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 .
Information Disclosure Statement
The information disclosure statements submitted on February 23, 2024 and September 23, 2024 have been considered by the Examiner.
Claim Objections
The Examiner notes that claim 20 is a substantial duplicate of claim 10. Applicant is advised that should claim 10 be found allowable, claim 20 will be objected to under 37 CFR 1.75 as being a substantial duplicate thereof. When two claims in an application are duplicates or else are so close in content that they both cover the same thing, despite a slight difference in wording, it is proper after allowing one claim to object to the other as being a substantial duplicate of the allowed claim. See MPEP § 608.01(m).
Claims 1-20 are objected to because of the following informalities. Appropriate correction is required.
In claim 1, there appears to be a typographical error in the phrase, “a set of the questions-answer threads.” Claims 2-10 and 20 depend from claim 1 and thereby include all of the limitations of claim 1, and are therefore objected to for the same reason.
Further regarding claim 2, there appears to be a typographical error in the phrase, “any subsequent messaged in that question-answer thread.”
Further regarding claim 9, there is no antecedent basis for “the selecting information” recited therein. Claim 9 depends from claim 8, which recites “selected information” but not “selecting information” per se.
In claim 11, there appears to be a typographical error in the phrase, “a set of actual questions-answer threads.” Claims 12-14 depend from claim 11 and thereby include all of the limitations of claim 11, and are therefore objected to for the same reason.
In claim 15, there appears to be a typographical error in the phrase, “a set of the questions-answer threads.” Also in claim 15, there is no antecedent basis for “the computing system” recited therein. Claims 16-19 depend from claim 15 and thereby include all of the limitations of claim 15, and are therefore objected to for the same reason.
Further regarding claim 16, there appears to be a typographical error in the phrase, “for a subsequent messaged in a given question-answer thread.”
Claim Interpretation
The following is a quotation of 35 U.S.C. 112(f):
(f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The following is a quotation of pre-AIA 35 U.S.C. 112, sixth paragraph:
An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The claims in this application are given their broadest reasonable interpretation using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The broadest reasonable interpretation of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked.
As explained in MPEP § 2181, subsection I, claim limitations that meet the following three-prong test will be interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph:
(A) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;
(B) the term “means” or “step” or the generic placeholder is modified by functional language, typically, but not always linked by the transition word “for” (e.g., “means for”) or another linking word or phrase, such as “configured to” or “so that”; and
(C) the term “means” or “step” or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function.
Use of the word “means” (or “step”) in a claim with functional language creates a rebuttable presumption that the claim limitation is to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites sufficient structure, material, or acts to entirely perform the recited function.
Absence of the word “means” (or “step”) in a claim creates a rebuttable presumption that the claim limitation is not to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is not interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites function without reciting sufficient structure, material or acts to entirely perform the recited function.
Claim limitations in this application that use the word “means” (or “step”) are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. Conversely, claim limitations in this application that do not use the word “means” (or “step”) are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action.
This application includes one or more claim limitations that do not use the word “means,” but are nonetheless being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, because the claim limitations use a generic placeholder (i.e. “module”) that is coupled with functional language (e.g. “to obtain incoming electronic messages from corresponding users”) without reciting sufficient structure to perform the recited function and the generic placeholder is not preceded by a structural modifier. Such claim limitations are the “electronic message module,” “thread processing module,” and “agent training module” first introduced in claim 15 and further required by claims 16-19 by virtue of their dependency from claim 15.
Because these claim limitations are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, they is/are being interpreted to cover the corresponding structure (i.e. a programmed processor) described in the specification as performing the claimed function, and equivalents thereof.
If applicant does not intend to have this/these limitation(s) interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, applicant may: (1) amend the claim limitation(s) to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph (e.g., by reciting sufficient structure to perform the claimed function); or (2) present a sufficient showing that the claim limitation(s) recite(s) sufficient structure to perform the claimed function so as to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph.
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claims 15-19 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description and enablement requirements. The claims contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention. These claims also contain subject matter which was not described in the specification in such a way as to enable one skilled in the art to which it pertains, or with which it is most nearly connected, to make and/or use the invention.
In particular, claim 15 recites “an electronic message module,” a “thread processing module” and an “agent training module,” which as noted above, have been interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, and thereby cover the corresponding structure described in the specification that achieves the claimed function, and equivalents thereof. As demonstrated below, these limitations are indefinite under 35 U.S.C. 112(b) for failure to disclose sufficient corresponding structure (e.g., the computer and the algorithm) in the specification that performs the entire claimed function (e.g., for the “thread processing module”, “correlate each responsive message with…). Such a limitation also lacks an adequate written description as required by 35 U.S.C. 112(a) or pre-AIA 35 U.S.C. 112, first paragraph, because an indefinite, unbounded functional limitation would cover all ways of performing a function and indicate that the inventor has not provided sufficient disclosure to show possession of the invention. See MPEP § 2163.03, subsection VI. Under a similar rationale, the specification does not enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the invention commensurate in scope with these claims. Claims 16-19, which depend from claim 15 and thereby include all of the limitations of claim 15, are rejected for the same reasons as claim 15.
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 15-19 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor, or for pre-AIA the applicant regards as the invention.
In particular, claim 15 recites “an electronic message module,” a “thread processing module” and an “agent training module,” which as noted above, have been interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, and thereby cover the corresponding structure described in the specification that achieves the claimed function, and equivalents thereof. The “electronic message module” is configured to perform a specialized function (i.e. “to obtain incoming electronic messages from corresponding users…”), as is the “thread processing module” (i.e. “to correlate each responsive electronic message with…”) and the “agent training module” (i.e. “to receive the question-answer thread…”). In computational contexts, a specialized function must be supported in the specification by the computer and the algorithm that the computer uses to perform the claimed specialized function. See, e.g., In re Katz Interactive Call Processing Patent Litigation, 639 F.3d 1303, 1316, 97 USPQ2d 1737, 1747 (Fed. Cir. 2011). In this case, the specification generally describes a computing system (see e.g. paragraphs 0072-0075 of the specification as originally filed) but does not disclose the algorithm that the computer uses to perform the claimed specialized functions. Accordingly, the scope of the recited limitations is indefinite, and claim 15 thereby fails to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor, or for pre-AIA the applicant regards as the invention. Claims 16-19, which depend from claim 15 and thereby include all of the limitations of claim 15, are rejected for the same reasons.
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 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea (i.e. a mental process) without significantly more.
