DETAILED ACTION
Status of the Application
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claims 1-20 are currently pending and have been examined.
Claim Objections
Claims 1-20 are objected to because of the following informalities: Claims 1, 13, and 20 recite "the prompt further comprises and text" (bold emphasis) should read "the prompt further comprises text"; and Claim 6, "one or more of the content item" should read "one or more of the content items"(bold emphasis). Claims 2-12 and 14-19 depend from claims 1 and 13 above and therefore inherit the claim objection of their parent claim. Appropriate correction is required.
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 without significantly more.
Step 1: Is the claim to a process, machine, manufacture or composition of matter? (MPEP 2106.03)
In the present application, claims 1-12 are directed to a method (i.e., a process), claims 13-19 are directed to a computer product (i.e., an article of manufacture), and 20 is directed to an apparatus (i.e., a machine). Thus, the eligibility analysis proceeds to Step 2A.1.
Step 2A. prong one: Does the claim recite an abstract idea, law of nature, or natural phenomenon? (MPEP 2106.04)
While claims 1, 13, and 20, are directed to different categories, the language and scope are substantially the same and have been addressed together below.
The abstract idea recited in claims 1, 13, and 20, is
A method comprising: receiving a plurality of content items, wherein each content item is associated with a content creator sponsored by a user;
generating a prompt for each of the content items, wherein the prompt includes free text describing instructions to generate a risk score for the respective content item and an explanation for the generated risk score in response to being provided with the prompt, wherein the prompt further comprises and text describing or associated with the respective content item;
receiving a response for each generated prompt, wherein each response comprises the respective risk score and respective explanation for the respective risk score;
generating a dashboard based on the received response, wherein the dashboard includes a row for each content item, wherein each row describes at least one of the content creator who created the content item, the risk score associated with the content item, and a link to view the content item; and
transmitting instructions to display the dashboard.
The claimed invention is directed to an abstract idea of evaluating content for compliance with rules/guidelines and assigning a risk score.
Under the broadest reasonable interpretation, the claim limitations above recite steps of reviewing a piece of content (e.g., a social media post), comparing it against a set of instructions or guidelines (e.g., brand safety standards), determining a risk score based on that comparison, and formulating an explanation for the score are concepts that can practically be performed in the human mind or by a person using a pen and paper. Human moderators have been historically performing the above-mentioned steps before the invention of computers. Thus, the claims recite an abstract idea consistent with the “mental processes” grouping of the abstract ideas, set forth in MPEP 2106.04(a)(2)(III).
Additionally, the above-mentioned limitations are fundamentally directed to managing commercial risk (i.e., mitigating risk) for brand safety and suitability for an advertiser sponsoring a content creator. Managing commercial risk and applying administrative rules to content are fundamental economic practices and commercial interactions, consistent with the “certain methods of organizing human activity” grouping of the abstract ideas, set forth in MPEP 2106.04(a)(2)(II).
Accordingly, the above-mentioned limitations are considered as a single abstract idea, therefore, the claims recite an abstract idea and the analysis proceeds to Step 2A. prong two.
Step 2A. prong two: Does the claim recite additional elements that integrate the judicial exception into a practical application? (MPEP 2106.04)
This judicial exception is not integrated into a practical application because the additional elements merely add instructions to apply the abstract idea to a computer.
The additional elements considered include:
Claim 1: “by a risk assessment system,” “from one or more online platforms,” “to a generative language model,” “that cause the generative language model,” “from the generative language model,” “user interface,” “transmitting instructions to a client device to display the dashboard user interface.”
Claim 13: “A non-transitory computer-readable storage medium storing instructions that, when executed, cause a processor to perform steps comprising:” “by a risk assessment system,” “from one or more online platforms,” “to a generative language model,” “that cause the generative language model,” “from the generative language model,” “user interface,” “transmitting a second set of instructions to a client device to display the dashboard user interface.”
