DETAILED ACTION
A. This action is in response to the following communications: Amendment filed: 05/26/2026. This action is made Final.
B. Claims 1-20 and 22 remain pending.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-20 and 22 is/are rejected under 35 U.S.C. 103 as being unpatentable over Howard, Todd A (US Pub. 2021/0232632 A1), herein referred to as “Howard” in view of Malkiel, Itzik et al. (US Pub. 2025/0238629 A1), herein referred to as “Malkiel” in further view of Sadr, Arash et al. (US Pub. 2024/0202796 A1), herein referred to as “Sadr”.
As for claims 1, 16 and 19, Howard teaches. A system and corresponding method of claim 16 and non-transitory computer readable medium of claim 19 storing instructions that when executed by a processor, programs the processor to, comprising: a processor programmed to (par. 243 hardware environment used to execute application):
automatically generate content comprising text content, visual content, and/or audio content (fig. 4A, item 401 a beholder (or user of the system) makes a request; par. 34-35 subject matter prompts (herein SMP) are comprised of text, audio and visual input) in response to an input prompt (paragraphs 120-122 the user inputs (prompts) images and captions (text) is automatically generated based upon prompts. In paragraph 192 the user (beholder) prompts (requests) and automated audio is generated. In paragraph 150-151 user (beholder) prompts (request) communicates multiple, tangible data structures related to the subject matter desired in the virtual experience, a beholder request may be initiated, generated, or modified in a variety of ways, including, but not limited to, as a result of a user's interaction with user interface elements of a user experience device, application, web site, or mobile device “app,” as described in regard to FIG. 1 and FIG. 6. The beholder request can also be generated via automated processes resulting from automated agents; in such the subject matter is prompted into the system and virtual experience including images, captions and sound is automatically generated for beholder (user);
perform, based on a harmonization model and/or consistency model, a harmonization check and/or a consistency check on the content (par. 156-157, fig. 4B, item 414, par. 4 For example, techniques and systems described herein advantageously provide technical methods for the extraction and separation of information facets from the base content returned from element sources, thus transforming content and metadata from the technical form usually returned from an element source (e.g., JSON, XML, HTML, images, video files, document files, and other proprietary formats) into discrete content elements that can be individually and separately used to construct a facet-segmented repository (and, ultimately, a bespoke virtual experience container). Searching a facet segmented repositories (FCRs) of selected target identities (prompts from user) to retrieve discrete content elements(DCEs) (objects) where primary concept aligns with target primary and secondary concepts (prompts) and stored separately as Cache set A and B);
recognize, based on the harmonization check and/or the consistency check, a conflict to be corrected (fig. 4B, 416 par. 158 Supplement, refine, and/or reduce cache set A 413 using cache set B 415, forming cache set C 417. Cache set C contains discrete content elements (DCEs) having content with a particular conceptual classification stemming from the beholder request and related to the subject matter context; the system is able to add to make refinements too and/or reduce Cache set A with Cache set B to create new Cache set C);
identify a property of the content that should be changed based on the recognized conflict (Cache set C created; par. 169 FIG. 4A, a virtual experience container may be assembled from the selected discrete content elements in accordance with the user experience device parameters (403)); and
generate a corrective action based on the property of the content that should be changed (par. 215 FIG. 4A, the virtual experience container may be provided to the user experience device (404)) .
Howard does not specifically teach automatic prompt generating; however in the same field of endeavor Malkiel teaches automatically generate, via a prompt generator (par. 95 automatic submitting a forward prompt), a verification prompt based on contextual and/or semantic information associated with the content (par. 95 804 obtaining primary answer generated by first language model and submitting a backward prompt for that is used as a function of verification, (determine hallucination)); execute a language model with the verification prompt to determine whether one or more portions of the content are inconsistent across one or more modalities (par. 95 acquiring 808 at least one candidate question generated by the backward language model set and getting a digital vector similarity measurement between primary question and candidate question)
perform, based on a harmonization model and/or consistency model that accesses output of the executed language model, a harmonization check and/or a consistency check on the content, the harmonization check and/or the consistency check comprising a comparison of at least one of: (i) different versions of a content element using a difference representation and/or (ii) numerical embedding representations of the content element generated at different times; recognize, based on the harmonization check and/or the consistency check, a conflict to be corrected including at least one of a semantic conflict, a visual conflict, or a mood conflict (par. 95 assigning 812 a digital hallucination extent to the primary answer computed from at least digital vector similarity measurement this is collected without any input 452 which represents human review 454 of either primary answer or digital vector similarity measurement by use of the standalone language model, first language model, backward language model and other text embedding model).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Malkiel into Howard since Malkiel suggests in paragraphs 3 and 4 Language models can be useful for computational tasks such as machine translation between natural languages, generation of text in natural languages, generation of source code text in programming languages, device control, image generation, information retrieval, information classification, information summarization, and many other computational tasks that involve communication efforts in written or spoken or other form. The capabilities of many language models depend at least in part on the training data which is used to train the model in question. The training data is embedded into the language model in an encoded form, so that more training on a given topic tends to produce a model which performs better on that topic than another language model which is less well trained on the topic. However, despite advances, language models sometimes produce undesirable, irrelevant, or insufficient results. Accordingly, improvements in technical areas involving language models would be beneficial.