As described in MPEP § 2106, the analysis as to whether a claim qualifies as eligible subject matter under 35 U.S.C. § 101 includes the following determinations:
(1) Whether the claim is to a statutory category, i.e. to a process, machine, manufacture or composition of matter (“Step 1”) – see MPEP §§ 2106, subsection III, and 2106.03
(2) If the claim is to a statutory category, whether the claim recites any judicial exceptions, including certain groupings of abstract ideas (i.e., mathematical concepts, certain methods of organizing human activity, or mental processes) (“Step 2A, Prong One”) – see MPEP §§ 2106, subsection III, and 2106.04
(3) If the claim recites a judicial exception, whether the claim recites additional elements that integrate the judicial exception into a practical application (“Step 2A, Prong Two”) – see MPEP §§ 2106, subsection III, and 2106.04
(4) If the claim does not recite additional elements that integrate the judicial exception into a practical application, whether the claim recites additional elements that amount to significantly more than the judicial exception (“Step 2B”) – see MPEP §§ 2106, subsection III, and 2106.05
Claim 1
Claim 1 is to a statutory category, as claim 1 is directed to a method, i.e. a process. However, claim 1 recites a judicial exception. In claim 1, “correlating…each responsive electronic message with a given one of the incoming electronic messages as a question-answer thread” is considered a mental process, as is “…learn an answer that addresses the question.” Such tasks can practically be performed in the human mind.
Other than this mental process, claim 1 recites: “obtaining…incoming electronic messages from corresponding users, the incoming electronic messages each including a question; obtaining…responsive electronic messages to the corresponding users, the responsive electronic messages each including an answer to the question;…routing…the question-answer thread for each correlated incoming and responsive electronic message pair to an agent training module of the computing system;…and storing the trained large language model in a database of the system.” Such tasks are indicative of insignificant extra-solution activity, i.e. mere data gathering, and are therefore insufficient to integrate the abstract idea into a practical application. See MPEP § 2106.5(g). Data gathering is also well-understood, routine and conventional and therefore does not amount to significantly more than the judicial exception. See, e.g., Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014).
Claim 1 also recites that the tasks listed therein are performed by “one or more processors” of a computing system, and further recites “performing, by one or more processors of the computing system using an agent training module, training of a large language model using a set of the questions-answer threads as inputs to learn an answer that addresses the question.” However, such recitations represent no more than mere instructions to apply the mental process on a computer, and thus do not integrate the judicial exception into a practical application or amount to significantly more than the judicial exception. See MPEP § 2106.05(f). Particularly, the claim omits any details of the agent training module and large language model; these appear to be invoked merely as a tool for performing the judicial exception.
Consequently, claim 1 recites an abstract idea but does not include additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea. As a result, and for the reasons described above, claim 1 is rejected as being patent ineligible under 35 U.S.C. § 101.
Claims 2 and 3
The additional elements recited in claims 2 and 3 are considered a mental process; as indicated above, correlating messages can practically be performed in the human mind. Claims 2 and 3 thus fail to recite any additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea, and as a result, claims 2 and 3 are also patent ineligible under 35 U.S.C. § 101.
Claim 4
Claim 4 recites that “the training of the large language model comprises fine-tuning a previously trained model for a specific question-answer situation.” However, such a recitation represents no more than mere instructions to apply the mental process on a computer, and thus does not integrate the judicial exception into a practical application or amount to significantly more than the judicial exception. See MPEP § 2106.05(f). Claim 4 thus fails to recite any additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea, and as a result, claim 4 is also patent ineligible under 35 U.S.C. § 101.
Claims 5 and 6
Claims 5 and 6 provide details of the data gathering recited in claim 1 (i.e. that the responsive electronic messages are obtained from a shared inbox or group email address, and are associated with a single person). As noted above, data gathering tasks are considered insignificant extra-solution activity and are insufficient to integrate the abstract idea into a practical application. See MPEP § 2106.5(g). Data gathering is also well-understood, routine and conventional and therefore does not amount to significantly more than the judicial exception. See, e.g., Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014). Claims 5 and 6 thus fail to recite any additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea, and as a result, claims 5 and 6 are also patent ineligible under 35 U.S.C. § 101.
Claim 7
Claim 7 recites that “the trained large language model is configured for use as a virtual assistant for the single person.” However, such a recitation represents no more than mere instructions to apply the mental process on a computer, and thus does not integrate the judicial exception into a practical application or amount to significantly more than the judicial exception. See MPEP § 2106.05(f). Claim 7 thus fails to recite any additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea, and as a result, claim 7 is also patent ineligible under 35 U.S.C. § 101.
Claims 8 and 9
The additional elements recited in claims 8 and 9 are considered a mental process. Processing messages to remove information therefore can practically be performed in the human mind. Claims 8 and 9 thus fail to recite any additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea, and as a result, claims 8 and 9 are also patent ineligible under 35 U.S.C. § 101.
Claim 10
Claim 10 recites that “performing the training includes discarding a learned answer that does not satisfy a question-answer criterion.” However, such a recitation represents no more than mere instructions to apply the mental process on a computer, and thus does not integrate the judicial exception into a practical application or amount to significantly more than the judicial exception. See MPEP § 2106.05(f). Claim 10 thus fails to recite any additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea, and as a result, claim 10 is also patent ineligible under 35 U.S.C. § 101.
Claim 11
Claim 11 is to a statutory category, as claim 11 is directed to a method, i.e. a process. However, claim 11 recites a judicial exception. In claim 11, “generating…a responsive…message according to the learned answer that addresses the question” is considered a mental process. Answering a question can practically be performed in the human mind.
Other than this mental process, claim 11 recites: “receiving…an incoming electronic message from a user, the incoming electronic message including a question; [and] routing…the incoming electronic message to trained autonomous agent.” Such tasks are indicative of insignificant extra-solution activity, i.e. mere data gathering, and are therefore insufficient to integrate the abstract idea into a practical application. See MPEP § 2106.5(g). Data gathering is also well-understood, routine and conventional and therefore does not amount to significantly more than the judicial exception. See, e.g., Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014).
Claim 11 also recites that the above-noted receiving and routing tasks are performed by an “electronic messaging system,” that the above-noted generating task is performed by “one or more processors of the electronic message system,” and that the trained autonomous agent comprises “a large language model trained using a set of actual question-answer threads as inputs to learn an answer that addresses the question.” However, such recitations represent no more than mere instructions to apply the mental process on a computer, and thus do not integrate the judicial exception into a practical application or amount to significantly more than the judicial exception. See MPEP § 2106.05(f). Particularly, the claim omits any details of the large language model or the training thereof; the model appears to be invoked merely as a tool for performing the judicial exception.