Claim 20: “A system comprising: a processor; and a non-transitory computer-readable storage medium storing instructions that, when executed, cause the processor to perform steps comprising:” “by a risk assessment system,” “from one or more online platforms,” “to a generative language model,” “that cause the generative language model,” “from the generative language model,” “user interface,” “transmitting a second set of instructions to a client device to display the dashboard user interface.”
In particular, the claim only recites the additional elements - the use of risk assessment system, online platform, generative language model, and client device to receive, generate, transmit, and display information. The computer in the steps is recited at a high-level of generality (i.e., See Applicant’s Specification at least at paragraphs [0015]-[0018] and Fig. 1. generic computer components performing a generic computer function; In para. [0002] and [0023]-[0024] the specification acknowledges the invention riles on the inherent capabilities of the off-the-shelf large language models to understand natural language) such that it amounts to no more than mere instructions to apply the exception using a generic computer component.
That is, the function of limitations [A]-[E] are steps of adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea as discussed in MPEP 2106.05(f). The combination of these additional elements is no more than mere instructions to apply the exception using a generic computer.
Accordingly, even in combination, these additional element(s) do not integrate the abstract idea into a practical application because they do not improve a computer or other technology, do not transform a particular article, do not recite more than a general link to a computer, and do not invoke the computer in any meaningful way; the general computer is effectively part of the preamble instruction to “apply” the exception by the computer. Therefore, the claims are directed to an abstract idea and the analysis proceeds to Step 2B.
Step 2B: Does the claim recite additional elements that amount to significantly more than the judicial exception? (MPEP 2106.05)
The claim(s) does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the bold portions of the limitations recited above, were all considered to be an abstract idea in Step2A-Prong Two. The additional elements and analysis of Step2A-Prong two is carried over. For the same reason, these elements are not sufficient to provide an inventive concept. Applicant has merely recited elements that instruct the user to apply the abstract idea to a computer or other machinery. When considered individually and in combination the conclusion, as discussed above with respect to integration of the abstract idea into a practical application, the additional element of using a computer to perform the above-mentioned limitations of [A]-[E] amount to no more than mere instructions to apply the function of the limitations to the exception using generic computer component, as discussed in MPEP 2106.05(f). The claims as a whole merely describes how to generally “apply” the concept for evaluating content for compliance with rules/guidelines and assigning a risk score. Thus, viewed as a whole, nothing in the claim adds significantly more (i.e., an inventive concept) to the abstract idea. For these reasons there is no inventive concept in the claims and thus are ineligible.
The dependent claims do not recite addition limitations that overcome the 101 rejection. As for dependent claims 2, 9, and 14 recites the content includes images, video, or audio. Processing different generic data does not change the abstract idea (inventive concept). As for dependent claims 3, 5-8, 11, 12, 15, 18, and 19 further recite additional abstract steps and information regarding generic user interface elements (interactive elements), specific types of guidelines (words, phrases, topics), or specific formats for the score (category or number). These are mere data classification and GUI display, which do not change the abstract idea of the independent claims, nor provide additional element integrates the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself. As for dependent claims 4 and 16 further recites receiving different scores based on whether content is allowed or not. This is merely reciting expected logical outcome of abstract idea under mental process. Claims 10 and 17 recite comparing the score to a threshold and performing a remedial action (e.g., removing the content, removing an association, or flagging for review), which further define the abstract idea for administrative rules performable by human moderators, and does not change the abstract idea of the independent claims.
Therefore, claims 1-20 are rejected under 35 U.S.C. 101.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries set forth in Graham v. John Deere Co., 383 U.S. 1, 148 USPQ 459 (1966), that are applied for establishing a background for determining obviousness under pre-AIA 35 U.S.C. 103(a) are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-3, 7, 10-15, and 17-20 are rejected under 35 U.S.C. 103 as being unpatentable over Goeldi (US 20100119053 A1) in view of Tunstall-Pedoe et al. (US 20230274094 A1, hereinafter, “Tunstall-Pedoe”).
