DETAILED ACTION
This action is in response to the application filed 2 April 2024, claiming benefit back to 20 February 2024.
Claims 1 – 20 are pending and have been examined.
This action is Non-Final.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Information Disclosure Statement
The information disclosure statements (IDSs) have been considered by the examiner.
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 non-statutory subject matter. The claimed invention, when the claims are taken as a whole, is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
Step 2A – 1: The claims recite a Judicial Exception. Exemplary independent claim 1 recites the limitations of:
[a] receiving a query for completing a task using the web in natural language;
[b] creating a first plan having a first sequence of website action steps from the query using a first large language model (LLM), each website action step specifying a website and including a request for performing an action on the website, comprising:
[c] executing a current website action step specifying a current website and including a current request for performing a current action on the current website, comprising:
[d] creating a current plan having a current sequence of function action steps from the current request using a current LLM based on an application programming interface (API) of the current website, each function action step specifying a function in the current website's API and including one or more values for one or more parameters of the function, comprising:
[e] executing a current function action step of the current sequence of function action steps; and
[f] generating a next function action step of the current sequence of function action steps based on a result of executing the current function action step; and
[g] generating a next website action step of the first sequence of website action steps based on a result of executing the current sequence of function action steps as a result of executing the current website action step;
[h] transmitting a result of executing the first sequence of website action steps in response to the query.
These limitations (bolded and italicized), as drafted, are a process that, under its broadest reasonable interpretation, covers encompasses mental processes practically performed in the human mind. See MPEP 2106.04(a)(2)1. The claim limitations [b] and [d], [f], and [g] recite creating a plan or a step, and as such encompass mental processes practically performed in the human mind using, evaluation, judgment, and opinion; limitations [c] and [e] are executing a function, which again encompass mental processes practically performed in the human mind using, evaluation, judgment, and opinion, e.g., performing an action based on a workflow plan or step of a workflow.
Step 2A – 2: This judicial exception is not integrated into a practical application, and the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception.
The limitations [b] and [d] recite using an LLM, however this is recited at a high level of generality, and provide nothing more than mere instructions to implement an abstract idea on a generic computer. See MPEP 2106.05(f)2. The limitations [a] and [h] recite the receiving or outputting / transmitting of data, which amount to mere data gathering and output recited at a high level of generality, and thus are insignificant extra-solution activity. See MPEP 2106.05(g). In addition, all uses of the recited judicial exceptions require such data gathering and output, and, as such, these limitations do not impose any meaningful limits on the claim. These limitations amount to necessary data gathering and outputting. See MPEP 2106.05.
Further, the claims do not provide for or recite any improvements to the functioning of a computer, or to any other technology or technical field; applying or using a judicial exception to effect a particular treatment or prophylaxis for a disease or medical condition; applying the judicial exception with, or by use of, a particular machine; effecting a transformation or reduction of a particular article to a different state or thing; or applying or using the judicial exception in some other meaningful way beyond generally linking the use of the judicial exception to a particular technological environment, such that the claim as a whole is more than a drafting effort designed to monopolize the exception.
The claim is directed to the abstract idea.
The dependent claims have the same deficiencies as their parent claims as being directed towards an abstract idea, as the dependent claims merely narrow the scope of their parent claims, and it has been held that “[i]n defining the excluded categories, the Court has ruled that the exclusion applies if a claim involves a natural law or phenomenon or abstract idea, even if the particular natural law or phenomenon or abstract idea at issue is narrow.” (buySAFE, Inc. v. Google, Inc., 765 F.3d 1350. )
Turning to the dependent claims, none of the claimed features of the dependent claims further limit the claimed invention in such a way to direct the claimed invention to statutory subject matter (e.g. change the scope of the claimed invention as to no longer be directed towards an abstract idea, or include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements or combination of elements in the claims other than the abstract idea per se), nor do they add limitations that, when taken as a combination, result in the claim as a whole amounting to significantly more than the judicial exception.
In respect to exemplary dependent claims 2 – 14, the claims either further describe the details of the plan, or merely describe further instructions to instructions to implement an abstract idea on a generic computer, in regards to the further use of the LLM or the use of APIs of websites.
Step 2B: The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception because, explained with respect to Step 2A, Prong Two, the additional elements or combination of elements in the claims other than the abstract idea per se amount to no more than mere instructions to implement the idea on a computer, or the recitation of generic computer structure that serves to perform generic computer functions previously known to the industry3 [e.g. performing repetitive calculations; receiving, processing, and storing data; electronically scanning or extracting data from a physical document; electronic recordkeeping; automating mental tasks; receiving or transmitting data over a network, e.g., using the Internet to gather data] .