Consequently, claim 11 recites an abstract idea but does not include additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea. As a result, and for the reasons described above, claim 11 is rejected as being patent ineligible under 35 U.S.C. § 101.
Claim 12
In claim 12, “creating…a proposed answer to the question” and “evaluating…the proposed answer” are considered a mental process. Claim 12 additionally recites that the “creating” is performed “by the trained autonomous agent” and that the “evaluating” is performed “by the one or more processors using a scorer module.” However, these additional recitations represent no more than mere instructions to apply the mental process on a computer, and thus do not integrate the judicial exception into a practical application or amount to significantly more than the judicial exception. See MPEP § 2106.05(f).
Claim 13
In claim 13, the limitation reciting “when evaluating determines that the proposed answer does not satisfy a threshold criterion” is considered a mental process. The additional limitations therein (i.e. “discarding the generated responsive electronic message; and forwarding the incoming electronic message to a specific inbox or email address for manual answer generation”) are considered insignificant extra-solution activity, i.e. data gathering, and are insufficient to integrate the abstract idea into a practical application. See MPEP § 2106.5(g). Data gathering is also well-understood, routine and conventional and therefore does not amount to significantly more than the judicial exception. See, e.g., Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014). Claim 13 thus fails to recite any additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea, and as a result, claim 13 is also patent ineligible under 35 U.S.C. § 101.
Claim 14
In claim 14, the limitation reciting “when evaluating determines that the proposed answer does satisfy a threshold criterion” is considered a mental process. The additional limitation therein (i.e. “causing the generated responsive electronic message to be transmitted to the user”) is considered insignificant extra-solution activity, i.e. data gathering, and are insufficient to integrate the abstract idea into a practical application. See MPEP § 2106.5(g). Data gathering is also well-understood, routine and conventional and therefore does not amount to significantly more than the judicial exception. See, e.g., Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014). Claim 14 thus fails to recite any additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea, and as a result, claim 14 is also patent ineligible under 35 U.S.C. § 101.
Claim 15
Claim 15 is directed to a statutory category, as claim 15 is directed to a system, which is considered a machine or manufacture. However, claim 15 recites a mental process. In claim 15, “correlate each responsive electronic message with a given one of the incoming electronic messages as a question-answer thread” and “…learn an answer that addresses the question” are considered indicative of a mental process.
Other than this mental process, claim 15 recites: “to obtain incoming electronic messages from corresponding users, the incoming electronic messages each including a question, and to obtain responsive electronic messages to the corresponding users, the responsive electronic messages each including an answer to the question” and “to receive the question-answer thread for each correlated incoming and responsive electronic message pair.” Such tasks are indicative of insignificant extra-solution activity, i.e. mere data gathering, and are therefore insufficient to integrate the abstract idea into a practical application. See MPEP § 2106.5(g). Data gathering is also well-understood, routine and conventional and therefore does not amount to significantly more than the judicial exception. See, e.g., Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014).
Claim 15 also recites that an “electronic message module” performs the above-noted task of obtaining the incoming and responsive electronic messages, that a “thread processing module” performs the above-noted task of correlating the messages, and that “an agent training module” performs the above-noted task of receiving the question-answer threads and which is further “configured to train a large language model using a set of the questions-answer threads as inputs to learn an answer that addresses the question, and to store the trained large language model in a database of the computing system.” However, such recitations represent no more than mere instructions to apply the mental process on a computer, and thus do not integrate the judicial exception into a practical application or amount to significantly more than the judicial exception. See MPEP § 2106.05(f). Particularly, the claim omits any details of the agent training module and large language model; these appear to be invoked merely as a tool for performing the judicial exception.
Consequently, claim 15 recites an abstract idea but does not include additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea. As a result, and for the reasons described above, claim 15 is rejected as being patent ineligible under 35 U.S.C. § 101.
Claims 16 and 17
The additional elements recited in claims 16 and 17 are considered a mental process; as indicated above, correlating messages can practically be performed in the human mind. Claims 16 and 17 thus fail to recite any additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea, and as a result, claims 16 and 17 are also patent ineligible under 35 U.S.C. § 101.
Claim 18
Claim 18 recites that “the agent training module is configured to train the large language model by fine-tuning a previously trained model for a specific question-answer situation.” However, such a recitation represents no more than mere instructions to apply the mental process on a computer, and thus does not integrate the judicial exception into a practical application or amount to significantly more than the judicial exception. See MPEP § 2106.05(f). Claim 18 thus fails to recite any additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea, and as a result, claim 18 is also patent ineligible under 35 U.S.C. § 101.
Claim 19
The additional elements recited in claim 19 are considered a mental process. Processing messages to remove information therefore can practically be performed in the human mind. Claim 19 thus fails to recite any additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea, and as a result, claim 19 is also patent ineligible under 35 U.S.C. § 101.
Claim 20
Claim 20 recites that “performance of the training includes discarding a learned answer that does not satisfy a question-answer criterion.” However, such a recitation represents no more than mere instructions to apply the mental process on a computer, and thus does not integrate the judicial exception into a practical application or amount to significantly more than the judicial exception. See MPEP § 2106.05(f). Claim 20 thus fails to recite any additional elements that integrate the abstract idea into a practical application or that amount to significantly more than the abstract idea, and as a result, claim 20 is also patent ineligible under 35 U.S.C. § 101.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claims 1, 3, 4, 8, 15, 17, 18 and 19 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by the article entitled, “Leveraging Large Language Models for Generating Responses to Patient Messages” by Liu et al. (“Liu”).