Claim 1, Goeldi discloses a method (Abstract) comprising:
receiving, by a risk assessment system, a plurality of content items from one or more online platforms (para. [0054] “harvesting layer 100 includes locating or discovering social media sources (e.g., websites) from the Internet related to a particular industry or other category, and harvesting the relevant content from those sources.” Goelidi discloses a social media analytics platform (a risk assessment system) that ingests and aggregates social media posts (content items) from various online platforms), wherein each content item is associated with a content creator sponsored by a user (para. [0050]-[0051], Goeldi discloses a social medial analytics platform (risk assessment system) that harvests online social media conversations (content items) form online platforms, including tracking posts authored by opinion leaders/influencers (content creators) regarding a specific brand/product for an enterprise or advertising agency (sponsored by a user));
generating a dashboard user interface based on the received response, wherein the dashboard user interface includes a row for each content item, wherein each row describes at least one of the content creator who created the content item, the risk score associated with the content item, and a link to view the content item (para. [0083], [0084], and [0089], Goeldi discloses generating a dashboard GUI that displays lists (rows) of individual posts. The lists display the author, a calculated sentiment index/rating (risk score), and link (hyperlink) to drill down the original content); and
transmitting instructions to a client device to display the dashboard user interface (para. [0049] discloses user interface to be browser-based interface transmitted to user device. In para. [0083]-[0084] discloses displaying of dashboard user interface).
While Goeldi in para. [0069] discloses the use of algorithmic natural language processing and sentence structure analysis for keyword matching, Goeldi fails to explicitly disclose the sentiment (risk) score is generated by generating a prompt to a generative language model that includes free text describing instructions and the text of the content item.
Specifically, Goeldi fails to expressly disclose the limitations (italic emphasis included):
generating a prompt to a generative language model for each of the content items, wherein the prompt includes free text describing instructions that cause the generative language model to generate a risk score for the respective content item and an explanation for the generated risk score in response to being provided with the prompt, wherein the prompt further comprises and text describing or associated with the respective content item;
receiving a response from the generative language model for each generated prompt, wherein each response comprises the respective risk score and respective explanation for the respective risk score;
based on the received response from the generative language model.
However, Tunstall-Pedoe is in the similar field that teaches systems and methods for the processing of content data including social media content, using a generative large language model (LLM) to assess risk, which Tunstall-Pedoe specifically teaches:
generating a prompt to a generative language model for each of the content items, wherein the prompt includes free text describing instructions that cause the generative language model to generate a risk score for the respective content item and an explanation for the generated risk score in response to being provided with the prompt, wherein the prompt further comprises and text describing or associated with the respective content item (Examiner note: for the purpose of prior art analysis, the Examiner interpreted the typographical error “comprises and text” to be “comprises text”. In para. [0187], [0603]-[0604], [0608], [0611], [0623], and [0858]-[0874], Tunstall-Pedoe teaches providing a prompt to an LLM that includes free-text instructions follow by the raw text of the content item such as social media postings for analysis for scoring);
receiving a response from the generative language model for each generated prompt, wherein each response comprises the respective risk score and respective explanation for the respective risk score; …based on the received response from the generative language model (para. [0608]-[0609] Tunstall-Pedoe teaches the LLM analysis generates an overall score (risk score) used to take action on the post, as well as natural language explanation detailing why the post was flagged).
Therefore, it would have been obvious for one of ordinary skill in the art, before the effective filling of the invention to modify the social media moderation dashboard system and method of Goeldi to substitute the traditional NLP keyword matching scoring algorithm with the generative language model (LLM) prompting system as taught by Tunstall-Pedoe, for the motivation of improving the accuracy (para. [0036], [0103]), context-awareness, and semantic understanding of the risk assessment system, as recognized in Tunstall-Pedoe that by using its sematic LLM techniques, “abusive content can be identified automatically in a way that is superior to prior art methods. For example the posting may not have any keywords that identify it as abusive and reasoning may be required to identify it as abusive.” (Para. [0605]).
