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 .
DETAILED ACTION
Status of the Application
This action is in response to the Amendment filed on 5/8/2026, and is a Final Office Action. Claims 1-5, 8-15, 18-20, 22, 24-30 are pending in the application.
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 1-5, 8-15, 18-20, 22, 24-30 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.
Claim 1 is directed towards a server, thus meeting the Step 1 eligibility criterion. Claim 1 does recite the abstract concept of a commercial interaction/fundamental economic practice, which has been identified as an abstract idea by the MPEP. The relevant claimed limitations include: upon receiving of an image, identify a customer based on the received image, and acquire customer information corresponding to the customer / upon receipt of location information, determine a location of the customer in the store based on the location information / generate first text indicating the location of the customer and one or more attributes corresponding to the customer information / calculate a semantic vector of the item text that is generated / stores promotional text data in which each of a plurality of promotional texts for items sold in the store is associated with a corresponding semantic vector indicating a meaning of the promotional text, for a semantic vector that is most approximate to the calculated semantic vector of the item text and select second text for promoting a first item indicating by the item text, the second text being the promotional text associated with the semantic vector that is most approximate to the calculated semantic vector / transmit the second text selected as output promotional text in place of the item text. Applicant’s Spec. further describes the context of the claimed invention as pertaining to the commercial interaction realm, and describes the claimed invention as seeking to, when implemented, at best optimize a business practice/goal: “For example, the item text generation unit 205 inputs the query text generated by the query text generation unit 204 into the LLM for sales promotion 242 stored in the storage unit 24. The item text generation unit 205 acquires the text data output from the LLM for sales promotion 242 as the item text”, “As a result, the information processing system S can generate query text considering information on a customer and a location in a store where the customer is located. By query text considering information about a customer or a location in a store where a customer is located is entered into the LLM for sales promotion 242 as a generative AI to generate item text, it is possible to generate item text optimized for the customer. Further, by generating promotional text obtained by converting the item text into a representation suitable for the promotion of the item, it is possible to promote the item unified with a more appropriate representation. Further, since the information processing system S causes the mobile terminal 1 used by the customer to display the promotional text, the information processing system S can provide a promotion service for an item regardless of where the customer is in the store. In other words, according to the information processing system S, it is possible to improve the convenience of promotion to which the generative AI is applied.” Claim 1 also recites the abstract concept of a mental concept – i.e. mental process that can be performed in the human mind or using pen/paper, including an observation/evaluation/judgment, which has been identified as an abstract idea by the MPEP: generate first text indicating the location of the customer and one or more attributes corresponding to the customer information / calculate a semantic vector of the item text that is generated / search for a semantic vector that is most approximate to the calculated semantic vector of the item text and select second text for promoting a first item indicated by the item text, the second text being the promotional text associated with the semantic vector that is most approximate to the calculated semantic vector. These claimed limitations, under their broadest reasonable interpretation, cover performance in the human mind but for the recitation of generic computing elements- see below, thus still being in the mental process category.
This judicial exception is not integrated into a practical application. Claim 1 includes the additional elements of a network interface connectable to a customer terminal in the store / a memory/ a customer terminal / a processor configured to execute a program stored in the memory / first and second database that store data / store server/ analyzing and determining data using the machine learning model (‘input the first text to a machine learning model that has been trained to generate, item text indicating one of the items sold in the store and to be promoted to a customer of a particular attribute at a particular location in the store ‘, ‘selected from the second database as output promotional text in place of the item text that is generated by and output from the machine learning model’). The network interface / customer terminal/memory/processor /databases / store server represent generic computing elements. Using a machine learning model to determine/generate data does no more than apply or link the use of the recited judicial exception to a particular technological environment. The additional elements do not, alone or in combination, improve the functioning of the computing device or another technology/technical field, nor do they apply or use the judicial exception in some other meaningful way beyond generally linking its use to a particular technological environment. The claim is directed to an abstract idea.
Claim 1 does not include additional elements that are sufficient to amount to significantly more than the judicial exception, because as noted above, the claimed computing elements represent generic computing elements; they are recited at a high level of generality. Using a machine learning model to determine/generate data does no more than apply or link the use of the recited judicial exception to a particular technological environment. The additional elements do not, alone or in combination, improve the functioning of the computing device or another technology/technical field, nor do they apply or use the judicial exception in some other meaningful way beyond generally linking its use to a particular technological environment. Therefore, Claim 1 does not amount to significantly more than the abstract idea itself. The claim is not patent eligible.
