Prosecution Insights
Last updated: November 05, 2026
Application No. 18/784,082

METHOD FOR BIDIRECTIONAL TRANSLATION BETWEEN SIGN LANGUAGE AND TEXT USING AI, DEEP LEARNING, AND DICTIONARY SEARCH TECHNIQUES

Final Rejection §103§112
Filed
Jul 25, 2024
Examiner
WOZNIAK, JAMES S
Art Unit
2655
Tech Center
2600 — Communications
Assignee
Dr Ahmed Mahgoub
OA Round
2 (Final)
59%
Grant Probability
Moderate
3-4
OA Rounds
1y 4m
Est. Remaining
99%
With Interview

Examiner Intelligence

Grants 59% of resolved cases
59%
Career Allowance Rate
244 granted / 413 resolved
-2.9% vs TC avg
Strong +40% interview lift
Without
With
+40.0%
Interview Lift
resolved cases with interview
Typical timeline
3y 7m
Avg Prosecution
17 currently pending
Career history
442
Total Applications
across all art units

Statute-Specific Performance

§101
19.2%
-20.8% vs TC avg
§103
42.8%
+2.8% vs TC avg
§102
16.3%
-23.7% vs TC avg
§112
16.9%
-23.1% vs TC avg
Black line = Tech Center average estimate • Based on career data from 413 resolved cases