Claim 13, Goeldi discloses a non-transitory computer-readable storage medium storing instructions that, when executed, cause a processor to perform steps (para. [0110]-[0113]).
Claim 13 recites the same limitations as claim 1. The combination of Goeldi and Tunstall-Pedoe teaches these limitations for the same reasons set forth above, mutatis mutandis respectively.
Claim 20, Goeldi discloses a system comprising: a processor; and a non-transitory computer-readable storage medium storing instructions that, when executed, cause the processor to perform steps (para. [0110]-[0113]).
Claim 20 recites the same limitations as claim 1. The combination of Goeldi and Tunstall-Pedoe teaches these limitations for the same reasons set forth above, mutatis mutandis respectively.
Claims 2 and 14, the combination of Goeldi and Tunstall-Pedoe makes obvious of the method of claim 1 and the non-transitory computer-readable storage medium of claim 13, further teaches
wherein the content item includes one or more of free text, an image, a video, and audio (Goeldi, para. [0004], “Online social media encompasses online media such as blogs and sub-blogs, online discussion forums, social networks, wiki sites such as Wikipedia, online reviews on e-commerce sites such as Amazon.com®, video sites such as YouTube®, micro-blogging services such as Twitter®,” which includes one or more of free text, image, video, and audio. Additionally and alternatively, Tunstall-Podoe para. [0607], “classifying the posting or identifying non-text content in the posting —such as the content of images, videos or audio”).
Claims 3 and 15, the combination of Goeldi and Tunstall-Pedoe makes obvious of the method of claim 1 and the non-transitory computer-readable storage medium of claim 13, Goeldi further discloses
generating a settings user interface including one or more interactive elements; and causing a user device to present the user interface (Goeldi para. [0082], “a GUI is utilized to present the quantified and analyzed online social media content in a manner relevant to the user. The GUI may be fully customizable giving users the ability to select which charts and graphs should appear on the login page of the interface. The GUI provides an intuitive display to visualize brand, product or service sentiment over time. This display is a quantitative measure of opinion or sentiment for a brand, product, services, or its competitors and is derived from an automated aggregation of sentiment ratings on each individual post to online social media about a brand, product, services and/or those of their competitors. The GUI includes various knobs or switches to manipulate the above information in a variety of ways.”).
Tunstall-Pedoe further teaches, configured to receive the free text describing the instructions for generating the risk score (para. [0187], [0603]-[0604], [0608], [0611], [0623], and [0858]-[0874], teaches inputting free-text/natural language instructions such as tenets via a user interface to generate scoring).
The rationales to modify the teachings of Goeldi with/and the teachings of Tunstall-Pedoe are presented in the examining of independent claim 1 and incorporated herein.
Claim 7, the combination of Goeldi and Tunstall-Pedoe makes obvious of the method of claim 3. Tunstall-Pedoe further teaches,
wherein the instructions for generating the risk score include instructions on how to assign a number or categorical value to a respective content item (para. [0547], “some examples, the prompt will request the LLM to provide a classification of Yes, No or Unknown. In other examples, a degree of certainty can be asked for.” teaches the prompt to the LLM includes explicit instructions on how to format the categorical or numerical output).
Claims 10 and 17, the combination of Goeldi and Tunstall-Pedoe makes obvious the method of claim 1 and the non-transitory computer-readable storage medium of claim 13, Goeldi further discloses,
comparing, for each content item, the risk score of the respective content item to a risk threshold (Goeldi, para. [0074], teaches comparing generated sentiment/risk scores to a threshold to trigger system actions); and
in response to the risk score exceeding the risk threshold, performing a remedial action in relation to the content item (Tunstall-Pedoe, para. [0608] teaches the generated overall score to automatically trigger a decision to take action on the content),
wherein the remedial action includes one or more of removing the content item from the one or more online platforms, removing an association with the content item to the user, and flagging the content item for review by the user (Tunstall-Pedoe, para. [0608] teaches the automated actions taken based on the score include hiding (removing) the content item and flagging it for human review).