Independent claims 11, 20 are directed to a method and system, respectively, for performing the method of claim 1; the claims recite the same abstract idea of Claim 1. Claims 11, 20 perform the method of claim 1 using only generic components of a networked computer system. Therefore, claims 11, 20 are directed to an abstract idea without significantly more for the reasons given in the discussion of claim 1.
Remaining dependent claims 2-5, 8-10, 12-15, 18-19, 22, 24-30 further recite and narrow the abstract ideas of independent claim 1. The claims further recite the additional elements of using a generative LLM to analyze/determine data (see claims 2/12), training a machine learning model and using the model to analyze/determine data (see claims 3/5/13/15), outputting voice sound (see claims 9/19), a POS terminal. The POS terminal represents a generic computing element that is recited at a high level of generality. Using a generative LLM to analyze/determine data does no more than apply or link the use of the recited judicial exception to a particular technological environment. Training a machine learning model and using the model to analyze/determine data does no more than apply or link the use of the recited judicial exception to a particular technological environment. Outputting audio content does no more than apply or link the use of the recited judicial exception to a particular technological environment/ field of use. The additional elements do not, alone or in combination with the other additional elements, improve the functioning of the computing device or another technology/technical field, or apply or use the judicial exception in some other meaningful way beyond generally linking its use to a particular technological environment. Therefore, the claims above do not amount to significantly more than the abstract idea itself. The claims are not patent eligible.
Relevant prior art
The prior art of record does not teach neither singly nor in combination the limitations of pending claims 1-5, 8-15, 18-24. The most relevant prior art identified, Ellison (20150026010), teaches: upon a receipt of an image from a customer terminal, identify the customer based on the image, and acquire customer information, upon receipt of location data from the terminal, determine the customer location in the store based on the location data, generate first text indicating the customer location and attributes of the customer, one of the items sold in the store and to be promoted for a customer, generate item text indicating one of the items sold in the store and to be promoted to a customer of a particular attribute at a particular location in the store. However, it lacks the combination of claimed elements of the pending independent claims. Smith (20190272557) teaches inputting the first text to a machine learning model that has been trained to generate item text indicating one of the items sold in the store and to be promoted to a customer of a particular attribute at a particular store location, using a first database in which items are associated with store locations, selecting second text for promoting a first item based on the item text that is generated by and output from the machine learning model from a second database that stored promotional text for each of the items sold in the store, control a network interface to transmit the second text to the customer terminal. However, it lacks the combination of claimed elements of the pending independent claims.
When taken as a whole, the claims are not rendered obvious as the available prior art does not suggest or otherwise render obvious the noted features nor does the available prior art suggest or otherwise render obvious further modification of the evidence at hand. Such modifications would require substantial reconstruction relying solely on improper hindsight bias, and thus would not be obvious.
Response to Arguments
Applicant’s arguments have been fully considered; Applicant argues with substance:
The Office Action asserts that the limitations of calculating a semantic vector and searching a database by comparing semantic vectors can be performed in the human mind. However, the amended claims recite that the store server is to "calculate a semantic vector of the item text that is generated by and output from the machine learning model" and to "search a second database, which stores promotional text data in which each of a plurality of promotional texts for items sold in the store is associated with a corresponding semantic vector indicating a meaning of the promotional text, for a
semantic vector that is most approximate to the calculated semantic vector of the item text." A semantic vector is a high-dimensional array of floating-point numbers that encodes the meaning of a text string in a vector space. "Reminders on evaluating subject matter eligibility of claims under 35 U.S.C. 101" memorandum dated August 4, 2025 (hereafter "the Memo") clarifies that "a claim does not recite a mental process when it contains limitation(s) that cannot practically be performed in the human mind." Memo, page 2. A human cannot practically calculate a semantic vector for a text string, nor can
a human practically perform a nearest-neighbor search ("most approximate") across a database of high-dimensional vectors to identify the closest match. These steps are not observations, evaluations, or judgments that a person could perform mentally or with pen and paper. Applicant therefore submits that the claims do not recite a mental process. Regarding the "commercial interaction" grouping, the claims do not recite the concept of a sale or promotion. The claims recite a specific data processing pipeline
involving a machine learning model, calculation of a semantic vector from the output of
that model, a nearest-neighbor search of a vector-indexed database, and the selection
and transmission of curated text from the database in place of the model's output. These
are technical data processing steps, not commercial activities.