Applicant’s specification, at, e.g., paragraphs [0020]-[0027], [0032], [0068]-[0088], provides evidence of generic computer hardware performing generic, well-known, computer functions.
Viewed as a whole, these additional claim elements, both individually and in combination, do not provide meaningful limitations to transform the above identified abstract idea into a patent eligible application of the abstract idea such that the claims amount to significantly more (e.g. improvements to another technology or technical fields, improvements to the functioning of the computer itself, or meaningful limitations beyond generally linking the use of an abstract idea to a particular technological environment) than the abstract idea itself. Thus, taken alone, the additional elements do not amount to significantly more than the above-identified judicial exception (the abstract idea). Looking at the limitations as an ordered combination adds nothing that is not already present when looking at the elements taken individually. There is no indication that the combination of elements improves the functioning of a computer or improves any other technology. Their collective functions merely provide conventional computer implementation4.
Therefore, the claims are rejected under 35 U.S.C. 101 as being directed to non-statutory subject matter. See Alice Corporation Pty. Ltd. v. CLS Bank International, 573 U.S. No. 13–298.
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 of this title, 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, 4, 6, 8, 9, 10, 12, 15, 16, 17, 18, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Sodhi et al. (U.S. 2024/0386216, hereinafter Sodhi) in view of Jouhier (U.S. 2020/0409776).
In respect to claim 1, Sodhi discloses a non-transitory, computer-readable storage medium storing one or more sequences of instructions ([0135]) which when executed cause one or more processors to perform:
receiving a query for completing a task using the web in natural language ([0075] At step 520, instruction text is received for accomplishing a task. Any appropriate instruction text may be received. In some implementations, the instruction text may be from a customer support session between a customer and an agent. In some implementations, the instruction text may be received from a user who would like to automate completion of a web browsing task) ;
creating a first plan having a first sequence of website action steps from the query using a first large language model (LLM), each website action step specifying a website and including a request for performing an action on the website ([0069] Task specification 480 instructs the language model of the desired task to be accomplished. This section of the template includes variables or placeholders to be filled according to a current task to be accomplished. For example, {context} may be replaced by instruction text of the desired task, {url} may be replaced by the URL of a current web page, {browser_content} may be replaced by a representation of a current web page, and {previous_actions} may be replaced be zero or more actions already performed to complete the desired task; see further [0076] At step 525, a prompt template is selected using the instruction text. The prompt template may be selected using any appropriate techniques. In some implementations, the instruction text may be processed with a classifier to select a natural language intent from a set of possible natural language intents. Prompt templates may be associated with a natural language intent, and a prompt template may be selected by selecting a prompt template whose intent matches the intent of the instruction text; see further [0113] At step 1010, instruction text is received for accomplishing a task. Step 1010 may be performed as described for step 520. At step 1015, a task prompt template is selected using the instruction text), comprising:
executing a current website action step specifying a current website and including a current request for performing a current action on the current website ([0077] At step 530, an initial web page is be requested. The initial web page may be requested using any appropriate techniques. For example, the initial web page may be determined from the prompt template or may be provided by a user. [0078] In some implementations, the initial web page may be requested using a web browser plugin or extension. For example, a web browser (e.g., Chrome, Firefox, or Safari) may allow users to install plugins or extensions that allow software (e.g., JavaScript) to control the web browser, perform actions on a web page, or obtain data relating to web pages shown in the browser), comprising:
creating a current plan having a current sequence of function action steps from the current request using a current LLM based on an application programming interface (API) of the current website, each function action step specifying a function in the current website's API [and including one or more values for one or more parameters of the function], comprising:
executing a current function action step of the current sequence of function action steps ([0080] At step 535, a prompt is created or generated using the selected prompt template, the instruction text, a representation of a web page (e.g., a representation of the initial web page for the first iteration of step 535), and information about previous actions performed during the automation process (which may be empty for a first iteration of step 535). The prompt may be created using any appropriate techniques, such as any of the techniques described herein); and
generating a next function action step of the current sequence of function action steps based on a result of executing the current function action step ([0081] At step 540, a query is submitted to a language model using the prompt to obtain a response from the language model. The query may be submitted using any appropriate techniques, such as any of the techniques described herein. In some implementations, the prompt may be submitted to third-party language model using an API call. The response of the language model may indicate an action to be performed, such as an operation to be performed on a web page (e.g., clicking an element with a mouse or entering text)); and
While Sodhi discloses an API call for a website, it may not explicitly disclose the API including one or more values for one or more parameters of the function.