Therefore, it would have been obvious for one of ordinary skill in the art, before the effective filling of the invention to modify the threshold-based trigger logic for social media moderation dashboard system and method of Goeldi to include LLM-based scoring and remedial actions as taught by Tunstall-Pedoe, for the motivation of eliminating the need for manual human review of every post, thereby improving the efficiency, scalability, and response time of the risk assessment system (para. [0604], “Many social media applications need to identify abusive posts and many operate at a scale where human identification of such posts is not practical.”).
Claims 11 and 18, the combination of Goeldi and Tunstall-Pedoe makes obvious of the method of claim 1 and the non-transitory computer-readable storage medium of claim 13, Goeldi further discloses,
wherein each row further describes an online platform on which the respective content item was posted, a type of the respective content item, a topic associated with the respective content item (para. [0084], [0090], and [0105], discloses the dashboard’s list/rows of content can be filtered by and describe the source/platform, the type of post (e.g., positive/negative/neutral), and the topic (e.g., product, service, or brand).
Claims 12 and 19, the combination of Goeldi and Tunstall-Pedoe makes obvious of the method of claim 1 and the non-transitory computer-readable storage medium of claim 13, Goeldi further discloses,
wherein the risk score is a category or a number (para. [0087] discloses the score is a number (index) and a category (positive/negative/neutral)).
Claims 4-6, 8, 9, and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Goeldi (US 20100119053 A1) in view of Tunstall-Pedoe et al. (US 20230274094 A1, hereinafter, “Tunstall-Pedoe”) and further in view of Mysore et al (US 12111754 B1, hereinafter, “Mysore”).
Claims 4 and 16, the combination of Goeldi and Tunstall-Pedoe makes obvious of the method of claim 1 and the non-transitory computer-readable storage medium of claim 13. The combination of Goeldi and Tunstall-Pedoe teaches prompting an LLM with free-text instructions (tenets/rules) to generate a risk score and explanation for social media posts.
The combination teaches (italic emphasis),
wherein the free text further includes one or more guidelines provided via the user device (Tunstall-Pedoe para. [0580], [0623] teaches the free-text instructions (tenets) are provided via user interface), wherein the guidelines indicate what type of content is allowed in the content item and what type of content is not allowed in the content item (Tunstall-Pedoe, para. [0648] recites “These example tenets can be split into 2 categories: 1-4 are goal-like tenets (as mentioned above, these specify things to optimise) while 5-11 are constraints (typically preventing bad behaviour). The goal-like tenets provide the system with a way of generating actions that it should carry out [allowed] and the constraint tenets then provide a way of preventing bad actions [not allowed].” which Tunstall-Pedoe teaches the instructions/guidelines explicitly dictate what content is allowed (goals) and what content is not allowed (constraints)),
the method of claim 1 further comprising: receiving a first response associated with a first content item that includes content that is not allowed, wherein the response includes a first risk score (Tunstall-Pedoe para. [0608] teaches the LLM evaluates a content item that breaches the guidelines/policies (i.e., content that is not allowed), it generates a response that includes an overall score indicating the violation (a first risk score)); and
receiving a second response associated with a second content item includes only content that is allowed, wherein the second response includes a second risk score that is lower than the first risk score (Goeldi para. [0087] teaches generating differential scores based on the evaluation of the content, wherein negative/violating content receives one score and positive/allowed content receives a different, lower-risk score).
However, the combination fails to expressly teach the claim language of guidelines (italic emphasis):
wherein the free text further includes one or more guidelines provided via the user device, wherein the guidelines indicate what type of content is allowed in the content item and what type of content is not allowed in the content item,
the method of claim 1 further comprising: receiving a first response associated with a first content item that includes content that is not allowed, wherein the response includes a first risk score; and receiving a second response associated with a second content item includes only content that is allowed, wherein the second response includes a second risk score that is lower than the first risk score.