Office Action

§103 §112
DETAILED ACTION Notice of Pre-AIA or AIA Status The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA . Response to Amendment In response to the Non-Final Office Action from 4/15/2026, Applicant has filed an amendment on 7/13/2026. In this reply, Applicant has amended independent claim 1 to properly set forth the method claim as including a series of steps by converting the recited system components into step. In some cases, the functionality of the system component converted to a method step has been altered as in the case of "generating using a fine-tuned response-generation large language model (LLM), responses based on the coherent sentence by querying a document or database using a retrieve-and-generate (RAG) model" and "refining subsequent conversational interactions by feeding back the user input and contextual information directly into the RAG model query parameters." In summary, the amendment to independent claim 1 involves conversion of system functionality into proper method steps while adding RAG model processing. Independent claim 7 was amended to add details similar to claim 1 method steps regarding LLM-based sentence simplification, word embeddings that capture semantic relationships, video construction, and handling non-matching words while adding functionality for nearest-neighbor matching for mapping word embeddings to sign language words. Applicant has also argued that the prior art of record fails to teach the claimed particularly arranged processing pipeline (Remarks, Pages 10-12) as well as amended features such as the added RAG processing (Remarks, Pages 12-14). These arguments have been fully considered, however, are moot with respect to the new grounds of rejection, necessitated by the amended claims and further in view of Marelus, et al. (U.S. PG Publication: 2025/0349294 A1). Applicant argues that the replacement drawing sheets that resolves duplicate reference characters overcomes the objection to the drawings (Remarks, Page 6). While the specification was amended to relabel the query sign language dictionary with reference number 306 to resolve the labeling conflict, the drawings have not been similarly amended. See Fig. 3 in the replacement drawings submitted on 7/13/2026: PNG media_image1.png 64 138 media_image1.png Greyscale Accordingly, as the drawings have not been corrected in a manner similar to the instant disclosure, the objection to the drawings has been maintained. Applicant argues that all claims have been amended to address the minor informalities regarding acronym expansion, add conjunctions, and begin claim limitations with lowercase letters (Remarks, Page 7). In response to the correction of the noted informalities, the claim objections directed towards minor informalities have been withdrawn with the exception of dependent claims 5 and 8 where the conjunction "and" was not added prior to the last limitations in each of the claims. Applicant requests withdrawal of the indefiniteness rejections of claims 1-8 under 35 U.S.C. 112(b) due to rewriting method 1 as a proper method claim by converting structural component functionality into method steps, correcting antecedent basis issues, and the removal of unclear "such as" language (Remarks, Pages 7-8). In response to the recitation of a proper method comprising steps, removal of unclear claim language, and clarification of antecedent basis via the instant amendment, the 35 U.S.C. 112(b) rejections are now moot and have been withdrawn. Applicant should be aware that the amended claims, while resolving past antecedent basis issues, have resulted in a multitude of other indefiniteness issues that have been set forth in the below 35 U.S.C. 112(b) rejections. Applicant argues that the 35 U.S.C. 101 rejection of claim 1 and its dependents for failing to fall within at least one of the four statutory categories due to a lack of any steps or acts so as to properly fall within the process statutory class has been overcome due to the instant amendment to claim 1 now reciting an ordered method instead of a collection of modules (Remarks, Pages 8-9). The examiner agrees that claim 1 as amended now includes at least an act, step, or series thereof and as such relates to a proper process claim falling within one of the four statutory categories of invention. As such, the rejection of claims 1-6 and 8 under 35 U.S.C. 101 has been withdrawn. Applicant argues that Claim 7 as amended is no longer directed towards a judicial exception in the form of a mental process under the broadest reasonable interpretation (BRI). In particular, Applicant contends that point out the recited processing steps carried out via the recited computer program execution and submit that a human cannot practically perform the recited process. Applicant further argues that the claimed invention relates to a practical application in the form of a sign language video output and that the claimed invention includes a particular ordered combination of API-based LLM restructuring, embedding generation, etc. that is not well-known, routine, and conventional per step 2B of the 2019 Patent Subject Matter Eligibility Guidelines (2019 PEG) (Remarks, Pages 9-10). In response, Applicant is reminded that the claim language considered under steps 2A prong 2 and 2B of the 2019 PEG are those that are in addition to the identified abstract idea under the BRI- "An inventive concept "cannot be furnished by the unpatentable law of nature (or natural phenomenon or abstract idea) itself." Genetic Techs. Ltd. v. Merial LLC, 818 F.3d 1369, 1376, 118 USPQ2d 1541, 1546 (Fed. Cir. 2016) and "if there are no additional claim elements besides the judicial exception, or if the additional claim elements merely recite another judicial exception, that is insufficient to integrate the judicial exception into a practical application," see, e.g., RecogniCorp, LLC v. Nintendo Co., 855 F.3d 1322, 1327, 122 USPQ2d 1377 (Fed. Cir. 2017). With this analysis in mind, it is noted that some of the limitations upon which Applicant relies to evidence eligibility under 35 U.S.C. 101 could be practically performed by a human (e.g., - mentally considering semantic distance representation to render a mental judgement as to the closest sign language word and mentally identifying a sequence of sign language words from that mapping), however, the claim does feature additional elements that would be set aside for consideration under step 2A prong 2 and 2B including the use of a specific simplification LLM in a process for sign language video generation using embeddings. Thus, amended claim 7 is found to be eligible to be directed to a technical practical application in the form of video rendering of a sign language sequence through the use of a simplification LLM accessible via an API and word embeddings under step 2A prong 1. Accordingly, the 