Analogous art Jouhier discloses the API including one or more values for one or more parameters of the function ([0056] When a web service operation provided by a web application is to be accessed by part of the client application code (this part of the client application code referred to as a "function caller"), an operation name parameter is passed to the 'operation' method 203. The 'operation' method 203 returns an operation function object to the function caller that can then be used to call that particular web service operation; see further [0065] In certain examples, the conversion between JavaScript input/output objects and the web API input/ output payloads are provided with additional context to deal with protocol options. [0066] For example, XML allows simple values to be passed either as elements or as attributes. The generic library can use a simple convention to distinguish between the two. For example, 'firstName: "Bob"' to format as an element and '$firstName: "Bob"' to format as an attribute).
It would have been obvious to one of ordinary skill in the art to include in the automatic task completion system with API calls of webpages of Sodhi the API including parameters for the webpage as taught by Jouhier since the claimed invention is merely a combination of old elements, and in combination each element would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that it would produce a predictable result of allowing the system and the website to interact, and allow the system to perform actions on or with the webpage via the API.
Sodhi, as combined with Jouhier, further discloses generating a next website action step of the first sequence of website action steps based on a result of executing the current sequence of function action steps as a result of executing the current website action step ([0082] At step 545, the web page operation indicated by the next action of the response of the language model [i.e., LLM] is implemented or executed to obtain a next web page. The web page operation may be executed using any appropriate techniques. The next web page may or may not require an HTTP request. For example, when the web page operation is to enter text, the web page may be updated by entering text into an input element, but an HTTP request ( e.g., a GET or a POST) may not be performed. For another example, the web page operation may be to click a button to submit a form and the execution of the action may cause an HTTP request to be performed; and
transmitting a result of executing the first sequence of website action steps in response to the query ([0085] If the task is complete, then processing proceeds to step 555, where a result is provided to a user. A result may be provided to the user using any appropriate techniques. In some implementations, a final web page may be presented to the user. In some implementations, a description of the completed task may be presented to the user).
In respect to claim 2, the combined invention of Sodhi and Jouhier disclose the non-transitory, computer-readable storage medium of claim 1, Sodhi further disclosing creating the first plan further comprising sending an initial prompt having in-context learning examples showing a trajectory of user requests for performing actions on websites and responses from the websites to the first LLM ([0055] In FIG. 3, user 310 may use computer 320 to perform a task using a web browser. Computer 320 may use network 330 to communicate with web server 360. For example, web server 360 may be the web server of an airline, and user 310 may make an airline reservation with computer 320. The actions of user 310 may be recorded and used to generate prompt templates for tasks and/or subtasks. [0056] After the web browsing data is collected and prompt templates are generated, the prompt templates may be used with a language model to automate web browsing tasks. User 310 may provide instructions for task to be performed. The user may provide the instructions using any appropriate techniques, such as typing the instructions or speaking the instructions. In some implementations, user 310 may be communicating with another user (not shown), such as a customer support agent, and the instructions may be determined from the conversation between the users. Note that the user who is using the prompt templates to automate a task may be different from the user whose web browsing session was recorded. After prompt templates are created, they may be used with any user and need not be specific to a particular user).
In respect to claim 4, the combined invention of Sodhi and Jouhier disclose the non-transitory, computer-readable storage medium of claim 2, Sodhi further disclosing generating the next website action step comprising sending a subsequent prompt including the result of executing the current website action step to the first LLM ([0083] In some implementations, the next action of the response of the language model may be implemented using a web browser extension or plugin or using a headless browser. The next action may indicate an element or portion of a web page, such as indicated by element attributes ( e.g., a class or identifier) or any other appropriate techniques (e.g., XPath syntax). The next action may also indicate an action to be performed, such as a click or entering text into an input element).
In respect to claim 6, the combined invention of Sodhi and Jouhier disclose the non-transitory, computer-readable storage medium of claim 1, Sodhi further disclosing the first plan including a system-level user action step indicating a user action and including a request for user data ([0057] In some cases, when available, prompt templates may be generated from webpage flow specifications or flow diagrams. Webpage flow specifications may be any data that provide specifications for data entry requirements for each page and/or the user actions necessary to accomplish a task on a page or a series of pages) .