Nonetheless, Mysore is in the similar field of AI application with predefined guidelines for compliance which specifically teaches,
wherein the free text further includes one or more guidelines provided via the user device, wherein the guidelines indicate what type of content is allowed in the content item and what type of content is not allowed in the content item (Mysore, Col. 14 lines 42-58, teaches validating AI applications by obtaining specific, user-provided guidelines that define the operational boundaries of the AI. In Col. 12 lines 5-18 teaches the provided guidelines establish boundaries that indicates what the AI model is permitted to generate or accept (i.e., allowed vs. not allowed content)),
the method of claim 1 further comprising: receiving a first response associated with a first content item that includes content that is not allowed, wherein the response includes a first risk score; and receiving a second response associated with a second content item includes only content that is allowed, wherein the second response includes a second risk score that is lower than the first risk score (Col. 19 lines 32-62 teaches generating a probability score indicating how well the outcome aligns with the expected guidelines for example, LLM prompted to evaluate content against safety guidelines will output higher risk score for a first content item containing abusive/prohibited material and lower risk score for second content item for safe/allowed material).
Therefore, it would have been obvious for one of ordinary skill in the art, before the effective filling of the invention to modify the LLM-based risk assessment system and method of Goeldi and Tunstall-Pedoe to include the user provided guidelines of Mysore for the motivation of allowing highly customizable validation of social media content against an advertiser’s specific organizational polices, (Mysore, Col. 9 lines 29-33, “allowing organizations to implement customized security measures and compliance policies tailored to the organization's specific needs.”). By allowing a user to input free-text guidelines defining what is allowed and not allowed, the moderation dashboard of Goeldi becomes more adaptable to different brands and allowing more business opportunities.
Claim 5, the combination of Goeldi, Tunstall-Pedoe, and Mysore makes obvious of the method of claim 4. Tunstall-Pedoe further teaches,
wherein the type of content that is not allowed in the content item includes one or more particular words, phrases, topics, or depictions (para. [0630], “If the system believes that the action is prohibited by the tenets, the action will not be completed. If the tenets allow the action, the action is then performed. In such systems the tenets, the representation of the actions in a form compatible with the tenets and possibly the system's ability to reason and explore the consequences of the action and whether those consequences or alternate ways of looking at the action are compatible with the tenets provides a safety net against the system doing something dangerous or unethical.” para. [0634], “It can also represent constraints such as not helping users break the law [topic] or never using profane language [particular words or phrases].” Para. [0931] teaches refusing to answer the question or generate the inappropriate content, “providing information that could be physically dangerous for the user or simply breaking policies that the provider of such a system might have such as not advocating for one political viewpoint over another.” Para. [0603]-[0604], “Identifying Abusive or Untrue Postings in Social Media… Abusive posts can include postings or media which are racist or otherwise offensive to users, depict things which are disturbing [depictions], are illegal, have national security or crime implications, break intellectual property rights, propagate false information in a way that is damaging, are defamatory or otherwise break the rules of the application or website where they appear”).
Claim 6, the combination of Goeldi, Tunstall-Pedoe, and Mysore makes obvious of the method of claim 4. Goeldi further discloses,
the content item being created by one or more particular content creators (Goeldi, para. [0105] discloses filtering and analyzing content based on the specific identity of the content creator (e.g., opinion leaders or specific authors)), and
the content being related to one or more particular topics (para. [0085] discloses filtering and assessing content based on particular topics (e.g., specific brands, products, or services)).