Step 2A, Prong Two
Even if the claims were found to recite an abstract idea under Prong One, the
claimed subject matter integrates any such exception into a practical application because
the claims reflect an improvement to the functioning of a computer. The Memo instructs examiners to "evaluate the claim to ensure it reflects the disclosed improvement." Memo, page 4. The Specification discloses a specific technical problem inherent to generative Al models. As described at paragraph [0095], "the LLM for sales promotion 242 generates different texts as item texts each time. That is, even
if text promoting the same item is generated, the contents thereof are different each time."
As further described at paragraph [0096], "there is a possibility that text including a
difficult-to-understand expression is generated, or text including an inappropriate
expression is generated when a promotion is performed in a store." This is a technical
deficiency of the generative model itself, specifically the non-deterministic nature of its
output, not a business problem. The amended claims recite a specific technical solution to this technical problem. Amended claim 1 requires the store server to input the first text to a machine learning
model to generate item text, but the item text is not transmitted to the customer terminal.
Instead, the store server calculates "a semantic vector of the item text that is generated
by and output from the machine learning model," searches a second database of precurated
promotional texts indexed by semantic vectors for "a semantic vector that is most
approximate to the calculated semantic vector of the item text," and selects "the
promotional text associated with the semantic vector that is most approximate to the
calculated semantic vector." The store server then transmits "the second text selected
from the second database as output promotional text in place of the item text that is
generated by and output from the machine learning model." As described at paragraph [0097], "The output promotional text is text data obtained by converting the item text into a representation suitable for promoting the item text in the store." As further described at paragraph [0135], "Since the information processing system S generates the output promotional text using the generated item text
and the promotional text DB 245 storing the promotional text appropriate for the promotion
of the item, the information processing system S can promote the item with a more
appropriate expression to the customer." This architecture addresses the identified technical problem through a specific technical mechanism. The machine learning model's non-deterministic output is used
solely as an intermediate computational input. That intermediate output is converted
through a mathematical operation, specifically semantic vector calculation, into a query
key. That key is used to perform a nearest-neighbor search against a pre-indexed vector
database of curated texts. The curated text retrieved from the database, not the model's
raw output, is what is transmitted to the customer terminal. This pipeline constrains and
stabilizes the output of the non-deterministic generative model through deterministic
vector-based retrieval, ensuring that the text delivered to the customer terminal is
controlled and consistent regardless of the variability in the model's output.
The Memo states that "claims that are determined to improve computer capabilities
or improve technology or a technical field support a finding that the claim integrates the judicial exception into a practical application." Memo, page 5. The claimed architecture
improves the functioning of a generative Al system by addressing the inherent unreliability
of its output through a specific post-processing pipeline. This is not merely applying an
abstract idea using a generic computer. The claim recites a particular technical
implementation, namely semantic vector computation and nearest-neighbor retrieval
against a vector-indexed database, to achieve a particular technical result, namely stable,
reliable output from a non-deterministic model. The Office Action states that "selecting and presenting targeted product content represents a business practice/goal, not other technology/technical field; thus, improving this practice pertains to a business practice optimization, not to an improvement to other
technology/technical field." However, this conflates the domain of application with the
nature of the improvement. The technical problem addressed by the claims, that a
generative Al model produces different, potentially unsuitable text each time it is invoked,
exists regardless of whether the model is used for sales promotion, medical reporting,
legal drafting, or any other application. The technical solution, using the model's output
as an intermediate key for vector-based retrieval of curated text, likewise operates at the
level of the computing architecture, not the business practice. The improvement is to how
the computing system processes and constrains generative Al output, not to how a store
promotes items. The Memo further notes that the analysis should consider whether "the claim
covers 'a particular solution to a problem' using a specific technical implementation."
Memo, page 4. Amended claim 1 does not claim the abstract idea of stabilizing Al output.