In respect to claim 8, the combined invention of Sodhi and Jouhier disclose the non-transitory, computer-readable storage medium of claim 1, Jouhier further disclosing the one or more sequences of instructions which when executed further cause the one or more processors to perform accessing a web API manifest indicating the current website's API, the web API manifest including a tag or an attribute signaling information regarding a particular function in the current website's API ([0044] A web API manifest comprises, for example, a XML, JSON or YAML file specifying the service operations available to client applications from the web application.; see further [0045] Typically, SOAP web services manifests are expressed in WSDL which is an XML dialect; REST web API manifests are often expressed in swagger (aka. OpenAPI), with two possible syntaxes: YAML or JSON, and GraphQL APIs use a dedicated syntax for its manifest (called a schema). This syntax borrows elements from JSON but is different from JSON; see further [0046] More particularly, a typical web API manifest file may include definitions for the service operations exposed by the web API, as well as definitions for the structures of the request and response payloads of each service operation. Manifests use a powerful type description language that can describe arbitrary types composed of scalar values, arrays and records; [0047] Conventionally, at development time, each web API manifest is used to generate a client library. The client library can be generated automatically from the manifest file to generate a “stub” or can be manually written; see further [0078] As depicted in FIG. 5a, the TypeScript definition file is generated from the relevant web API manifest file in accordance with a definition file generation process. As can be seen in FIG. 5a, a web API manifest file 501 is input to a definition file generation process 502 which generates a corresponding TypeScript definition file 503; see further [0080] In the definition file generation process 502 during a first stage S501, the web API manifest file 501 is parsed to obtain a list of web service operations performed by the web application to which the web API relates along with definitions of the input and output payloads associated with each web service operation).
In respect to claim 9, the combined invention of Sodhi and Jouhier disclose the non-transitory, computer-readable storage medium of claim 1, Sodhi further disclosing the current website's API including a client-side web API that enables manipulation of webpage elements using DOM or computer vision ([0072] At step 510, web browser session data is collected. Any appropriate web browser session data may be collected using any appropriate techniques, such as using any of the techniques described herein. For example, a browser extension may be installed to record web browsing sessions of users. The web browser session data may include sequences of representations of web pages (e.g., HTML or DOM) and actions performed between the web pages (e.g., entering text into a specific element for clicking a specific element with a mouse)).
In respect to claim 10, the combined invention of Sodhi and Jouhier disclose the non-transitory, computer-readable storage medium of claim 1, Sodhi further disclosing the result of executing the current function action step being executable code ([0081] At step 540, a query is submitted to a language model using the prompt to obtain a response from the language model. The query may be submitted using any appropriate techniques, such as any of the techniques described herein. In some implementations, the prompt may be submitted to third-party language model using an API call. The response of the language model may indicate an action to be performed, such as an operation to be performed on a web page (e.g., clicking an element with a mouse or entering text)).
In respect to claim 12, the combined invention of Sodhi and Jouhier disclose the non-transitory, computer-readable storage medium of claim 1, Sodhi further disclosing generating the next function action step comprising sending a prompt including the result of executing the current function action step to the current LLM ([0081] At step 540, a query is submitted to a language model using the prompt to obtain a response from the language model. The query may be submitted using any appropriate techniques, such as any of the techniques described herein. In some implementations, the prompt may be submitted to third-party language model using an API call. The response of the language model may indicate an action to be performed, such as an operation to be performed on a web page (e.g., clicking an element with a mouse or entering text)).
Claims 15, 16, 17, 18, and 20 recite a method reciting the same limitations as those found in claims 1, 2, 4, 6, 8, 9, 10, and 12, above, and are rejected using the same rationale.
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Sodhi et al. (U.S. 2024/0386216, hereinafter Sodhi) in view of Jouhier (U.S. 2020/0409776), in further view of Shribman et al. (US 2020/0356618, hereinafter Shribman).
In respect to claim 3, the combined invention of Sodhi and Jouhier disclose the non-transitory, computer-readable storage medium of claim 2, however they may not explicitly disclose the trajectory including retrying to communicate with a specific website that was unreachable or returned an error message.