Tunstall-Pedoe further teaches:
wherein a subset of the guidelines are specific guidelines for content items that meet one or more requirements (Tunstall-Pedoe, para. [0732] teaches a specific subset of guidelines only when certain conditional requirements are met),
the requirements including one or more of the content items including a particular media item (Tunstall-Pedoe, para. [0607], [0732], teaches applying specific rules/guidelines when the content items include a particular type of media (audio track, images, or video)),
the content item being created by one or more particular content creators (Tunstall-Pedoe, para. [0850] teaches applying guidelines based on the specific identity or details of the person involved in the communication), and
the content being related to one or more particular topics (para. [0850] teaches applying specific guidelines to avoid or handle particular topics (e.g., adult subjects)).
The rationales to modify the teachings of Goeldi with/and the teachings of Tunstall-Pedoe are presented in the examining of independent claim 1 and incorporated herein.
Claim 8, the combination of Goeldi and Tunstall-Pedoe makes obvious of the method of claim 3. The combination fails to expressly teach,
wherein the instructions for generating the risk score include one or more standards, wherein the one or more standards indicate unacceptable types of content within content items.
Nonetheless, Mysore is in the similar field of AI application with predefined guidelines for compliance which specifically teaches,
wherein the instructions for generating the risk score include one or more standards, wherein the one or more standards indicate unacceptable types of content within content items (Mysore, Col. 14 lines 42-58, teaches validating AI applications by obtaining specific, user-provided guidelines that define the operational boundaries of the AI. In Col. 12 lines 5-18 teaches the provided guidelines establish boundaries that indicates what the AI model is permitted to generate or accept (i.e., allowed vs. not allowed content)).
Therefore, it would have been obvious for one of ordinary skill in the art, before the effective filling of the invention to modify the LLM-based risk assessment system and method of Goeldi and Tunstall-Pedoe to include the user provided guidelines of Mysore for the motivation of allowing highly customizable validation of social media content against an advertiser’s specific organizational polices, (Mysore, Col. 9 lines 29-33, “allowing organizations to implement customized security measures and compliance policies tailored to the organization's specific needs.”). By allowing a user to input free-text guidelines defining what is allowed and not allowed, the moderation dashboard of Goeldi becomes more adaptable to different brands and allowing more business opportunities.
Claim 9, the combination of Goeldi, Tunstall-Pedoe, and Mysore makes obvious of the method of claim 8. Tunstall-Pedoe further teaches,
wherein the prompt further comprises an image or video associated with the respective content item (Tunstall-Pedoe, para. [0448], [0607], teaches the automated sematic analysis system for identifying abusive posting evaluates non-text content including images and videos, by processing the content of those images/videos into the UL for risk assessment).
The rationales to modify the teachings of Goeldi with/and the teachings of Tunstall-Pedoe are presented in the examining of independent claim 1 and incorporated herein.
Relevant Prior Art Not Relied Upon
The prior art made of record and not relied upon is considered pertinent to Applicant’s disclosure. The additional cited art, including but not limited to the excerpts below, further establishes the state of the art at the time of Applicant’s invention and shows the following was known:
O’Neill (US 20240412542 A1) is directed to system and method for using AI to analyze social media content, which includes the use of LLM for generating scoring of social media content.
Centner (US 20160294753 A1) is in the similar field and directed to a system for facilitating integrity based communications among users of a social network, that teaches the threshold value in the context of removing a social media post.
A. Gopalan, V. Mohanavel, A. V. A. Geo, G. V. Rajkumar, T. Kavitha and P. P, "Experimental Evaluation of Robust Cyberbullying Detection over social media using Intelligent Learning Scheme," 2023 International Conference on Research Methodologies in Knowledge Management, Artificial Intelligence and Telecommunication Engineering (RMKMATE), Chennai, India, 2023, pp. 1-7, doi: 10.1109/RMKMATE59243.2023.10368747; teaches detecting negative semantics on social media using Intelligent learning scheme.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to WENREN CHEN whose telephone number is (571)272-5208. The examiner can normally be reached Monday - Friday 10AM - 6PM.
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, Nathan C Uber can be reached on (571) 270-3923. 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.
/WENREN CHEN/Primary Examiner, Art Unit 3626
Prosecution Insights