Howard as modified by Malkiel does not specifically teach the combination of automatically generate, via a prompt generator, a verification prompt based on contextual and/or semantic relationships between a plurality of portions of the content across one or more modalities, wherein the verification prompt is configured to prompt a language model to determine whether the plurality of portions of the content are semantically consistent with one another across the one or more modalities: execute the language model with the verification prompt to determine whether the plurality of portions of the content are inconsistent across the one or more modalities; however in the same field of endeavor Sadr teaches in figure 4 and paragraphs 106-108 which describes example machine-learned model processing and search system 400 that has external inputs 404 and learned information 402 that is utilized to engineer a prompt 408 and is processed by an image generation model 410 to generate model generated images which can then be searched. The stored and learned information from external input includes fashion knowledge, personalization based upon stored data associated with user, trends relating to purchase, social media trends and search trends this data is obtained and utilized to generate a prompt and/or display suggested selectable button input for prompt selections. The external inputs come from more than one modalities such as image, video, audio and terms. The learned information 402 and external inputs 404 can be used to generate a palette 406 of data to be selected for prompt generation. For example the palette can include suggested images, suggested music and suggested terms all of which can be selected, these suggestions come from different modalities (i.e. pictures, audio, text) and are related/ have relationship based upon learned information 402, in which these suggestions are used to engineer a prompt 408 that can be processed with an image generation model 410 to generate a plurality of model-generated images that can be provided for display on canvas. Verification takes place with one or more model-generated images that can be selected. The one or more selected model-generated images can be processed with a search block 414 which can include object recognition, image matching, image embedding, and/or searching. The search block 414 can include feature recognition and search to determine one or more resources that include similar data. In some implementations, the search block 414 can utilize a multitask unified model (Pandu Nayak, “MUM: A new AI milestone for understanding information,” GOOGLE: THE KEYWORD (May 18, 2021). Additionally and/or alternatively, the model-generated images displayed in the canvas interface 412 may be interacted with to modify 418 a model-generated image (e.g., make a variation of a depicted object and/or to redo at least a portion of the image).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Sadr into Howard as modified by Malkiel, this is true because Sadr suggests in paragraph 4 , the utilization of artificial intelligence techniques to generate images and/or other datasets can be non-intuitive, may be open-ended, and may be time consuming. Current image generation systems utilize a prompt input box for receiving freeform text to be processed to generate one or more images. However, as a user utilizes the prompt input box, the user may struggle with which words to utilize and/or may be dissatisfied with the generated image as one or more of the input words may not be utilized in the direction the user desired (e.g., “houndstooth” may be entered by the user in association with the pattern; however, the model may generate an image with a dog's teeth).
As for claims 2, 17 and 20, Howard teaches. The system of claim 1, wherein the content comprises multi-modal content and wherein to recognize the conflict to be corrected, the processor is programmed to: identify an inconsistency between text in the multi-modal content and a visual in the multi-modal content (par. 46 the SMP is too broad and fixes the prompt from beholder; a subject matter prompt may be developed iteratively or interactively with the beholder, for example, to narrow down an open-ended prompt or suggest options to a beholder. For instance, the beholder request component 105 may, at the direction of the virtual experience service 120, present additional options on the UED 100 to the beholder 101 after an initial beholder request has been formulated. For example, if the prompt “Dad as a soldier” elicits too much content, the beholder request component 105 might render a prompt asking the beholder “How about Dad's experiences during the Korean War?”).