Regarding claim 1, Liu generally teaches developing fine-tuned large language models to generate responses to patient messages sent via an electronic health record patient portal (see e.g. the ABSTRACT). Liu particularly teaches:
obtaining, by one or more processors of a computing system, incoming electronic messages from corresponding users, the incoming electronic messages each including a question (see e.g. “Data Collection and Preprocessing” within the “METHODS” section: Liu teaches extracting patient messages sent to primary care providers at a medical center. Liu demonstrates that the patient messages can each comprise a question – see e.g. “Evaluation Dataset” and Table 1 within the “METHODS” section. Liu further teaches that the messages are obtained by a computing system, understandably via one or more processors thereof, so as to fine-tune a large language model – see e.g. “Model Development” within the “METHODS” section. Accordingly, Liu teaches obtaining, by one or more processors of a computing system, incoming electronic messages from corresponding users, i.e. from patients, the incoming electronic messages each including a question.);
obtaining, by one or more processors of the computing system, responsive electronic messages to the corresponding users, the responsive electronic messages each including an answer to the question (see e.g. “Data Collection and Preprocessing” within the “METHODS” section: Liu teaches extracting provider responses corresponding to the patient messages sent to the primary care providers. Liu demonstrates that each response can include an answer to a corresponding patient message/question – see e.g. “Evaluation Dataset” and Table 1 within the “METHODS” section. Like the patient messages, the responses are also obtained by the computing system, understandably via one or more processors thereof, so as to fine-tune a large language model – see e.g. “Model Development” within the “METHODS” section. Accordingly, Liu teaches obtaining, by one or more processors of the computing system, responsive electronic messages to the corresponding users, the responsive electronic messages each including an answer to the question.);
correlating, by one or more processors of the computing system, each responsive electronic message with a given one of the incoming electronic messages as a question-answer thread (see e.g. “Data Collection and Preprocessing” within the “METHODS” section: as noted above, Liu teaches extracting provider responses corresponding to the patient messages sent to the primary care providers. The patient messages and corresponding provider responses form pairs within a training set used to fine-tune a large language model – see e.g. the ABSTRACT and the first paragraph of the “RESULTS” section. Accordingly, Liu teaches correlating, necessarily using the one or more processors, each responsive electronic message with a given one of the incoming patient electronic messages as a question-answer thread, i.e. as a patient message and provider response pair.);
routing, by one or more processors of the computing system, the question-answer thread for each correlated incoming and responsive electronic message pair to an agent training module of the computing system (see e.g. the ABSTRACT, “Model Development” within the “METHODS” section, and the first paragraph of the “RESULTS” section: Liu teaches that pairs of patient messages and corresponding provider responses are provided to a computing system so as to fine-tune a large language model. The hardware and/or software necessary for fine-tuning the large language model is considered an “agent training module” like claimed. Accordingly, Liu teaches routing, by one or more processors of the computing system, the question-answer thread for each correlated incoming and responsive electronic message pair, i.e. each patient message and provider response pair, to an agent training module of the computing system, which fine-tunes a large language model.);
performing, by one or more processors of the computing system using the agent training module, training of a large language model using a set of the question-answer threads as input to learn an answer that addresses the question (see e.g. the ABSTRACT, “Model Development” within the “METHODS” section, and the first paragraph of the “RESULTS” section: like noted above, Liu teaches that pairs of patient messages and corresponding provider responses are provided to a computing system so as to fine-tune a large language model. As further noted above, the hardware and/or software necessary for fine-tuning the large language model is considered an “agent training module” like claimed. Liu demonstrates that such fine-tuning enables the large language model to learn a response that addresses a corresponding question – see e.g. “Evaluation Dataset” and “Primary Care Physicians Review of Responses” within the “METHODS” section, and “Results of Physician Review of Responses,” Table 2 and “CLAIR-Short Generated Responses” within the “RESULTS” section. Accordingly, Liu teaches training, i.e. fine-tuning, a large language model using a set of the question-answer threads, i.e. using pairs of patient messages and corresponding provider responses, as input to learn an answer that addresses the question. The training is performed by one or more processors of the computing system using the agent training module, i.e. using the software/hardware necessary for fine-tuning the large language model.); and
storing the trained large language model in a database of the computing system (see e.g. the ABSTRACT, “Model Development” within the “METHODS” section, and the first paragraph of the “RESULTS” section: like noted above, Liu teaches that pairs of patient messages and corresponding provider responses are provided to a computing system so as to fine-tune a large language model. The fine-tuned large language model is necessarily stored, i.e. in a database, of the computing system so as to perform inference on additional, evaluation questions – see e.g. “Evaluation Dataset” within the “METHODS” section. Accordingly, Liu teaches storing the trained large language model in a database of the computing system.).
Accordingly, Liu teaches a method like that of claim 1.
As per claim 3, Liu further teaches that the correlating includes adding any new incoming message or new responsive message to a related question-answer thread when the new incoming or responsive messages are obtained before a dormancy threshold has been reached (see e.g. “Data Collection and Preprocessing” within the “METHODS” section: Liu discloses that, when multiple messages are sent by a patient or a provider prior to receiving a response, the multiple messages are combined. The period until a response to a message is received is considered a “dormancy threshold” like claimed. Accordingly, Liu teaches adding any new incoming or new responsive message to a related question-answer thread, particularly to a previous message in the thread, when the new incoming or responsive messages are obtained before a dormancy threshold has been reached, i.e. before a response to the previous message is received.). Accordingly, Liu further teaches a method like that of claim 3.
As per claim 4, Liu further teaches that training the large language model comprises fine-tuning a previously trained model for a specific question-answer situation (see e.g. “Data Collection and Preprocessing” and “Model Development” within the “METHODS” section: Liu discloses that the large language model is pretrained to gain basic conversation capabilities, and is then fine-tuned on the pairs of patient messages and corresponding provider responses. Accordingly, Liu teaches that training the large language model comprises fine-tuning a previously trained model. The pairs of patient messages and corresponding provider responses used to fine-tune the large language model understandably provide specific question-answer situations – see e.g. “Evaluation Dataset” and Table 1 within the “METHODS” section. Fine-tuning the large language model enables it to learn a response that answers a corresponding question – see e.g. “Evaluation Dataset” and “Primary Care Physicians Review of Responses” within the “METHODS” section, and “Results of Physician Review of Responses,” Table 2 and “CLAIR-Short Generated Responses” within the “RESULTS” section. Consequently, Liu further teaches that the previously trained model is particularly fine-tuned for specific question-answer situations.). Accordingly, Liu further teaches a method like that of claim 4.
As per claim 8, Liu further teaches preprocessing one or more of the incoming electronic messages or responsive electronic messages to remove selected information therefrom (see e.g. “Data Collection and Preprocessing” within the “METHODS” section: Liu discloses that the patient messages and provider responses are preprocessed with an automated deidentification pipeline that removes, for example, patient and provider names.). Accordingly, Liu further teaches a method like that of claim 8.