35 U.S.C. 101 rejection of claim 7 has been withdrawn. Response to Arguments Applicant has traversed the prior art rejection of claims 1-5 and 7-8 under 35 U.S.C. 102(a)(2) as being anticipated by Meged, et al. (U.S. PG Publication: 2025/0308409 A1) for several reasons. First, applicant contends that the claims require a particularly ordered pipeline (note: that includes newly introduced limitations related to use of a a RAG model for response generation) and then generally alleges that "Meged does not disclose this pipeline arranged as claimed" (Remarks, Pages 10-12). In response, see MPEP 2111.01 (II) and Altiris Inc. v. Symantec Corp., 318 F.3d 1363, 1371, 65 USPQ2d 1865, 1869-70 (Fed. Cir. 2003)- "Although the specification discussed only a single embodiment, the court held that it was improper to read a specific order of steps into method claims where, as a matter of logic or grammar, the language of the method claims did not impose a specific order on the performance of the method steps, and the specification did not directly or implicitly require a particular order." Thus, while certain steps do contain specific ordering via antecedent basis, the entire sequence need not be carried out in the order claimed as claim 1 recites a method "comprising" the indicated steps. Thus, Applicant's arguments are directed towards a scope of the claim that is not recited. Additionally, Applicant arguments do not point out a particular step that is related to a prior step via antecedent basis that is lacking in the teachings of Meged and so amounts to a general allegation of patentability lacking such a specific identification of claim language and rationale. Lastly, it should be pointed out that recited method has been amended so as to necessitate new grounds of rejection further in view of Marelus, et al. (U.S. PG Publication: 2025/0349294 A1) thus rendering these arguments moot. Next, Applicant argues that Meged "does not disclose monitoring dictionary updates and substituting previously spelled output with newly available signs" as a "lifecycle limitation." In response, it is noted that Applicant's arguments rely upon features not claimed. Even as amended, claim 1 does not recite any dictionary update monitoring. Instead, claim 1 merely recites "spelling out non-matching words that do not have corresponding sign language words in the provided sign language dictionary" that is addressed by Meged's finger gloss spellings found at Paragraphs 0147 and 0170 and "replacing spelled-out output for the non-matching word with a corresponding sign language gesture when the non-matching word is added to the provided sign language dictionary" that is addressed by Meged's "replacement" operation and new gloss creation detailed in Paragraphs 0141 and 0170. Thus, since Applicant's argument does not address the language of claim 1 that is addressed by Meged under the broadest reasonable interpretation (BRI), this argument is not found to be persuasive. Applicant further arguments directed towards RAG (Remarks, Page 13-14) are moot with respect to the above-noted new grounds of rejection necessitated by the amended claims. Applicant next argues that the word-embedding nearest neighbor mapping recited in the processing sequence is not provided in the teachings of Meged (Remarks, Page 13). In response, it is pointed out that the features upon which the Applicant relies are not present in claim 1 and so this argument is considered moot. Instead, such limitations are found in independent claim 7 as amended. Applicant arguments, if applied to claim 7, would be considered moot, necessitated by the amended claims and further in view of Yang, et al. (CN 116561245). Applicant argues that combining the Applicant's RAG with the teachings of Meged amounts to impermissible hindsight (Remarks, 14-15). This argument is moot and confusing since no such combination was made or suggested in the Non-final Office Action. Lastly, Applicant arguments directed towards claim 4 (Remarks, Page 14) are moot with respect to the new grounds of rejection further in view of the new grounds of rejection necessitated by the amendment of this claim while Applicant arguments directed towards claim 6 (Remarks, Pages 14-16) are moot since this rejection has been withdrawn. Drawings The drawings are objected to as failing to comply with 37 CFR 1.84(p)(4) because reference character “201” has been used to designate both the vector embedding model and query sign language dictionary. Corrected drawing sheets in compliance with 37 CFR 1.121(d) are required in reply to the Office action to avoid abandonment of the application. Any amended replacement drawing sheet should include all of the figures appearing on the immediate prior version of the sheet, even if only one figure is being amended. Each drawing sheet submitted after the filing date of an application must be labeled in the top margin as either “Replacement Sheet” or “New Sheet” pursuant to 37 CFR 1.121(d). If the changes are not accepted by the examiner, the applicant will be notified and informed of any required corrective action in the next Office action. The objection to the drawings will not be held in abeyance. Claim Objections Claims 5-6 and 8 are objected to because of the following informalities: Since claims 5 and 8 feature a collection of clauses in a combination, the word --and-- should be inserted between the table mapping words and data related to interactions in claim 5 and between the training and utilizing steps in claim 8. The dependent claims inherit the objected to limitations of their respective parent claims, and thus, have also been objected to due to minor informalities by virtue of their dependency. Appropriate correction is required. Claim Rejections - 35 USC § 112 The following is a quotation of the first paragraph of 35 U.S.C. 112(a): (a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention. The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112: The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention. Claim 6 is rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention. Claim 6 has been amended to recite in part: "updating the trained model used for mapping the word embeddings based on the archived new words and the user feedback to refine subsequent translations by updating vector representations stored in an embedded words database used for nearest-neighbor semantic matching, without retraining global parameters of a vector embedding model used to generate the word embeddings." While the specification does teach updating a sign language dictionary "for each new word" (Paragraph 0055) and storing an embedded word database that includes current data tables for storing ongoing translation data and historical data tables for archiving completed translations and user feedback" to allow for "continuous