As for claims 3 and 18, Howard teaches. The system of claim 1, wherein the content comprises multi-modal content and wherein to recognize the conflict to be corrected, the processor is programmed to:
identify an inconsistency between a first visual in the multi-modal content and a second visual in the multi-modal content (par. 48 beholder input of visual content “A beholder can also be creating, using, or consuming content (such as a document, article, photo, or video) in another application and can indicate a subject matter prompt using an indication motif provided across the entire user experience device by the beholder request component 105.” Par. 49 visual content can be sent by user experience device (UED) to the virtual experience service to create a virtual experience container (object) content can be text, audio, photo, video etc.…).
As for claim 4, Howard teaches. The system of claim 1, wherein the content comprises a prompt for input to a language model, and wherein to recognize the conflict, the processor is programmed to: identify an inconsistency between one or more first words in the prompt and one or more second words in the prompt (par. 57, 156-158; searching (using queries or full text) may be supplemented with advanced search techniques and operators, such as fuzzy search, proximity search, word stemming, synonym databases, and acronym databases; in some cases, therefore, a primary concept facet being “aligned with” the target primary concept may not mean “strictly identical to”. Supplement, refine, and/or reduce cache set A 413 using cache set B 415, forming cache set C 417. Thus various inconsistencies can be found within subject matter prompt to create a new cache set of results to return to UED. A virtual experience service 120 (e.g., using context analysis component 121 or other subcomponents) may interact with or direct requests to content interpretation service(s) 130 to assist in the identification of concepts in various kinds of content, including subject matter prompts, beholder-provided content, content repository media, and information feeds. Content interpretation service(s) 130 can be used to, for example: identify the grammatical or semantic structure of text, discern key concepts in text, translate text, and identify entities in text; classify objects or places or identify people in images, caption images, perform video and speech interpretation to elicit concepts, identify a speaker, translate speech, identify and track faces, and index content; and analyze the “sentiment” expressed by speech, text, or images).
As for claim 5, Howard teaches. The system of claim 4, wherein to generate a corrective action, the processor is programmed to: generate a recommendation to modify the prompt based on recognized conflict (par. 65 query formulation can change inconsistencies with a query submitted by beholder to communicate with repositories more effectively. Transforming the subject matter context and other content into query terms and formulating queries from the query terms are described more fully with respect to the example process flow in FIG. 2A. It should also be noted that content and information searching capabilities of the query formulation and search module(s) 122 may be employed incrementally, iteratively, in multiple stages, synchronously and asynchronously, and by multiple sub-components of the virtual experience service 120 (e.g., by context analysis 121, content element deconstruction 123, and experiential synthesis 125 components).
As for claim 6, Howard teaches. The system of claim 4, wherein to generate a corrective action, the processor is programmed to: modify the prompt based on recognized conflict (par. 64 query formulation modifies the beholders query to make it more effective for the repositories it is communicating with).
As for claim 7, Howard teaches. The system of claim 1, wherein the content comprises a prompt for input to a language model, and wherein to recognize the conflict, the processor is programmed to: identify an inconsistency between one or more words in the prompt and one or more words in a previous prompt (par. 158Cache set A and cache set B may be processed together using set arithmetic, such as unions, joins, and intersections. Searches and filtering (including Boolean logic) may be applied to either cache set A or B to select DCEs for cache set C. For example, DCEs for cache set C may be selected by taking the union of cache set A and cache set B to form cache set C; DCEs for cache set C may be selected by taking the union of cache set A and B and then removing the DCEs that do not share a common content reference pointer, effectively meaning that a particular content segment must contain both a primary and supplemental concepts to be included in cache set C. par. 73-75 a virtual experience system/service needs to transform the structure of the query terms into a query language form suitable for functioning with a particular database service or to a form compatible with the particular API of a targeted element source).
As for claim 8, Howard teaches. The system of claim 7, wherein to generate a corrective action, the processor is programmed to: generate a recommendation to modify the prompt based on recognized conflict (par. 73-75 a virtual experience system/service needs to transform the structure of the query terms into a query language form suitable for functioning with a particular database service or to a form compatible with the particular API of a targeted element source).
As for claim 9, Howard teaches. The system of claim 7, wherein to generate a corrective action, the processor is programmed to: modify the prompt based on recognized conflict (par. 73-75 a virtual experience system/service needs to transform the structure of the query terms into a query language form suitable for functioning with a particular database service or to a form compatible with the particular API of a targeted element source).