Regarding claim 15, Liu generally teaches developing fine-tuned large language models to generate responses to patient messages sent via an electronic health record patient portal (see e.g. the ABSTRACT). Liu particularly teaches a system comprising:
an electronic message module configured to obtain incoming electronic messages from corresponding users, the incoming electronic messages each including a question, and to obtain responsive electronic messages to the corresponding users, the responsive electronic messages each including an answer to the question (see e.g. “Data Collection and Preprocessing” within the “METHODS” section: Liu teaches extracting patient messages sent to primary care providers and provider responses corresponding to the patient messages. Liu demonstrates that the patient messages can each comprise a question and that each response can include an answer to a corresponding patient message/question – see e.g. “Evaluation Dataset” and Table 1 within the “METHODS” section. Liu further teaches that the messages and responses are obtained by a computing system, understandably via one or more processors thereof, so as to fine-tune a large language model – see e.g. “Model Development” within the “METHODS” section. The one or more processors of the system programmed for receiving the messages and responses is considered an “electronic message module” like claimed, which is configured to obtain incoming electronic messages from corresponding users, i.e. patients, the incoming electronic messages each including a question, and to obtain responsive electronic messages from providers to the corresponding users, the responsive electronic messages each including an answer to the question.);
a thread processing module configured to correlate each responsive electronic messages with a given one of the incoming electronic messages as a question-answer thread (see e.g. “Data Collection and Preprocessing” within the “METHODS” section: as noted above, Liu teaches extracting patient messages sent to primary care providers and provider responses corresponding to the patient messages. The patient messages and corresponding provider responses are correlated to form pairs within a training set used to fine-tune a large language model – see e.g. the ABSTRACT and the first paragraph of the “RESULTS” section. The one or more processors necessary for identifying the patient messages and the provider response corresponding each patient message, and for associating each patient message and corresponding provider response in a pair, i.e. in a question-answer thread, is considered a “thread processing module” like claimed.);
an agent training module configured to receive the question-answer thread for each correlated incoming and responsive electronic message pair, the agent training module being configured to train a large language model using a set of the question-answer threads as inputs to learn an answer that addresses the question, and to store the trained large language model in a database of the computing system (see e.g. the ABSTRACT, “Model Development” within the “METHODS” section, and the first paragraph of the “RESULTS” section: like noted above, Liu teaches that pairs of the patient messages and corresponding provider responses, i.e. question-answer threads, are provided to a computing system so as to fine-tune the large language model. Liu teaches that fine-tuning the large language model enables it to learn responses that answer corresponding questions – see e.g. “Evaluation Dataset” and “Primary Care Physicians Review of Responses” within the “METHODS” section, and “Results of Physician Review of Responses,” Table 2 and “CLAIR-Short Generated Responses” within the “RESULTS” section. The fine-tuned large language model is necessarily stored, i.e. in a database, of the computing system so as to later perform inference on additional, evaluation questions – see e.g. “Evaluation Dataset” within the “METHODS” section. The one or more processors necessary for receiving the pairs of patient messages and corresponding provider responses, for fine-tuning the large language model, and for storing the large language model is considered an “agent training module” like claimed. Accordingly, Liu teaches an agent training module configured to receive the question-answer thread for each correlated incoming and responsive electronic message pair, the agent training module being configured to train a large language model using a set of the question-answer threads as inputs to learn an answer that addresses the question, and to store the trained large language model in a database of the computing system.).
Consequently, Liu teaches a system like that of claim 15.
As per claim 17, Liu further teaches that the correlation includes addition of any new incoming message or new responsive message to a related question-answer thread when the new incoming or responsive messages are obtained before a dormancy threshold has been reached (see e.g. “Data Collection and Preprocessing” within the “METHODS” section: Liu discloses that, when multiple messages are sent by a patient or a provider prior to receiving a response, the multiple messages are combined. The period until a response to a message is received is considered a “dormancy threshold” like claimed. Accordingly, Liu teaches adding any new incoming or new responsive message to a related question-answer thread, particularly to a previous message in the thread, when the new incoming or responsive messages are obtained before a dormancy threshold has been reached, i.e. before a response to the previous message is received.). Accordingly, Liu further teaches a system like that of claim 17.
As per claim 18, Liu further teaches that the agent training module is configured to train the large language model by fine-tuning a previously trained model for a specific question-answer situation (see e.g. “Data Collection and Preprocessing” and “Model Development” within the “METHODS” section: Liu discloses that the large language model is pretrained to gain basic conversation capabilities, and is then fine-tuned on the pairs of patient messages and corresponding provider responses. Accordingly, Liu teaches that training the large language model comprises fine-tuning a previously trained model. The pairs of patient messages and corresponding provider responses used to fine-tune the large language model understandably provide specific question-answer situations – see e.g. “Evaluation Dataset” and Table 1 within the “METHODS” section. Fine-tuning the large language model enables it to learn a response that answers a corresponding question – see e.g. “Evaluation Dataset” and “Primary Care Physicians Review of Responses” within the “METHODS” section, and “Results of Physician Review of Responses,” Table 2 and “CLAIR-Short Generated Responses” within the “RESULTS” section. Like noted above, the one or more processors necessary for fine-tuning the large language model is considered an “agent training module” like claimed. Consequently, Liu further teaches that the agent training module is configured to train the large language model by fine-tuning a previously trained model for a specific question-answer situation.). Accordingly, Liu further teaches a system like that of claim 18.
As per claim 19, Liu further teaches that the system is configured to preprocess one or more of the incoming electronic messages or responsive electronic messages to remove selected information therefrom (see e.g. “Data Collection and Preprocessing” within the “METHODS” section: Liu discloses that the patient messages and provider responses are preprocessed with an automated deidentification pipeline that removes, for example, patient and provider names.). Accordingly, Liu further teaches a system like that of claim 19.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 2, 6 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over the article to Liu cited above, and also over the article entitled “Building a Chatbot: Fine-Tune LLMs with WhatsApp Data” by Daniel Pleus (“Pleus”).
Regarding claim 2, Liu teaches a method like that of claim 1, as is described above, which entails obtaining incoming electronic messages and responsive electronic messages, and correlating each responsive electronic message with a given one of the incoming electronic messages as a question-answer thread. Liu, however, does not disclose that the correlating includes tracking each given incoming electronic message and each responsive electronic message for a selected amount of time, and prior to routing the question-answer thread for each correlated incoming and responsive electronic message pair to the agent training module, discarding any subsequent message in that question-answer thread obtained after the selected amount of time, as is required by claim 2.