Analogous art Sodhi discloses the trajectory including retrying to communicate with a specific website that was unreachable or returned an error message ([0769] In many cases, the failure of a content fetching action may be intermittent or temporary. In such a case, a retry, such as by an additional attempt, of the fetching action may be useful, and the repeated fetching action is expected to result in a proper content fetching. Such attempts may be repeated number of times, until successfully completed. The attempts may be the same as the original process, such as by repeating the same URL and sending to the same destination. Alternatively or in addition, the attempts may use different parameters, attributes, or characteristics, and may use different paths, different routes, or different intermediate devices for fetching the content. Further, different fetching schemes may be used in the retries actions; see further [0784] … Further, the number of attempts to retrieve content by the proxy server 53 from the data server #1 22a may be equal to, or different from, the number of attempts to retrieve content by the client device #1 31a from the proxy server 53. For example, Nmax may be equal 5 for the number of attempts to retrieve content by the proxy server 53 from the data server #1 22a, while Nmax may be equal 7 for the number of attempts to retrieve content by the client device #1 31a from the proxy server 53. Further, Nmax may be equal 1, meaning only a single fetching attempt, for each of the fetching schemes…).
It would have been obvious to one of ordinary skill in the art to include in the n the automatic task completion system the retrying of fetching as taught by Shribman since the claimed invention is merely a combination of old elements, and in combination each element would have performed the same function as it did separately, as it allows number of attempts to retrieve content by the client device (See Shribman [0784]).
Claims 5, 13, and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Sodhi et al. (U.S. 2024/0386216, hereinafter Sodhi) in view of Jouhier (U.S. 2020/0409776), in further view of SODHI, Paloma, et al., (Applicant Information Disclosure Statement, filed 3 June 2025, Non-Patent Literature Documents, Number 1), hereinafter Sodhi P.
In respect to claim 5, the combined invention of Sodhi and Jouhier disclose the non-transitory, computer-readable storage medium of claim 1, and while Sodhi discloses few-shot prompting ([0043], [0045], [0047], [0048]), the combined invention may not explicitly disclose the first plan including a website reasoning step indicating a reasoning of a current state of the first LLM.
Analogous art Sodhi P discloses a website reasoning step indicating a reasoning of a current state of the first LLM5 (page 19 “ We additionally include a reasoning section between the input and output sections of the prompt that forces the LLM to generate a series of short sentences that provide justifications for the actions it predictions. We found this to uniformly improve performance for both task and policy prompts (Appendix B). We subsequently added in a simple reason generator prompt (Appendix G.6) that works across both high-level tasks and low-level policies. The prompt takes the current browser content St, previous actions at-1:t-k, current action at and generates a reason r).
It would have been obvious to one of ordinary skill in the art to include in the in the automatic task completion system of Sodhi the LLM reasoning as taught by Sodhi P, since the claimed invention is merely a combination of old elements, and in combination each element would have performed the same function as it did separately, and provides for uniformly improving performance for both task and policy prompts (Sodhi P, page 19).
In respect to claim 13, the combined invention of Sodhi and Jouhier disclose the non-transitory, computer-readable storage medium of claim 1, and while Sodhi discloses few-shot prompting ([0043], [0045], [0047], [0048]), the combined invention may not explicitly disclose the current plan including a function reasoning step indicating a reasoning of a current state of the current LLM.
Analogous art Sodhi P discloses the current plan including a function reasoning step indicating a reasoning of a current state of the current LLM6 (page 19 “ We additionally include a reasoning section between the input and output sections of the prompt that forces the LLM to generate a series of short sentences that provide justifications for the actions it predictions. We found this to uniformly improve performance for both task and policy prompts (Appendix B). We subsequently added in a simple reason generator prompt (Appendix G.6) that works across both high-level tasks and low-level policies. The prompt takes the current browser content St, previous actions at-1:t-k, current action at and generates a reason r).
It would have been obvious to one of ordinary skill in the art to include in the in the automatic task completion system of Sodhi the LLM reasoning as taught by Sodhi P, since the claimed invention is merely a combination of old elements, and in combination each element would have performed the same function as it did separately, and provides for uniformly improving performance for both task and policy prompts (Sodhi P, page 19).
In respect to claim 14, the combined invention of Sodhi and Jouhier disclose the non-transitory, computer-readable storage medium of claim 1, however the combined invention may not explicitly disclose the current plan including a website-level user action step indicating a user action and including a request for user data.
Analogous art Sodhi P discloses the current plan including a website-level user action step indicating a user action and including a request for user data (page 4; page 19 “This is a single step task, that requires the source and destination airports and flight dates to be entered in the Ut and the search button to be clicked”; FIG. 1).