As for claim 10, Howard teaches. The system of claim 1, wherein to recognize the conflict to be corrected, the processor is programmed to: identify and store an object in the content; and determine a difference between a first version of the object and a second version of the object, wherein the difference indicates the conflict to be corrected (par. 65 virtual experience service 120 performs deconstruction of the obtained content into content elements (e.g., via 123) and re-classification, synthesis, and (in some cases) dimension-wise or facet-wise expansion of the content elements (e.g., via 125 and 126) into a unified virtual experience container 150 tuned to the capabilities of the user experience device 100. This can be seen in figure 4B with cache sets A and B used to create cache set C as described in analysis of claim 1 above).
As for claim 11, Howard teaches. The system of claim 1, wherein to recognize the conflict to be corrected, the processor is programmed to: identify and store an object in the content; and generate a first embedding for the object at a first time; generate a second embedding for the object at a second time; and determine a difference between the first embedding and the second embedding, wherein the difference indicates the conflict to be corrected (par. 87 The content reference pointer can reference the physical or logical location of a single content file, a set of multiple content files, a compound content file, a content file that embeds the desired content, a specific range in a content file (e.g., the range of video in a file from time location 1:00:12.36 to 1:01:18:04), and/or a location in a database, data stream, or other data structure).
As for claim 12, Howard teaches The system of claim 10, wherein to identify and store the object in the content, the processor is programmed to: store the object in a graph database that includes one or more other objects in a timeline of the content (par. 88 A schema of element facets contains a canonical structural organization for how discrete content elements are classified so that they can be used to construct the facet-segmented repository; par. 74 virtual experience system can sometimes visualize its data as a graph using a graph API).
As for claim 13, Howard teaches. The system of claim 1, wherein to identify a property of the content that should be changed based on the recognized conflict, the processor is further programmed to: determine a first mood of a first portion of the content; determine a second mood of a second portion of the content; determine that the first mood and the second mood conflict with one another; wherein to generate the corrective action, the processor is programmed to: modify the first portion and/or the second portion so that the first mood is consistent with the second mood
(par. 129-132 Sentiment analysis of content is used to determine an emotion (mood) of content to build sentiment facets to data in order to select and deselect content elements for virtual experience container this is paired with previous teachings as another layer into collecting and modifying content based upon SMP inputted; par. 48 beholder input of visual content “A beholder can also be creating, using, or consuming content (such as a document, article, photo, or video) in another application and can indicate a subject matter prompt using an indication motif provided across the entire user experience device by the beholder request component 105.” Par. 49 visual content can be sent by user experience device (UED) to the virtual experience service to create a virtual experience container (object) content can be text, audio, photo, video etc…)..
As for claim 14, Howard teaches. The system of claim 13, wherein the first portion comprises first text and the second portion comprises second text, and wherein the processor is further programmed to: determine the first mood based on the first text; and determine the second mood based on the second text (par. 129-132 Sentiment analysis of content is used to determine an emotion (mood) of content to build sentiment facets to data in order to select and deselect content elements for virtual experience container this is paired with previous teachings as another layer into collecting and modifying content based upon SMP inputted; par. 46 the SMP is too broad and fixes the prompt from beholder; a subject matter prompt may be developed iteratively or interactively with the beholder, for example, to narrow down an open-ended prompt or suggest options to a beholder. For instance, the beholder request component 105 may, at the direction of the virtual experience service 120, present additional options on the UED 100 to the beholder 101 after an initial beholder request has been formulated. For example, if the prompt “Dad as a soldier” elicits too much content, the beholder request component 105 might render a prompt asking the beholder “How about Dad's experiences during the Korean War?”).