Pleus generally teaches obtaining a chat history that comprises incoming electronic messages and responsive electronic messages, correlating the messages as a thread (i.e. into a conversation), and fine-tuning a large language model understandably using a set of the threads as inputs:
PNG
media_image1.png
310
762
media_image1.png
Greyscale
PNG
media_image2.png
992
763
media_image2.png
Greyscale
As noted in the above excerpt, Pleus discloses that the messages were separated into conversations, wherein “[a]ll messages that were written within one hour of the previous message were grouped into a conversation.” Pleus thus teaches that the correlating includes tracking each message (including each incoming electronic message and each responsive electronic message) for a selected amount of time (i.e. 1 hour), and discarding any subsequent messages in that thread obtained after the selected amount of time (i.e. any messages obtained after 1 hour are not included in the conversation/thread).
It would have been obvious to one of ordinary skill in the art, having the teachings of Liu and Pleus before the effective filing date of the claimed invention, to modify the method taught by Liu such that the correlating instead entails tracking each incoming electronic message and each responsive electronic message for a selected amount of time, and prior to training/fine-tuning the large language model (i.e. prior to routing the question-answer thread for each correlated incoming and responsive electronic message pair to the agent training module), discarding any subsequent messages in that thread obtained after the selected amount of time, as is taught by Pleus. It would have been advantageous to one of ordinary skill to utilize such a combination, whereby the large language model is trained on the resulting thread, because it would enable the large language model to better emulate conversations (i.e. a series of exchanges between participants), as is taught by Pleus (e.g., Pleus recites: “Overall, the results were surprisingly good. The model knows when you want to chat and responds in that form…The answers are logical and often correct and also keep context over various messages.”). Accordingly, Liu and Pleus are considered to teach, to one of ordinary skill in the art, a method like that of claim 2.
Regarding claim 6, Liu teaches a method like that of claim 1, as is described above, which entails obtaining incoming electronic messages and responsive electronic messages, and correlating each responsive electronic message with a given one of the incoming electronic messages as a question-answer thread. Liu, however, does not disclose that the responsive electronic messages are associated with a single person, as is required by claim 6.
As demonstrated above (see the rejection for claim 2), Pleus generally teaches obtaining a chat history that comprises incoming electronic messages and responsive electronic messages, correlating the messages as a thread (i.e. into a conversation), and fine-tuning a large language model understandably using a set of the threads as inputs. Regarding the claimed invention, Pleus suggests that the responsive electronic messages are associated with (i.e. written to, or by) a single person (e.g. the author, or his girlfriend):
PNG
media_image3.png
174
766
media_image3.png
Greyscale
PNG
media_image4.png
356
754
media_image4.png
Greyscale
It would have been obvious to one of ordinary skill in the art, having the teachings of Liu and Pleus before the effective filing date of the claimed invention, to modify the method taught by Liu such that the responsive electronic messages are associated with a single person, as is taught by Liu. It would have been advantageous to one of ordinary skill to utilize such a combination because it would enable the large language model to emulate the single person, as is suggested by Pleus (e.g., Pleus recites, “So, I decided to export the seven years' worth of chat history with my girlfriend Lisa and train a model to imitate our conversations.”). Accordingly, Liu and Pleus are considered to teach, to one of ordinary skill in the art, a method like that of claim 6.
Regarding claim 16, Liu teaches a system like that of claim 15, as is described above, which comprises an electronic message module configured to obtain incoming electronic messages and responsive electronic messages, and a thread processing module configured to correlate each responsive electronic message with a given one of the incoming electronic messages as a question-answer thread. Liu, however, does not explicitly disclose that the correlation includes tracking each given incoming electronic message and each responsive electronic message for a selected amount of time, and wherein for a subsequent message in a given question-answer thread obtained after the selected amount of time, the thread processing module is configured to discard the subsequent message, as is required by claim 16.
As demonstrated above (see the rejection for claim 2), Pleus generally teaches obtaining a chat history that comprises incoming electronic messages and responsive electronic messages, correlating the messages as a thread (i.e. into a conversation), and fine-tuning a large language model understandably using a set of the threads as inputs. Like further indicated above, Pleus discloses that the messages are correlated by grouping messages that were written within one hour of a previous message into a conversation:
PNG
media_image4.png
356
754
media_image4.png
Greyscale
Pleus thus teaches that the correlation includes tracking each message (including each incoming electronic message and each responsive electronic message) for a selected amount of time (i.e. 1 hour), and for a subsequent message in a given question-answer thread obtained after the selected amount of time, the subsequent message is discarded (i.e. any messages obtained after 1 hour are not included in the conversation/thread).
It would have been obvious to one of ordinary skill in the art, having the teachings of Liu and Pleus before the effective filing date of the claimed invention, to modify the system taught by Liu such that the correlation instead includes tracking each incoming electronic message and each responsive electronic message for a selected amount of time, and for a subsequent message in a given question-answer thread obtained after the selected amount of time, the subsequent message is discarded (i.e. by the thread processing module). It would have been advantageous to one of ordinary skill to utilize such a combination, whereby the large language model is trained on the resulting thread, because it would enable the large language model to better emulate conversations (i.e. a series of exchanges between participants), as is taught by Pleus (e.g., Pleus recites: “Overall, the results were surprisingly good. The model knows when you want to chat and responds in that form…The answers are logical and often correct and also keep context over various messages.”). Accordingly, Liu and Pleus are considered to teach, to one of ordinary skill in the art, a system like that of claim 16.
Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over the article to Liu cited above, and also over WIPO Publication No. WO 01/17276 A2 to Jones et al. (“Jones”).
Regarding claim 5, Liu teaches a method like that of claim 1, as is described above, which entails obtaining incoming electronic messages and responsive electronic messages, and correlating each responsive electronic message with a given one of the incoming electronic messages as a question-answer thread. Liu, however, does not explicitly disclose that the responsive electronic messages are obtained from a shared inbox or group email address, as is required by claim 5.
Jones nevertheless generally teaches applying automation techniques to business email, particularly with respect to a group email address (see e.g. page 1).
It would have been obvious to one of ordinary skill in the art, having the teachings of Liu and Jones before the effective filing date of the claimed invention, to apply the method taught by Liu to business email obtained via a group email address like taught by Jones, i.e. where the incoming electronic messages would be addressed to the group email address of the business and the responsive electronic messages are obtained from the group email address. It would have been advantageous to one of ordinary skill to utilize this combination because it is important that such email “gets a response in a timely fashion,” as is taught by Jones (see page 1); applying the method taught by Liu to automatically generate responsive messages to incoming email messages would help in responding to the emails in a timely fashion, as would have been appreciated by one of ordinary skill in the art. Accordingly, Liu and Jones are considered to teach, to one of ordinary skill in the art, a method like that of claim 5.