refinement" (0029; see also original claim 6). The originally filed specification makes no mention of any details of the underlying updating process with the claimed details of specifically training the model based upon such data where updating occurs "without retraining global parameters of a vector embedding model used to generate the word embeddings." Thus, Applicant fails to show possession of the subject matter of amended claim 6 based upon the originally filed disclosure of the invention. The following is a quotation of 35 U.S.C. 112(b): (b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention. The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph: The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention. Claims 1-8 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention. In Claim 1, Line 46, "the complete sentence in sign language" lacks antecedent basis and it is unclear what prior limitation is being referenced. For claim interpretation in the interest of compact prosecution, "the complete sentence in sign language" will be construed as --the video from a sequence of the images of sign language gestures--. In Claim 1, Line 48, "the user input" lacks antecedent basis and it is unclear what prior limitation is being referenced. For claim interpretation in the interest of compact prosecution, "the user input" will be construed as --a user input--. In Claim 1, Line 49, "the RAG model query parameters"" lacks antecedent basis and it is unclear what prior limitation is being referenced. For claim interpretation in the interest of compact prosecution, "the RAG model query parameters" will be construed as --RAG model query parameters--. In Claim 1, Line 54, "the non-matching word" lacks antecedent basis and it is unclear whether this limitation finds antecedent basis in the preceding "non-matching words" or represents a new term. For claim construction in the interest of compact prosecution, "non-matching word" will be construed as --the non-matching words--. Note that another instance of this limitation is also present in line 55 and also requires correction. In Claim 5, "the user feedback" lacks antecedent basis and it is unclear what prior limitation is being referenced. For claim interpretation in the interest of compact prosecution, "the user feedback" will be construed as --user feedback--. Dependent claims 2-6 and 8 inherit and fail to resolve the indefinite claim language of parent claim 1 by resolving antecedent basis issues, and thus, have also been rejected under 35 U.S.C. 112(b) for being indefinite by virtue of their dependency. The term “more accessible” in Claim 7 is a relative term which renders the claim indefinite. The term “more accessible” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. The use of the recited term renders the corresponding “language” that is a restructuring result unclear and indefinite. Claims 1-5 and 7-8 Claim Rejections - 35 USC § 103 The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action: A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made. Claims 1-2, 5, and 8 are rejected under 35 U.S.C. 103 as being unpatentable over Meged, et al. (U.S. PG Publication: 2025/0308409 A1) in view of Marelus, et al. (U.S. PG Publication: 2025/0349294 A1). With respect to Claim 1, Meged discloses: A method for automatic sign language video conversation by translating between sign language and text using word embeddings, comprising: capturing video input from a user performing sign language gestures (recording a signer making gestures/movements in motion capture (MoCap), Paragraphs 0190 and 0193); sampling the captured video to identify individual sign language words (recorded gestures are "digitally translated" leveraging computer vision for each gloss for the sign language interpretation, Paragraphs 0190 and 0193); converting, using a trained computer vision model, the sampled video containing the sign language gestures into corresponding sign language words (translation of input signs for an LLM of a personal assistant, Paragraphs 0024-0025; translation involves generating a gloss corresponding to a word or words in an input to convert or translate a video/sign language input for processing by an LLM using trained “computer vision” modeling techniques, Paragraphs 0019, 0111, 0118-0119, 0169, and 0188-0193); converting the identified sign language words into a coherent sentence (pre-processing a sequence of input words into a sentence such as a question or a request for information for understanding by an LLM, Paragraphs 0019, 0024-0025, 0111, 0117, and 0169); generating, using a fine-tuned response-generation large language model (LLM), responses based on the coherent sentence (examples of fine-tuned LLMs to generate responses using the formatted input such as an answer to a question or requested information, Paragraphs 0024-0025); simplifying, using a pre-trained simplification large language model (LLM) accessed via an application programming interface (API) (use of an API in sign language generation processing, Paragraphs 0021-0022, 0024, and 0097), the generated responses into a simplified sentence focusing on key verbs and nouns, removing complex or unnecessary phrases, and ensuring use of existing words in a provided sign language dictionary to simplify the reconstructed sentence, focusing on key verbs and nouns, and removing complex or unnecessary phrases, and ensuring the use of existing words in the provided dictionary ("input is pre-processed and restructured" where pre-processing involves conversion into "language that's more accessible" and "easier...to understand" and wherein pre-processing can involve the use of "models like BERT or GPT" that are specific examples of pre-trained large language models, Paragraphs 0114-0117; see that generation involves a focus on verbs/actions and important nouns in the sentence as well as the removal of complex terms/words (e.g., "condition") and ensuring the use of dictionary terms such as via the use of exiting synonyms, 0114-0116, 0123, 0125, and 0141-0146); tokenizing the simplified sentence and representing the simplified sentence using word embeddings that capture semantic relationships between words (identification of "each word in the input" constituting a token and the generation of semantic/word sense/contextual embeddings for disambiguation, Paragraphs 0011, 0117, 0127-0128, 0134, and 0169); mapping the word embeddings to a predefined set of sign language words using a trained model, accommodating variations in sentence structure and context (linguistic elements in the form of words utilizing the contextual embeddings are translated into a gloss ("written text to capture the essence of sign language" or "sign language words" as claimed) where "each gloss corresponds to a specific word or to a sequence of words from the...pre-processed