As for claim 15, Howard teaches. The system of claim 13, wherein the first portion comprises text and the second portion comprises an image, and wherein the processor is further programmed to: determine the first mood based on the text; and determine the second mood based on the image (par. 129-132 Sentiment analysis of content is used to determine an emotion (mood) of content to build sentiment facets to data in order to select and deselect content elements for virtual experience container this is paired with previous teachings as another layer into collecting and modifying content based upon SMP inputted; par. 135 For example, if Mom (from the immediately prior example) also posted, along with the photo, “Y′all KNOW I just LOVE cheeseburgers!” the surface meaning of this text might be interpreted as expressing a positive sentiment toward the concept “cheeseburgers,” though the impression given by the photo and the capitalization pattern suggests otherwise. To ameliorate the problems associated with sarcasm when performing sentiment facetization, some implementations may use a trained AI classifier that detects sarcasm in text (e.g., TrueRatr from Cornell University). In some implementations, expressions of contrasting sentiment, such those as obtained from social network metadata or facial expression analysis, may be used to signal sarcasm in text content. A result of sarcasm detection via one or more such techniques may be that the “sarcastic” contrary text indicators are disregarded altogether or counted as an additional instance of the opposite sentiment toward the content element.).
As for claim 22, Howard teaches The system of claim 1. Howard does not specifically teach, wherein to recognize the conflict to be corrected, the processor is further programmed to: store, in a graph database, a plurality of content elements associated with respective positions in a timeline of the content; generate, for a first content element, a first embedding representation at a first time; generate, for a second content element associated with the first content element in the timeline, a second embedding representation at a second time; determine, based on a comparison between the first embedding representation and the second embedding representation, that a semantic inconsistency exists between the first content element and the second content element across different modalities; and automatically propagate the corrective action to modify at least one additional content element associated with the timeline; however in the same field of endeavor Sadr teaches (par. 41 querying database with model-generated dataset include processing model generated data with one more machine learned models such as the datasets can be processed by one or more embedding models (shown in fig. 11b is 4 different ones) and presented as shown in graph for user selection the user and build upon these embeddings to generate new queries based upon previous selections and thus establishes a timeline of user interaction and model generation, figure 7a-d demonstrate user input along a timeline of user interfaces displayed to the user which builds upon the previous user interface presented to the user. The embedding models are used to generate and display one or more features which can be utilized to que4ry for a database for associated embeddings (e.g. one or more embedding neighbors) which may be associated with one or more candidate search results.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Sadr into Howard as modified by Malkiel, this is true because Sadr suggests in paragraph 4 , the utilization of artificial intelligence techniques to generate images and/or other datasets can be non-intuitive, may be open-ended, and may be time consuming. Current image generation systems utilize a prompt input box for receiving freeform text to be processed to generate one or more images. However, as a user utilizes the prompt input box, the user may struggle with which words to utilize and/or may be dissatisfied with the generated image as one or more of the input words may not be utilized in the direction the user desired (e.g., “houndstooth” may be entered by the user in association with the pattern; however, the model may generate an image with a dog's teeth).
(Note :) It is noted that any citation to specific, pages, columns, lines, or figures in the prior art references and any interpretation of the references should not be considered to be limiting in any way. A reference is relevant for all it contains and may be relied upon for all that it would have reasonably suggested to one having ordinary skill in the art. In re Heck, 699 F.2d 1331, 1332-33, 216 USPQ 1038, 1039 (Fed. Cir. 1983) (quoting In re Lemelson, 397 F.2d 1006,1009, 158 USPQ 275, 277 (CCPA 1968)).
Response to Arguments
Applicant's arguments filed 05/26/2026 have been fully considered but they are not persuasive.
A1. Applicant argues Malkiel's backward prompt architecture described in paragraph 95 does not disclose generating a verification prompt based on semantic relationships between multiple portions of generated multimodal content to determine whether those portions are semantically consistent with one another across modalities, as presently recited.
R1. Examiner does not agree, Malkiel teaches digital 450 verified answer, e.g., answer 430 which is determined 1024 to match retrieved context or is other deemed free of hallucination content; digital 452 input data. In paragraph 95 Although a question will typically include text, a question also includes other kinds of data in some situations, such as image data, sound data, video data, or sensor data that the text of a question refers to, or non-text data which otherwise provides meaningful context to the text of a question. Hence, models for embedding non-text data are also used in some cases. This verification includes context which comes from multi-modal context such as image data, sound data, video data and sensor data that the text of the question (prompt) refers to. In an attempt to expedite prosecution Sadr is introduced to also teach this limitation.
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Inquires
Any inquiry concerning this communication should be directed to NICHOLAS AUGUSTINE at telephone number (571)270-1056.
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.
PNG
media_image1.png
213
559
media_image1.png
Greyscale
/NICHOLAS AUGUSTINE/Primary Examiner, Art Unit 2178 August 4, 2026
Prosecution Insights