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over the combination of Liu and Pleus introduced above, and also over U.S. Patent Application Publication No. 2019/0005021 to Miller et al. (“Miller”).
Regarding claim 7, Liu and Pleus teach a method like that of claim 6, as is described above, which includes obtaining incoming electronic messages and responsive electronic messages, wherein the responsive electronic messages are associated with a single person. Each responsive electronic message is correlated with a given one of the incoming electronic messages as a question-answer thread, which is used to train a large language model to learn an answer that addresses a question. Liu and Pleus, however, do not explicitly disclose that the trained large language model is configured for use as a virtual assistant for the single person, as is required by claim 7.
Miller nevertheless generally teaches employing one or more machine learning models as a virtual assistant for a user within a communication session to automatically generate responses to incoming messages in the communication session (see e.g. paragraphs 0022-0023). The automatically-generated responses are responses that the user would likely manually provide within the communication session (see e.g. paragraphs 0022 and 0024).
It would have been obvious to one of ordinary skill in the art, having the teachings of Liu, Pleus and Miller before the effective filing date of the claimed invention, to modify the method taught by Liu and Pleus such that the trained large language model is configured for use as a virtual assistant like taught by Miller for the single person. It would have been advantageous to one of ordinary skill to utilize such a virtual assistant because it would free up the person’s attention to attend to other activities, as is taught by Miller (see e.g. paragraph 0022). Accordingly, Liu, Pleus and Miller are considered to teach, to one of ordinary skill in the art, a method like that of claim 7.
Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over the article to Liu cited above, and also over the article entitled, “Smart Reply: Automated Response Suggestion for Email” by Kannan et al. (“Kannan”).
Regarding claim 9, Liu teaches a method like that of claim 8, as is described above, which entails obtaining incoming electronic messages and responsive electronic messages, and preprocessing one or more of the messages to remove selected information therefrom. Liu, however, does not explicitly teach that removing the selected information includes at least one of removing a signature block, removing quoted text, or removing an attachment, as is required by claim 9.
Kannan generally describes a method for automatically generating short email responses (see e.g. the ABSTRACT). Like Liu, Kannan particularly teaches that such a method comprises obtaining incoming electronic messages and responsive electronic messages, correlating each responsive electronic message with a given one of the incoming electronic messages as a thread, and training a model using a set of the threads to learn a response that addresses an incoming message (see e.g. section 3 “SELECTING RESPONSES” and section 7.1 “Data”). Kannan further teaches preprocessing one or more of the incoming electronic messages or responsive electronic messages to remove selected information therefrom, including by removing at least one of a signature block, quoted text or an attachment (see e.g. section 7.1 “Data”).
It would have been obvious to one of ordinary skill in the art, having the teachings of Liu and Kannan before the effective filing date of the claimed invention, to modify the method taught by Liu such that preprocessing the incoming electronic messages or responsive electronic messages comprises at least one of removing a signature block, removing quoted text or removing an attachment, as is taught by Kannan. It would have been advantageous to one of ordinary skill to utilize such a combination because it would put the messages in a better form for training the model, as is evident from Kannan (see e.g. section 7.1 “Data”). Accordingly, Liu and Kannan are considered to teach, to one of ordinary skill in the art, a method like that of claim 9.
Claims 10 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over the article to Liu cited above, and also over the article entitled “Learning with Bad Training Data via Iterative Trimmed Loss Minimization” by Shen et al. (“Shen”).
Regarding claims 10 and 20, Liu teaches a method like that of claim 1, as is described above, which entails training a large language model using a set of question-answer threads as inputs to learn an answer that addresses a question. Liu, however, does not explicitly teach that the training includes discarding a learned answer that does not satisfy a question-answer criterion, as is required by claims 10 and 20.
Shen nevertheless generally teaches, while training a machine learning model, discarding training samples with which the model produces outputs that have the largest loss (see e.g. the Abstract, Section 1. “Introduction” and Section 4 “Iterative Trimmed Loss Minimization”).
It would have been obvious to one of ordinary skill in the art, having the teachings of Liu and Shen before the effective filing date of the claimed invention, to modify the method taught by Liu such that the training comprises discarding training samples with which the model produces the largest loss (i.e. the learned answer output by the model for the training sample does not satisfy a question-answer criterion), as is taught by Shen. It follows that the learned answers for discarded the training samples would likewise be discarded. It would have been advantageous to one of ordinary skill to utilize such a combination because it would enable the model to still be trained when a fraction of the training samples are corrupted, as is taught by Shen (see e.g. the Abstract and Section 1. “Introduction”). Accordingly, Liu and Shen are considered to teach, to one of ordinary skill in the art, a method like that of claims 10 and 20.
Claims 11-14 are rejected under 35 U.S.C. 103 as being unpatentable over U.S. Patent No. 10,019,673 to Allen et al. (“Allen”), and also over the article entitled “Leveraging Large Language Models for Generating Responses to Patient Messages” by Liu et al. (“Liu”).
Regarding claim 11, Allen generally describes a method for generating responses to an electronic communication, particularly by using a question answering system (see e.g. column 1, lines 23-36). Allen teaches that such a method comprises:
receiving, by an electronic messaging system, an incoming electronic message from a user, the incoming electronic message including a question (see e.g. column 1, lines 23-36; and column 9, lines 48-57: Allen teaches receiving text from a first client for communication to a second client. Allen teaches that the text can comprise a question – see e.g. column 2, line 61 – column 3, line 9; column 3, lines 53-65; and column 6, line 60 – column 7, line 19. The received text from the first client is thus considered an incoming electronic message like claimed.);
routing, by the electronic messaging system, the incoming electronic message to an autonomous agent (see e.g. column 1, lines 23-36; column 2, line 61 – column 3, line 9; column 7, lines 35-52; and column 9, line 58 – column 10, line 10: Allen discloses that the received text is sent to a question-answering system, which generates candidate answers. The question-answering system is considered a type of autonomous agent, as it automatically generates and provides answers to the received questions.); and
generating, by one or more processors of the electronic messaging system, a responsive electronic message according to an answer that addresses the question (see e.g. column 1, lines 23-36; column 2, line 61 – column 3, line 9; column 7, lines 35-52; and column 9, line 58 – column 10, line 10: like indicated above, Allen discloses that the question-answering system generates an answer that addresses the text received from the first client. Allen discloses that the answer is then provided to the first and/or second client – see e.g. column 1, lines 23-36; column 2, line 61 – column 3, line 9; column 10, lines 19-53; and column 11, lines 17-26. The provided answer is considered a responsive electronic message like claimed, which is generated according to an answer that addresses the first client’s question.).