It would have been obvious to one of ordinary skill in the art to include in the automatic task completion system of Sodhi the request for user data as taught by Sodhi P since the claimed invention is merely a combination of old elements, and in combination each element would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that it would produce a predictable result of requesting needed user input to perform actions.
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Sodhi et al. (U.S. 2024/0386216, hereinafter Sodhi) in view of Jouhier (U.S. 2020/0409776), in further view of Gray et al. (U.S. 11,886,828, hereinafter Gray).
In respect to claim 7, the combined invention of Sodhi and Jouhier disclose the non-transitory, computer-readable storage medium of claim 1, however the combined invention may not explicitly disclose the one or more sequences of instructions which when executed further cause the one or more processors to perform: sending an enhancing prompt indicating a request for additional information based on the query and data of a user account for a user associated with the query to a second LLM; replacing the query by output data from the second LLM7.
Analogous art Gray discloses the one or more sequences of instructions which when executed further cause the one or more processors to perform: sending an enhancing prompt indicating a request for additional information based on the query and data of a user account for a user associated with the query to a second LLM; replacing the query by output data from the second LLM (col 2, lines 28 – 43: the additional content that is processed, using the LLM in generating the NL based surmnary to provide responsive to submission of a query, includes: content from query-responsive search result document(s) that are responsive to the query; and/or content from other search result document(s) that are each responsive to a corresponding other query, such as another query deter- 35 mined to have a relationship to the query and/or to the submission of the query. The one or more other queries can include one or more related queries (e.g., often issued, among a population of users, in close temporal proximity to the query), one or more recent queries (e.g., submitted 40 within close temporal proximity of the submission of the query and/or having topical overlap with the query), and/or one or more implied queries ( e.g., automatically generated based on, for example, context and/or profile data)…; col 2, line 59 – col 3: line assume a given query is submitted, such 60 as a given query formulated and submitted based on user input, or an implied query that is automatically formulated and optionally automatically submitted. In response to submission of the given query, a search can performed for the given query to obtain query-responsive search result documents, a search can be performed for a related query to generate related-query-responsive search result documents, and recent-search-responsive search result documents that were responsive to a recent query can be obtained. Further, search result documents A and B can be selected from the query-responsive search result documents, search result document C can be selected from the related-query-responsive search result documents, and search result document D can be selected from recent-search-responsive search result
documents. Yet further, content A can be selected from search result document A, and contents B, C, and D can be selected from respective ones of search result documents B, C, and D. Contents A, B, C, and D can then be included in the additional content that is processed using the LLM in generating the NL based summary to provide responsive to submission of the query. For instance, a prompt of "Summarize <Content A>, <Content B>, <Content C>, and <Content D>" (which omits the query itself) can be processed using the LLM to generate the NL based summary. Also, for instance, a prompt of"In the context of <query>, summarize <Content A>, <Content B>, <Content C>, and <Content D>" can be processed using the LLM to generate the NL based summary; see further col 9, line 7 – 19: As another example, the implied input engine 114 can generate an implied query based on profile data ( e.g., an implied query related to an interest of a user), submit the query at regular or non-regular intervals, and cause corresponding result(s) for the submission(s) to be automatically provided (or a notification thereof automatically provided). For instance, the implied query can be "patent news" based on profile data indicating interest in patents, the implied query periodically submitted, and a corresponding NL based summary result automatically rendered. It is noted that the provided NL based summary result can vary over time in view of e.g., presence of new/fresh search result document(s) over time).
It would have been obvious to one of ordinary skill in the art to include in the automatic task completion system if Sodhi the request for additional information based on the query and data of a user account as taught by Gray since the claimed invention is merely a combination of old elements, and in combination each element would have performed the same function as it did separately, and would allow the NL based summary to provide responsive to the submission of the query (Gray, col 2).
Claims 11 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Sodhi et al. (U.S. 2024/0386216, hereinafter Sodhi) in view of Jouhier (U.S. 2020/0409776), in further view of GUO, Zhen, et al., (Applicant Information Disclosure Statement, filed 3 June 2025, Non-Patent Literature Documents, Number 3), hereinafter Guo.
In respect to claim 11, the combined invention of Sodhi and Jouhier disclose the non-transitory, computer-readable storage medium of claim 1, while Sodhi discloses functions of the current website's API and responses from the current website (e.g., [0081]), it may not explicitly disclose creating the current plan further comprising fine-tuning a certain LLM with training data showing a trajectory of parameterized calls of functions of the current website's API and responses from the current website8.