input, Paragraphs 0117-0120, 0122, and 0126-0128; note the varying accommodating approaches for various sentence structure and context via homographs, context, intensity based upon context, etc., in Paragraphs 0127-0128, 0146-0148, and 0158); converting the mapped embeddings into a sequence of sign language words (generation of "Visual generation guidance can include one or more directives pertaining to the generation of visual signs based on notation extending to a gloss sequence," Paragraphs 0172-0174 and 0196); mapping the sequence of sign language words to images of sign language gestures using the provided sign language dictionary (use of existing visual data dictionaries in performing mapping/selection of an optimal visual image sequence, Paragraphs 0188 and 0193-0196); constructing a video from a sequence of the images of sign language gestures to be displayed to the user, enabling the user to see the complete sentence ("provide a visual character, such as an avatar 120 to perform the representation by displaying a video of the avatar 120 presenting the visual signs based on the guidance," Paragraphs 0098, 0101, and 0204 (discussing rendering or construction of such a video)); refining subsequent conversational interactions by feeding back (user preferences on avatar presentation and user input contextual information for ensuring that the sign language is translated correctly, Paragraphs 0026, 0113, 0120, and 0201); and spelling out non-matching words that do not have corresponding sign language words in the provided sign language dictionary and replacing spelled-out output for the non-matching word with a corresponding sign language gesture when the non-matching word is added to the provided sign language dictionary ("visual spelling" used in situations "where words without representation as a gloss in sign language are identified," Paragraph 0147; techniques such as finger spelling is used in situations "where words without representation as a gloss in sign language are identified" where words may be identified in the future when added via visual data capture and "creating new glosses," Paragraphs 0141, 0147, 0170, and 0188-0193). While Meged discloses that the digital assistant utilizing an LLM model can allow a user to receive a response comprising "information based on data available to the LLM model" (Paragraph 0024). Meged does not specifically indicate that the response is generated by querying a document or database using a retrieve-and-generate (RAG) model. Marelus, however, discloses a user interface that enables communication via sign language (Paragraphs 0277, 0294, 0314, 0385, 0398 ("camera 301 may further be used, for example, to allow the system to have conversations with people who have speech impediments or who have to (partially) converse with gestures"), and 0431) that utilizes a RAG model to query a document database to generate a user response where future conversational interactions are refined and rendered based upon user conversational inputs from a conversational history and contextual information related to a user when querying the database in the RAG module (Paragraphs 0210, 0219, and 0251-0253). Meged and Marelus are analogous art because they are from a similar field of endeavor in LLM-based digital assistants. Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date to modify the process of Meged to include the RAG database querying and context in response generation taught by Marelus to provide a predictable result of holding more enriched, human-like conversation with the digital agent (Marelus, Paragraph 0251). With respect to Claim 2, Meged further discloses: The method of claim 1, wherein the word embeddings are generated using a pre-trained natural language processing (NLP) embedding model selected from the group consisting of Word2Vec, GloVe, and BERT, the NLP embedding model being fine-tuned on a corpus of text to capture language-specific nuances (see known LLM BERT that is fine-tuned to process text and handle language nuances such as normalization, disambiguation, and context wherein a natural language text processing model needs to be previously fine-tuned on text so as to meaningfully process text, Paragraphs 0024, 0117, and 0128). With respect to Claim 5, Meged further discloses: The method of claim 1, further comprising storing, in a data storage platform (process implementation using a processor and associated memory, Paragraphs 0091-0092 and 0103-0104, wherein throughout the processing steps, the resulting data would require storage at least temporarily in order to proceed with the further processing steps/analysis): intermediate data representing the simplified sentence ("input is pre-processed and restructured" where pre-processing involves conversion into "language that's more accessible" and "easier...to understand", Paragraphs 0114-0117); data representing the tokenized sentence in a form of the word embeddings (identification of "each word in the input" constituting a token and the generation of semantic/word sense/contextual embeddings for disambiguation, Paragraphs 0011, 0117, 0127-0128, 0134, and 0169); data reflecting the mapping of word embeddings to sign language words (linguistic elements in the form of words utilizing the contextual embeddings are translated into a gloss ("written text to capture the essence of sign language" or "sign language words" as claimed) where "each gloss corresponds to a specific word or to a sequence of words from the...pre-processed input, Paragraphs 0117-0120, 0122, and 0126-0128); data representing a final sequence of sign language words (visual data corresponding to the gloss sequence and how to present it, Paragraphs 0197; Fig. 3, Element 370); a table mapping individual letters to their corresponding sign language images (mapping of letters to glosses and images in finger spelling, Paragraphs 0147 and 0170-0172); a table mapping words to their corresponding sign language images (visual data pertaining to a plurality of glosses such as the word he or left mapped to corresponding sign images, Paragraphs 0194-0195); and data related to interactions with human agents and the user feedback (stored preferences that the user can select "relate" to their preferences for interactions with humanoid agents and feedback to the type of avatar that is preferred, Paragraph 0201; Fig. 1C). With respect to Claim 8, Meged further discloses: The method of claim 1, further comprising: maintaining a collection of labeled sign language images, each associated with corresponding sign language words ("Visual data 232 may store visual data pertaining to a plurality of glosses" (i.e., "sign language words"), Paragraph 0194); training a computer vision model using the labeled sign language images to recognize and predict sign language words from new images (computer vision used for "sign language interpretation," Paragraphs 0188, 0192-0194, and 0202 wherein in order to be able to perform such interpretation, the model must first be trained in some manner, otherwise a computer would not be able to make an association between images and glosses/words); utilizing the trained computer vision model to output a necessary sign language image corresponding to a given word, enabling construction of an answer video to be displayed to the user (animation of the avatar using the computer vision model, Paragraph 0202; see also Paragraphs 0188 and 0192-0194; and digital assistant implementation providing answers or requested information, Paragraphs 0024-0025). Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Meged, et al. in view of Marelus, et al. and further in view of Yang, et al. (CN 116561245, accompanied by an English machine translation thereof). With respect to Claim 3, Meged in view of Marelus discloses the sign language video generation process utilizing a mapping of word embeddings to sign language words as applied to Claim 1. Meged in view of Marelus do not teach that this process comprises performing semantic search using nearest-neighbor matching in an embedding space to select sign language words that consider semantic similarity and grammatical structure specific to sign language. Yang, however, discloses vector proximity in a vector/embedding space wherein the "closer the corresponding words are in space, the closer the corresponding words are in semantics and meaning" (i.e., semantic and grammatical sense) wherein distance/proximity may be calculated as a Euclidean distance and the vectors/embeddings are those of a word vector and a "sign-language word" (Pages 6). Meged, Marelus, and Yang are analogous art because they are from a similar field of endeavor in sign language communication utilizing machine learning models. Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date to substitute in the method of embedding to sign language word mapping taught by Yang into the mapping process taught by Meged in view of Marelus to provide a predictable result of providing a useful metric to identify semantically similar sign language words. Claim 4 is rejected under 35 U.S.C. 103 as being unpatentable over Meged, et al. in view of Marelus, et al. and further in view of Cordell, et al. (U.S. PG Publication: 2020/0366790 A1). With respect to Claim 4, Meged in view of Marelus discloses the sign language video generation process utilizing a mapping of word embeddings to sign language words as applied to Claim 1. Meged in view of Marelus do not teach that the transmission of a user communication (i.e., coherent sentence generated by Meged) to a human agent via a human agent interaction application programming interface (API) when the user requests human agent assistance, and incorporating a response from the human agent into the generated responses prior to simplifying the generated responses. Cordell, however, discloses an API that is used to instantiate a communication channel between either an automated agent or a live human agent (Paragraph 0028) wherein a user may request a human agent to receive the interpreted request and a combination of these communications may be utilized to respond to the user (i.e., human and machine responses are incorporated) (Paragraph 0035). Meged, Marelus, and Cordell are analogous art because they are from a similar field of endeavor in sign language communication. Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate the human agent interaction in the sign language response process taught by Meged in view of Marelus to provide a predictable result in the form of enabling a human to better respond to a sign language request when a digital/machine assistant is able to properly/fully respond. Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over Meged, et al. in view of Marelus, et al. and further in view of Mishra (U.S. PG Publication: 2024/0304031 A1). With respect to Claim 6, Meged in view of Marelus discloses the virtual agent using bi-directional sign language communication as applied to Claim 5. Meged also disclsoes for storing ongoing translation data as a part of a processor-implemented process per the claim 5 rejection wherein ongoing translation data and new sign language words are stored as a part of a digital assistant/meeting session (Paragraph 0023-0025, 0102, 0141, and 0170). Meged does not specifically teach archiving completed translations, new words, human agent interaction, and user feedback for further refinement. Mishra, however, disclsoes archiving completed translations, new words, human agent interaction, and user feedback for further refinement in an implied table format where such data is inked to communication sessions (historical comm session database 248 storing completed translations in a session, user specific gestures having different word-based meanings, human agent interaction and user/agent feedback for model training/refinement, Paragraphs 0053-0054, 0059-0060, 0065, 0069 (words related to gestures), 0086, 0099, 0101, and 0104-0105). Meged and Mishra are analogous art because they are from a similar field of endeavor in sign language communication systems. Thus, it would have been obvious to one of ordinary skill in the art, before the effective filing date, to utilize the archiving of historical data taught by Mishra in the sign language translation models of Meged to provide a predictable result of a model that is more accurate in detecting, identifying, and translating gestures of a particular user (Mishra, Paragraph 0054). Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Meged, et al. in view of Yang, et al. With respect to Claim 7, Meged discloses: A non-transitory computer-readable storage medium storing executable instructions that, when executed by a processor, cause a system to perform operations for translating complete sentences into sign language using word embeddings, comprising: (non-transitory computer-readable memory storing a program of instructions executable by a computer processor, Paragraph 0209), comprising: Receiving an input sentence ("obtain input content for translation to sign language...can be text input such as a sentence," Paragraph 0111); simplifying the input sentence using a pre-trained simplification large language model (LLM) accessed via an application programming interface (API) (use of an API in sign language generation processing, Paragraphs 0021-0022, 0024, and 0097), the pre-trained simplification LLM restructuring the input sentence into language that is more accessible for sign language translation ("input is pre-processed and restructured" where pre-processing involves conversion into "language that's more accessible" and "easier...to understand" and wherein pre-processing can involve the use of "models like BERT or GPT" that are