what enhancements, nor why they represent an improvement in terms of a technical solution to a technical problem. These factors may be other than technical parameters. In fact, we do not know how the general purpose computer is impacted by implementation of the abstract idea on the computer because Applicant hasn’t recited the features or steps of the invention that result in an improvement to technology via demonstrating a technical solution to a technical problem. It is important to note, the judicial exception alone cannot provide the improvement. The improvement can be provided by one or more additional elements. Because a judicial exception alone is not eligible subject matter, if there are no additional claim elements besides the judicial exception, or if the additional claim elements merely recite another judicial exception, that is insufficient to integrate the judicial exception into a practical application.
The present specification explains how general purpose language model can be ill-suited to performing specific tasks, as they are not specifically trained to perform the task. These models also lack access to domain-specific information. The present specification describes an approach that solves this problem using in-context learning, summarized in paragraph 16:
To that end, specific-domain open information extraction using a large language model will be enhanced by in-context learning. Given an input sentence, the most import relations may be extracted, including a subject, an action, and an object. A set of domain specific sentences with ground-truth labels and an associated confident label will be used, with the ground-truth labels including a subject, action, and object. An initial prompt may be designed for general purpose language model, including a description of the information extraction task. A series of prompts may be provided some of with the domain-specific sentences and their associated labels to provide domain-specific knowledge to the language model. Additional prompts may be used to test the understanding of the language model and to provide corrections when the model generates erroneous results. The language model can then be used to extract information from unlabeled new sentences and to provide a confidence score as to the correctness of its output.
See response above which addresses this topic.
The remainder of the specification providers specific details on how this may be performed. It is therefore respectfully asserted that the specification does describe an improvement to a technology in improving the performance of a language model as applied to a specific task or to domain-specific information. The claims furthermore reflect his improvement. The steps of validating and correcting errors using respective prompts to the language mode, and then performing the recited task, implements the improvements that are described in the claims. Based on the interview with the Examiner on December 19, 2025, the claims have been amended to explicitly recite the context established by the prior prompts when performing the validation and when performing the information extraction.
Additionally, if it is asserted that the invention improves upon conventional functioning of a computer, or upon conventional technology or technological processes, a technical explanation as to how to implement the invention should be present in the specification. That is, the disclosure must provide sufficient details such that one of ordinary skill in the art would recognize the claimed invention as providing an improvement. The specification need not explicitly set forth the improvement, but it must describe the invention such that the improvement would be apparent to one of ordinary skill in the art. Conversely, if the specification explicitly sets forth an improvement but in a conclusory manner (i.e., a bare assertion of an improvement without the detail necessary to be apparent to a person of ordinary skill in the art), the examiner should not determine the claim improves technology. An indication that the claimed invention provides an improvement can include a discussion in the specification that identifies a technical problem and explains the details of an unconventional technical solution expressed in the claim, or identifies technical improvements realized by the claim over the prior art.
In computer-related technologies, the examiner should determine whether the claim purports to improve computer capabilities or, instead, invokes computers merely as a tool.
See relevant response above. Currently, in sum, Applicant has demonstrated an improved pre-existing abstract idea and has provided, with the additional elements, instructions to implement that abstract idea on the computer and there is uncertain specific impact on the computer, the environment to which the claimed limitations are confined.
35 U.S.C. 103 Rejection Arguments are Responses
Claim 1 recites distinct prompt including “an information extraction instruction prompt,” “at least one validation prompt,” and “a correction prompt.’ Claim 11 recite analogous language. The rejection cites Puri’s “user instructions” to read on all three of these prompts. Applicant argues that Puri does not disclose that the user include three distinct prompts.
The Examiner has reviewed the art again and would like to clarify what imitations Puri taught. The Examiner agrees that Puri was used in the original rejection to teach several limitations that included various distinct prompts. The Examiner purports that Puri does in fact teach several distinct prompts. The original independent claim read had Puri applied to the following actually recited limitations, not the paraphrased recitations of the prompts:
A computer-implemented method for information extraction comprising: configuring a language model with an information extraction prompt and at least one labeled example prompt;
Puri teaches at col. 5 line 38-42 in some embodiments a language model application framework will correspond to at least one of generation, open Quester-Answer (QA), closed QA, brainstorming, chat, rewriting, summarization, classification, extraction, or other. Puri teaches at col. 23 line 16-22 featurization engine 1120 will include feature annotating & labeling engine 1112 (e.g., configured to annotate or label features from a model or data, which will be extracted by feature extraction engine 1114) feature extraction engine 1114 (e.g., configured to extract one or more features from a model or data), and/or feature scaling and selection engine 1116. Puri teaches at claim 1 “accessing a language model using the at least one processor based on the input text prompt, the set of model parameters, and the one or more user instructions;”…. The input text prompt is interpreted as the information extraction instruction prompt.