Analogous art Guo discloses creating the current plan further comprising fine-tuning a certain LLM with training data showing a trajectory of parameterized calls of functions of the current website's API and responses from the current website (page 6: “we conduct a scaling experiment to investigate whether more API data improves a model's generalization ability for unseen API data. For this experiment we fine-tune models on progressively larger API datasets, all with unique API calls in Python. Specifically, we fine-tuned models with 20k, 40k, 80k, and 100k instances respectively. Our hypothesis is that exposure to a greater diversity of APIs during fine-tuning will improve the model's ability to generalize to new, unseen APIs… We integrate a subset of 50,000 entries from API Pack into the Magicoder datasets (Magicoder-OSS-Instruct plus Magicoder-EvolInstruct) and fine-tune the CodeLlama-13b model… Table 3 shows the evaluation results for the four models finetuned with 20,000 API Pack instances in Python. Overall, note that the fine tuning of CodeLlama-13b excels in API call (code generation) for the three-shot retrieved setting, where it achieved the top rates for the three evaluation levels (55.5% in Level 1, 51.4% in Level 2, and 49.5% in Level 3). Also note that the models fine-tuned with the threeshot template consistently show better performance than those fine-tuned with the zero-shot template. This result suggests that bootstrapping data with the three-shot template is important to improve the model's in-context learning abilities. Another key insight is the substantial improvement observed prompting with 3-shot (retre) versus 0-shot at testing time. This trend is consistent across all models and levels, indicating that providing models with relevant examples improves their accuracy in generating API calls…).
It would have been obvious to one of ordinary skill in the art to include in the API and responses from a current website of Sodhi the fine-tuning of LLM with training data showing as taught by Guo since the claimed invention is merely a combination of old elements, and in combination each element would have performed the same function as it did separately, and since it improves the model's in-context learning abilities (Guo, pg. 6).
Claim 19 recites a method reciting the same limitations as those found in claim 11, above, and is rejected using the same rationale.
Conclusion
The prior art made of record and not relied upon considered pertinent to Applicant’s disclosure.
Taheri; Shahriar US 20250117630 A1 Chatbot With Llm And Vectorized Database
Yadav; Nikhil et al. US 20210019309 A1 Mapping Natural Language To Queries Using A Query Grammar
Ross; Steven et al. US 20190179940 A1 Generating And/Or Utilizing A Machine Learning Model In Response To A Search Request
Ray; Soham et al. US 20240193190 A1 Automated Key-Value Extraction Using Natural Language Intents
Chen; Hailin et al. US 20240249077 A1 Systems And Methods For In-Context Learning Using Small-Scale Language Models
Tan; Li et al. US 20240303711 A1 Conversational And Interactive Search Using Machine Learning Based Language Models
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ALAN S MILLER whose telephone number is (571)270-5288. The examiner can normally be reached on M-F 10am-6pm. Examiner’s fax phone number is (571) 270-6288.
Examiner interviews are available via telephone 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, Beth Boswell can be reached at (571) 272-6737. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/ALAN S MILLER/Primary Examiner, Art Unit 3625
1 See, e.g., Recentive Analytics, Inc. v. Fox Corp. (Fed. Cir. 2025), which found that the claimed inventions directed to the use of a Machine Learning in different environments to be directed to an abstract idea.
2 See also Recentive Analytics, Inc. v. Fox Corp. (Fed. Cir. 2025), page 15 “…We have consistently held, in the context of computer-assisted methods, that such claims are not made patent eligible under § 101 simply because they speed up human activity. See, e.g., Content Extraction, 776 F.3d at 1347; DealerTrack, 674 F.3d at 1333. Whether the issue is raised at step one or step two, the increased speed and efficiency resulting from use of computers (with no improved computer techniques) do not themselves create eligibility. See, e.g., Trinity Info Media, LLC v. Covalent, Inc., 72 F.4th 1355, 1363 (Fed. Cir. 2023) (rejecting argument that “humans could not mentally engage in the ‘same claimed process’ because they could not perform ‘nanosecond comparisons’ and aggregate ‘result values with huge numbers of polls and members’”) (internal citation omitted); Customedia Techs., LLC v. Dish Network Corp., 951 F.3d 1359, 1365 (Fed. Cir. 2020) (holding claims abstract where “[t]he only improvements identified in the specification are generic speed and efficiency improvements inherent in applying the use of a computer to any task”)…”.