specific examples of pre-trained large language models, Paragraphs 0114-0117; see that generation involves a focus on verbs/actions and important nouns in the sentence as well as the removal of complex terms/words (e.g., "condition") and ensuring the use of dictionary terms such as via the use of exiting synonyms, 0114-0116, 0123, 0125, and 0141-0146); tokenizing the simplified sentence and representing the simplified sentence with word embeddings that capture semantic relationships between words (identification of "each word in the input" constituting a token and the generation of semantic/word sense/contextual embeddings for disambiguation, Paragraphs 0011, 0117, 0127-0128, 0134, and 0169); mapping the word embeddings to sign language words in a predefined sign language dictionary (linguistic elements in the form of words utilizing the contextual embeddings are translated into a gloss ("written text to capture the essence of sign language" or "sign language words" as claimed) where "each gloss corresponds to a specific word or to a sequence of words from the...pre-processed input, Paragraphs 0117-0120, 0122, and 0126-0128); generating a sequence of sign language words from the mapped word embeddings (generation of "Visual generation guidance can include one or more directives pertaining to the generation of visual signs based on notation extending to a gloss sequence," Paragraphs 0172-0174 and 0196); mapping the sequence of sign language words to a sequence of images of sign language gestures using the predefined sign language dictionary ("provide a visual character, such as an avatar 120 to perform the representation by displaying a video of the avatar 120 presenting the visual signs based on the guidance" where the plural “visual signs” used to generate the video avatar correspond to the claimed “sequence of images,” Paragraphs 0098, 0101, and 0204 (discussing rendering or construction of such a video)); constructing a video from the sequence of images of sign language gestures to be displayed to a user ("provide a visual character, such as an avatar 120 to perform the representation by displaying a video of the avatar 120 presenting the visual signs based on the guidance," Paragraphs 0098, 0101, and 0204 (discussing rendering or construction of such a video)); and handling non-matching words that lack corresponding sign language words in the predefined sign language dictionary by spelling out the non-matching words or substituting the non-matching words with corresponding sign language gestures when available in the predefined sign language dictionary (handling situations where a word/gloss does not have visual sign information by substituting sign language "synonyms" that "that can convey the same or very similar meanings but have established signs in ASL," Paragraphs 0125; see also "visual spelling" used in situations "where words without representation as a gloss in sign language are identified," Paragraph 0147; although the limitation is recited in the alternative where the prior art needs to only teach at least one of the alternatives, the prior art discloses both alternatives). Meged does not teach that the process of mapping word embeddings to sign language words comprises performing semantic search using nearest-neighbor matching in an embedding space to select sign language words that consider semantic similarity and grammatical structure specific to sign language. Yang, however, discloses vector proximity in a vector/embedding space wherein the "closer the corresponding words are in space, the closer the corresponding words are in semantics and meaning" (i.e., semantic and grammatical sense) wherein distance/proximity may be calculated as a Euclidean distance and the vectors/embeddings are those of a word vector and a "sign-language word" (Pages 6). Meged and Yang are analogous art because they are from a similar field of endeavor in sign language communication utilizing machine learning models. Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date to substitute in the method of embedding to sign language word mapping taught by Yang into the mapping process taught by Meged to provide a predictable result of providing a useful metric to identify semantically similar sign language words. Potentially Allowable Subject Matter Claim 6 would be potentially allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims and amended to overcome the rejections under 35 U.S.C. 112 (a and b) and claim objections directed towards minor informalities. The following is a statement of reasons for the indication of potentially allowable subject matter: With respect to Claim 6, Meged discloses a sign language communication system with a digital assistant utilizing an LLM and embeddings while Marelus teaches the ability to utilize RAG in response generation as applied to Claims 1 and 5 while additional prior art in the form of Mishra (U.S. PG Publication: 2024/0304031 A1) teaches the types of data being tracked in claim 6 via archiving completed translations, new words, human agent interaction, and user feedback for further refinement in an implied table format where such data is inked to communication sessions (historical comm session database 248 storing completed translations in a session, user specific gestures having different word-based meanings, human agent interaction and user/agent feedback for model training/refinement, Paragraphs 0053-0054, 0059-0060, 0065, 0069 (words related to gestures), 0086, 0099, 0101, and 0104-0105), the model refinement utilizing such data in Mishra does not involve updating vector representations stored in an embedded words database used for nearest-neighbor semantic matching, without retraining global parameters of a vector embedding model used to generate the word embeddings. Thus, the prior art of record fails to teach or fairly suggest the subject matter set forth in claim 6. 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 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. The prior art made of record and not relied upon is considered pertinent to applicant's disclosure: Xu (U.S. PG Publication: 2025/0363145 A1)- teaches a system for sign language communication with a user (Paragraph 0060) using Retrieval Augmented Generation (RAG) (Paragraph 0047). Any inquiry concerning this communication or earlier communications from the examiner should be directed to JAMES S WOZNIAK whose telephone number is (571)272-7632. The examiner can normally be reached 7-3, off alternate Fridays. Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant may 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, Andrew Flanders can be reached at (571)272-7516. 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. JAMES S. WOZNIAK Primary Examiner Art Unit 2655 /JAMES S WOZNIAK/ Primary Examiner, Art Unit 2655
Read full office action