Therefore, Puri teaches an information extraction instruction prompt AND at least one labeled example prompt.
validating configuration of the language model using at least one validation prompt;
Puri teaches at col. 11 line 53-54 teaches that the system 200 will further include output validation engine 214. Puri teaches at col. 14 line 56-59 in some embodiments, demonstration data will be used as validation model to validate a language model as part of a machine learning process (e.g., as discussed below with respect to step 817). Puri teaches at col. 7 line 48-51 output validation engine 114 will execute a ranking of the received model outputs based on the set of user instructions, the output from context analysis engine 108, or the output from sentiment analysis. Puri teaches at col. 11 line 54-56 in some embodiments, output validation engine 214 will receive a set of model outputs, user-labelled output, or set of comparison data. Puri teaches at col. 11 56-59 output validation engine 214 will execute a ranking of the received model outputs based on the set of user instructions, the output from context analysis engine 208, or the output from sentiment analysis. The set of user instructions received is interpreted as the validation prompt. Collectively, the teachings of Puri encompass validating configuration of the language model using at least one validation prompt.
correcting errors made by the language model in response to the at least one validation prompt using a correction prompt;
Puri teaches at col. 14 line 56-59 in some embodiments, demonstration data will be used as validation model to validate a language model as part of a machine learning process (e.g., as discussed below with respect to step 817). Puri teaches at col. 14 line 63-67 in some embodiments, the system will use the labeled data based on (E.g., using) an engine, such as language model (LM) optimization engine 116 in Figure 1 or 216 in Figure 2, to fine-tune the alignment of a language model to the desired output behavior. Puri teaches at col. 16 line 44-48 in some embodiments, the at least one processor will use the demonstration data as validation data to determine quality score or other metrics of model output, to train a language model to generate improved digital text outputs. Puri at col. 18 line 14-15 teaches in some embodiments, training data sets will comprise example user prompts. Puri teaches at col. 7, line 57-59 teaches that the system 100 will further include LM optimization engine 116. Puri teaches at col. 7, line 59-62 that LM optimization engine 116 will perform optimization by aligning or fine tuning a language model from language model access engine 106, based on one or more desired output behaviors or user intent derived from a set of user instructions as in 101b. The set of user instructions is interpreted as the correction prompt. Collectively, Puri is interpreted to teach correcting errors made by the language model in response to the at least one validation prompt using a correction prompt.
Regardless, the rejection has been updated in response to amendment, including application of Puri in response to the amendment.
Applicant argues that Puri, Wing, Agrawal, and/or Ezhov, taken alone or in any combination, fail to disclose or suggest an information extraction instruction prompt, a validation prompt, and a correction prompt.
The Examiner disagrees. See answer above and the updated art rejection above. Line by line application of the art to the actually recited limitations is given in the updated rejection.
Puri teaches at col. 4 line 45 the model will also be used to insert text within text by providing a suffix prompt in addition to a prefix prompt, when writing long-form text, transitioning between paragraphs, following an outline, guiding the model towards an ending, or inserting code in the middle of a function or file. Puri teaches at col. 4 line 67 -col. 5 line 3 data input engine will obtain user instructions, comprising text data in the form of at least one of a sentence, a paragraph, or a user prompt. Puri teaches at col. 11 line 66- col. 12 line 4 LM optimization engine will perform optimization by aligning or fine-tuning a language model from language model access engine, based on one or more desired output behavior or user intent derived from prefix input data, suffix input data and/or output from content analysis engine. The data received from the LM optimization engine teaches, here, the validation prompt (prefix input data), the information extraction prompt (suffix input data), and the at least one labeled example prompt (output from content analysis engine). Puri teaches at col. 14 line 56-59 in some embodiments, demonstration data will be used as validation model to validate a language model as part of a machine learning process (e.g., as discussed below with respect to step 817). Collectively, Puri teaches validating configuration of the language model using at least one validation prompt, the information extraction prompt, and the at least one labeled example prompt. Note there is nothing in the broadest reasonable interpretation of the claim dictating the form of the prompts.
Conclusion
US 2005/0038678 A1 (hereafter Qian) teaches searching and obtaining medically related information from different sources and generating matrices of diagnostic information.
Dunn(Structured Information Extraction from Complex Scientific Text with Fine-Tuned Large Language Models) teaches approaches to intelligently extracting an linking complex scientific information from unstructured text.
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 TRISTAN ISAAC EVANS whose telephone number is (571)270-5972. The examiner can normally be reached Mon-Thurs 8:00am-12:00pm & 1:00pm-7:00pm, off Fridays.
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, Robert Morgan can be reached at 571-272-6773. 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.
/T.I.E./Examiner, Art Unit 3683
/CHRISTOPHER L GILLIGAN/Primary Examiner, Art Unit 3683
Prosecution Insights