3 “It is well-settled that mere recitation of concrete, tangible components is insufficient to confer patent eligibility to an otherwise abstract idea. Rather, the components must involve more than performance of “‘well understood, routine, conventional activit[ies]’ previously known to the industry.” Alice, 134 S. Ct. at 2359 (quoting Mayo, 132 S.Ct. at 1294)”. Id, pages 10-11. “Likewise, the server fails to add an inventive concept because it is simply a generic computer that “administer[ s]” digital images using a known “arbitrary data bank system.” Id. at col. 5 ll. 45–46. But “[f]or the role of a computer in a computer-implemented invention to be deemed meaningful in the context of this analysis, it must involve more than performance of ‘well-understood, routine, [and] conventional activities previously known to the industry.’” Content Extraction, 776 F.3d at 1347–48 (quoting Alice, 134 S. Ct at 2359). “These steps fall squarely within our precedent finding generic computer components insufficient to add an inventive concept to an otherwise abstract idea. Alice, 134 S. Ct. at 2360 (“Nearly every computer will include a ‘communications controller’ and a ‘data storage unit’ capable of performing the basic calculation, storage, and transmission functions required by the method claims.”); Content Extraction, 776 F.3d at 1345, 1348 (“storing information” into memory, and using a computer to “translate the shapes on a physical page into typeface characters,” insufficient confer patent eligibility); Mortg. Grader, 811 F.3d at 1324–25 (generic computer components such as an “interface,” “network,” and “database,” fail to satisfy the inventive concept requirement); Intellectual Ventures I, 792 F.3d at 1368 (a “database” and “a communication medium” “are all generic computer elements”); BuySAFE v. Google, Inc., 765 F.3d 1350, 1355 (Fed. Cir. 2014) (“That a computer receives and sends the information over a network—with no further specification—is not even arguably inventive.”)”. TLI Communications LLC v. AV Automotive L.L.C., (No. 15-1372, (Fed. Cir. May 17, 2016)), at *12-13.
See additionally MPEP 2106.05(d).
4 “Nor, in addressing the second step of Alice, does claiming the improved speed or efficiency inherent with applying the abstract idea on a computer provide a sufficient inventive concept. See Bancorp Servs., LLC v. Sun Life Assurance Co. of Can., 687 F.3d 1266, 1278 (Fed. Cir. 2012) (“[T]he fact that the required calculations could be performed more efficiently via a computer does not materially alter the patent eligibility of the claimed subject matter.”); CLS Bank, Int’l v. Alice Corp., 717 F.3d 1269, 1286 (Fed. Cir. 2013) (en banc) aff’d, 134 S. Ct. 2347 (2014) (“[S]imply appending generic computer functionality to lend speed or efficiency to the performance of an otherwise abstract concept does not meaningfully limit claim scope for purposes of patent eligibility.” (citations omitted))”. Intellectual Ventures I LLC v. Capital One Bank (USA), 792 F.3d 1363, 115 U.S.P.Q.2d 1636 (Fed. Cir. 2015).
5 Noting Applicant’s specification merely discloses - [0033] In some embodiments, the server 102 is programmed to interleave additional website reasoning steps with website action steps. The server 102 can then be programmed to also obtain in-context learning examples and guide performance of a trained LLM in accordance with the ReAct system discussed in the paper using few-shot prompting, as also illustrated in Table 6 of the paper. The website reasoning steps can be generated as appropriate without requiring additional prompts following the initial prompt; [0055] In some embodiments, the first plan includes a website reasoning step indicating a reasoning of a current state of the first LLM. In other embodiments, the first plan includes a system-level user action step indicating a user action and including a request for user data.
6 Noting Applicant’s specification merely discloses – [0045] In some embodiments, the user device 130 is programmed to plan and execute one or more function action steps (possibly interleaved with additional function reasoning steps) using a trained LLM… [0046] … As noted above, the user device 130 can be configured to interleave additional function reasoning steps with function action steps. The function reasoning steps can be generated as appropriate without requiring additional prompts;
7 Noting Applicant’s specification merely discloses - [0053] In some embodiments, the system is configured to send an enhancing prompt indicating a request for additional information based on the query and data of a user account for a user associated with the query to a second LLM. The system is further configured to replace the query by output data from the second LLM.
8 Noting Applicant’s disclosure merely discloses - [0059] “In some embodiments, in creating the current plan, the system is configured to finetune a certain LLM with training data showing a trajectory of parameterized calls of functions of the current website's API and responses from the current website”.