It claims a specific implementation: calculating a semantic vector from the machine
learning model's generated text, performing a nearest-neighbor search of a database that
stores promotional texts each associated with a pre-computed semantic vector, selecting
the promotional text associated with the most approximate vector, and transmitting that
selected text in place of the model's output. This is a specific technical solution to a
specific technical problem, not a result-oriented claim that preempts all approaches to the
problem. Amended claims 11 and 20 recite substantially similar features in method and
system form and are eligible for the same reasons. Therefore, the claimed subject matter integrates any recited judicial exception into a practical application, and the claims are patent eligible under Step 2A, Prong Two. By this Response, new claims 25-30 are added. As discussed above, amended
independent claims 1 and 11 are patent eligible, and by virtue of their dependence, new
claims 25-30 are patent eligible for at least the same reasons. Additionally, each new
claim recites features that further demonstrate the technical nature of the claimed
invention. New claims 25 and 28 recite determining an area in the store in which the customer
is located based on the location information and the first database, comparing the
determined area with a previously determined area corresponding to previously received
location information, and performing the generating of the first text in response to a
determination that the areas are different. This is a state-aware, event-driven triggering
mechanism that avoids redundant invocation of the machine learning model and the
downstream vector computation pipeline by conditioning text generation on a detected
change in the customer's area. This is a technical design choice that governs when the
computing system performs processing, not a commercial interaction or a mental process.
New claims 26 and 29 recite that the plurality of promotional texts stored in the
second database are prepared in advance for the items sold in the store, and that each
of the corresponding semantic vectors is computed from a respective one of the
promotional texts prior to the calculating of the semantic vector of the item text. This
makes explicit that the second database is a pre-indexed vector store, and that the
runtime operation is a similarity search of the item text vector against an index of precomputed
vectors. Pre-computing and indexing vectors for later retrieval is a specific
data structure and retrieval architecture, not an abstract concept or a step that can be
performed in the human mind. New claims 27 and 30 recite that the memory stores a plurality of templates for generating the first text, and that one of the plurality of templates is selected according to
a keyword capable of discriminating a genre of an item, and the first text is generated
using the selected template. This is a programmatic prompt construction mechanism in
which the input to the machine learning model is assembled from stored templates
selected by keyword-based classification, not composed by a human as free-form text.
Template selection by keyword matching to control the input to a generative model is a
specific technical implementation for structured data processing, not a mental process or
a commercial activity. Accordingly, entry and favorable consideration of the new claims are respectfully
requested.
The pending claims do recite an abstract idea, and the additional elements do not, alone or in combination, integrate the judicial exception into a practical application, nor do they represent significantly more than the abstract idea itself, as noted above. Applicant’s Spec. further describes the context of the claimed invention as pertaining to the commercial interaction realm, and describes the claimed invention as seeking to, when implemented, at best optimize a business practice/goal: “For example, the item text generation unit 205 inputs the query text generated by the query text generation unit 204 into the LLM for sales promotion 242 stored in the storage unit 24. The item text generation unit 205 acquires the text data output from the LLM for sales promotion 242 as the item text”, “As a result, the information processing system S can generate query text considering information on a customer and a location in a store where the customer is located. By query text considering information about a customer or a location in a store where a customer is located is entered into the LLM for sales promotion 242 as a generative AI to generate item text, it is possible to generate item text optimized for the customer. Further, by generating promotional text obtained by converting the item text into a representation suitable for the promotion of the item, it is possible to promote the item unified with a more appropriate representation. Further, since the information processing system S causes the mobile terminal 1 used by the customer to display the promotional text, the information processing system S can provide a promotion service for an item regardless of where the customer is in the store. In other words, according to the information processing system S, it is possible to improve the convenience of promotion to which the generative AI is applied.” Selecting and presenting targeted product content represents a business practice/goal, not other technology/technical field; thus, improving this practice pertains to a business practice optimization, not to an improvement to other technology/technical field. The pending claimed limitations pertaining to using a trained model/LLM model to analyze/determine data do no more than apply or link the use of the recited judicial exception to a particular technological environment, as noted above. The pending claims , when implemented, do not improve the functioning of the computing device itself, or other technology/technical field (including the technical field of machine learning); the pending claims do not recite an improved training method or machine learning architecture. There is no technical support/technical evidence in the Spec. paras noted above by the Applicant that the claimed invention, when implemented, improves the functioning of the computing device itself or other technology/technical field. See Office Action above for the detailed, reasoned 35 USC 101 analysis.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any extension fee 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 ALEXANDRU CIRNU whose telephone number is (571)272-7775. The examiner can normally be reached on M-F 9:00am-5pm. If attempts to reach the examiner by telephone are unsuccessful, the examiner's supervisor, Ilana Spar can be reached on (571) 270-7537. The fax phone number for the organization where this application or proceeding is assigned is 571- 273-8300. Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
Sincerely,
/Alexandru Cirnu/
Primary Patent Examiner, Art Unit 3622
5/12/2026
Prosecution Insights