Allen thus teaches a method similar to that of claim 11. However, Allen does not explicitly disclose that the autonomous agent is a trained autonomous agent comprising a large language model, which is trained using a set of actual question-answer threads as inputs to learn an answer that addresses the question, as is required by claim 11.
Nevertheless, like described above, Liu generally teaches developing fine-tuned large language models to generate responses to patient messages sent via an electronic health record patient portal (see e.g. the ABSTRACT). In particular, Liu teaches training (i.e. fine-tuning) the large language model using a set of actual question-answer threads as inputs to learn answers that address different questions (see e.g. “Data Collection and Preprocessing” within the “METHODS” section: Liu teaches extracting actual patient messages sent to primary care providers along with corresponding responses. Liu suggests that the patient messages can each comprise a question and that each response can include an answer to a corresponding patient message/question – see e.g. “Evaluation Dataset” and Table 1 within the “METHODS” section. The patient messages and corresponding provider responses form pairs within a training set used to fine-tune a large language model – see e.g. the ABSTRACT and the first paragraph of the “RESULTS” section. Liu demonstrates that such fine-tuning enables the large language model to learn responses to corresponding questions – see e.g. “Evaluation Dataset” and “Primary Care Physicians Review of Responses” within the “METHODS” section, and “Results of Physician Review of Responses,” Table 2 and “CLAIR-Short Generated Responses” within the “RESULTS” section. Accordingly, Liu teaches training a large language model using a set of actual question-answer threads, i.e. using pairs of patent messages and provider responses, as inputs to learn answers that address different questions.).
It would have been obvious to one of ordinary skill in the art, having the teachings of Allen and Liu before the effective filing date of the claimed invention, to modify the method taught by Allen such that the autonomous agent comprises a large language model like taught by Liu, which is trained using a set of actual question-answer threads as inputs to learn an answer that addresses the received question. It would have been advantageous to one of ordinary skill to utilize such a large language model because it is able to generate responses similar to a user’s responses, as is evident from Liu (see e.g. the ABSTRACT, which recites “CLAIR-Short exhibited the ability to generate concise responses similar to provider’s responses.”). Accordingly, Allen and Liu are considered to teach, to one of ordinary skill in the art, a method like that of claim 11.
As per claim 12, Allen further teaches (i) creating, by the autonomous agent, a proposed answer to the question, and (ii) evaluating, by the one or more processors using a scorer module, the proposed answer (see e.g. column 1, lines 23-36; and column 9, line 48 – column 10, line 10: Allen teaches that the question-answering system, which is considered an autonomous agent, creates a set of candidate answers to the received question. Allen further teaches that confidence scores can be calculated for each candidate answer – see e.g. column 2, line 61 – column 3, line 9; and column 10, lines 11-18: The software executed by one or more processors necessary for calculating the confidence scores is considered a “scorer module” like claimed, which evaluates the proposed answers.). As described above, it would have been obvious to modify the method taught by Allen such that the autonomous agent comprises a trained large language model like taught by Liu. Accordingly, the above-described combination of Allen and Liu is further considered to teach a method like that of claim 12.
As per claim 13, Allen further teaches that, when evaluating the proposed answer and determining that the proposed answer does not satisfy a threshold criterion, the method further comprises (i) discarding the generated responsive electronic message, and (ii) and forwarding the incoming electronic message to a specific inbox or email address for manual answer generation (see e.g. column 10, lines 11-31: Allen discloses that candidate answers with a confidence score above a specified value can be provided to the second client in response to the text being communicated to the second client. It thus follows that candidate answers below the specified value would not be communicated to the second client. That is, Allen teaches discarding the generated responsive electronic message, i.e. candidate answer, in response to determining that it does not satisfy a threshold criterion, i.e. in response to determining the candidate answer does not have a confidence score above the specified value. Allen further suggests that the user can manually generate an answer, even if candidate answers are provided to the second client – see e.g. column 10, lines 19-39. It is thus apparent that the electronic message is forwarded to the second client, i.e. a specific inbox or email address, for manual answer generation, including when candidate answers do not satisfy the threshold criterion.). As described above, it would have been obvious to modify the method taught by Allen such that the autonomous agent comprises a trained large language model like taught by Liu. Accordingly, the above-described combination of Allen and Liu is further considered to teach a method like that of claim 13.
As per claim 14, Allen further teaches that, when evaluating the proposed answer and determining that the proposed answer does satisfy a threshold criterion, the method further includes causing the generated responsive electronic message to be transmitted to the user (see e.g. column 10, line 54 – column 11, line 26: Allen further teaches that generated candidate answers with confidence scores above a specified value can be provided to the first user/client. That is, in response to determining that the proposed answer satisfies a threshold criterion, i.e. that the confidence score of the candidate answer is above the specified value, the candidate answer is transmitted to the first user.). As described above, it would have been obvious to modify the method taught by Allen such that the autonomous agent comprises a trained large language model like taught by Liu. Accordingly, the above-described combination of Allen and Liu is further considered to teach a method like that of claim 14.
Conclusion
The prior art made of record on form PTO-892 and not relied upon is considered pertinent to applicant’s disclosure. The applicant is required under 37 C.F.R. §1.111(C) to consider these references fully when responding to this action. In particular, the article by Feng et al. cited therein (“Reply Using Past Replies—A Deep Learning-Based E-Mail Client”) describes a Smart E-mail Management System for handling the issue of e-mail overload, and which particularly incorporates models for automatically generating and suggesting replies for incoming email messages. The U.S. Patent to Goldberg et al. cited therein describes an information processing system for generating answers to questions, and which particularly stores and analyzes a corpus of a predetermined entity (e.g. a person) to derive an emulated answer to a question, the emulated answer including an emulation of an actual answer that would be provided by the predetermined entity. The U.S. Patent to Mancuso cited therein teaches utilizing language neural networks to automatically generate draft electronic communications for a user account that reflect a composition style of the user account and that accurately address a context of a communication thread.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BLAINE T BASOM whose telephone number is (571)272-4044. The examiner can normally be reached Monday-Friday, 9:00 am - 5:30 pm, EST.
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, Matt Ell can be reached at (571)270-3264. 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.
/BTB/
9/5/2026
/TAN H TRAN/Primary Examiner, Art Unit 2141
Prosecution Insights