Prosecution Timeline

Jul 25, 2024
Application Filed
Apr 15, 2026
Non-Final Rejection mailed — §103, §112
Jul 13, 2026
Response Filed
Sep 18, 2026
Final Rejection mailed — §103, §112 (current)

Precedent Cases

Applications granted by this same examiner with similar technology

Patent 12784742
IMPROVING DETECTION OF VOICE-BASED KEYWORDS USING FALSELY REJECTED DATA
2y 9m to grant Granted Nov 03, 2026
Patent 12770640
DATA PROCESSING METHOD AND APPARATUS, ELECTRONIC DEVICE, COMPUTER-READABLE STORAGE MEDIUM, AND COMPUTER PROGRAM PRODUCT
2y 7m to grant Granted Oct 20, 2026
Patent 12748928
Service Platform Integration with Generative Natural Language Models
2y 4m to grant Granted Sep 29, 2026
Patent 12744028
GENERATING CONVERSATIONAL OUTPUT USING A LARGE LANGUAGE MODEL
2y 9m to grant Granted Sep 22, 2026
Patent 12738262
ENABLING TRAINING OF A MACHINE-LEARNING MODEL FOR TRIGGER-WORD DETECTION
3y 4m to grant Granted Sep 15, 2026
Study what changed to get past this examiner. Based on 5 most recent grants.

Strategy Recommendation AI-generated — please review before filing

Get a prosecution strategy drawn from examiner precedents, rejection analysis, and claim mapping.
Typically takes 5-10 seconds — AI-generated, attorney review required before filing

Prosecution Projections

3-4
Expected OA Rounds
59%
Grant Probability
99%
With Interview (+40.0%)
3y 7m (~1y 4m remaining)
Median Time to Grant
Moderate
PTA Risk
Based on 413 resolved cases by this examiner. Grant probability derived from career allowance rate.

Sign in with your work email

Enter your email to receive a magic link. No password needed.

Personal email addresses (Gmail, Yahoo, etc.) are not accepted.

Free tier: 3 strategy analyses per month