Computer-implemented method for providing responses to user queries

Abstract

To provide an improved method for providing responses to user queries.SOLUTION: In order to solve the problem, a method according to the present invention includes: receiving a user query from a user by an integrated control system; generating a pricing request based on the user query and / or parameters attached to the user; sending the pricing request to a large-scale language model, which is requesting the large-scale language model to generate a price indication; receiving the price indication generated by the large-scale language model; determining a response price based on the price indication; requesting confirmation from the user to allocate the response price; receiving the confirmation to allocate the response price; allocating the response price by using a payment system; and sending a response related to the user query to the user.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] explanation This application relates to a computer-implemented method for providing responses to user queries. [Background technology]

[0002] The increasing use of modern chatbot systems, especially those based on generative artificial intelligence (AI) models such as ChatGPT (Chat Generative Pre-trained Transformer), has encouraged continuous improvement of these chatbots to enhance their functionality and optimize the user experience.

[0003] Generative AI models that employ AI techniques to understand and process human language text and generate human-like responses to user queries and inputs are commonly known as large-scale language models (LLMs). With the continued advancement of technology, the growing need for LLM-based generative AI that extends beyond text-based interactions has led to the research, development, and integration of multimodal features, allowing users to interact with generative AI through audio or video inputs. To achieve this, generative AI tools commonly utilize LLMs in combination with speech-to-text models capable of converting audio or video audio into text. To develop these capabilities, LLMs are trained with substantial amounts of text, audio, and video data, which requires significant computational power and energy consumption that is expected to further increase over time. Beyond the environmental impact of LLMs, growing user preferences and expectations for improved quality, usefulness, and personalization of generated responses trigger a search for solutions that make LLMs more economically feasible, useful, and tailored to users.

[0004] Typically, LLMs operate without access to personal user information, and the responses they generate are not automatically customized to individual needs and preferences. Thus, in order for LLMs to generate more personalized responses, users are required to explicitly include all relevant information in their queries, which may be perceived by the user as unproductive, leading to potential user frustration, or having the effect of limiting the quality of the responses provided. Specifically, attributes that could potentially increase response quality may not be considered by the user.

[0005] The quality of the generated LLM response can be affected by a variety of factors, including the quality of the query or input, the amount and variety of data used to train the model, and the degree to which that data is relevant to the query or input. Furthermore, if the training data is not carefully pre-processed, there is a high probability that the training data will cause the LLM to learn inaccurate information, exhibiting biased behavior or experiencing what is known as AI hallucination.

[0006] Furthermore, some LLMs operate using data compiled from public sources and do not use information from proprietary internet pages, real-time data, or specific databases that often require payment of a fee or subscription for access. Access to these LLMs is typically unrestricted. In general, unrestricted use of LLMs may lead to excessive or irresponsible use of their computing resources, further contributing to their environmental footprint.

[0007] LLMs can also be trained with data that includes copyrighted material and / or may require payment for generating responses to user queries. Generally, paid use of an LLM can provide various benefits, including service improvements, feature advancements, and continuous development of the system. Additionally, generated revenues can be invested in clean energy sources, for example, to power data centers hosting the LLM. Payment-based LLMs, such as ChatGPT-4, primarily rely on subscription payments to use LLM functionality. However, subscription payments may be prone to high churn rates or may discourage potential users seeking greater flexibility from using the LLM.

[0008] Additionally, subscription payments are flat-rate payments, which can be somewhat unfair given that different users have different needs and preferences, or provide varying levels of detail and complexity in their queries or inputs. Thus, users looking for simpler information are charged the same fee as those asking more complex or personalized questions. Furthermore, reports from AI service providers indicate that there is enormous demand for AI services, exceeding the available hardware resources in many situations. Summary of the Invention [Problem to be solved by the invention]

[0009] It is therefore an object of the present invention to provide an improved method for providing responses to user queries, in particular where there is a need to appropriately allocate available hardware resources to improve the available service, e.g., enhance responsiveness.

[0010] In particular, it is an object of this application to provide an alternative method for charging for responses to user queries generated by a generative AI system.

[0011] Furthermore, an object of the present invention may be to establish an autonomous self-assessment method adapted to users to charge for responses, ensuring a seamless and safe user experience. [Means for solving the problem]

[0012] The present invention solves the above-mentioned problem by means of a method according to claim 1.

[0013] In particular, the problem is solved by a method for providing a response, the method comprising: - receiving a user query from a user by the integrated control system; - generating a bid request based on a user query and / or user-attached parameters; - sending a pricing request to the large-scale language model, where the integrated control system requests the large-scale language model to generate price instructions, in particular prices; - receiving, by an integrated control system, price instructions generated by the large-scale language model; - determining a response price based on the price indication; - requesting confirmation from the user for allocating a response price; - receiving, by the integrated control system, confirmation for allocating the response price; - allocating a response price, preferably using a payment system; - sending a response to the user related to the user query; It consists of:

[0014] Queries are user-initiated and may be issued through a client device, such as a mobile device, a personal computer, or any other type of digital device. However, a client device need not be dedicated hardware assigned to a single user. Indeed, it is possible for the client to interact with the same system that also hosts, at least in part, the LLM and / or any other components described above.

[0015] A user query is generally a question or prompt requesting information or assistance on a particular topic and may be provided as text, images, audio or video data, and / or a combination of these modalities. Correspondingly, a response to a user query may take the form of text, images, audio / video data, or a combination of these modalities. The received user query may be generated, for example, via a web interface. Alternatively, the received user query may be generated via any other means, for example, word processing software such as Microsoft Word, software running in the car, a third-party interface, an application programming interface (API), a messaging app, or any other type of device or interface.

[0016] It is an aspect of the present invention to use a generative AI, particularly an LLM, to arrive at a price indication for a user query that is also answered using an LLM. The LLM used to establish the price indication may be the same LLM used to process the user query and provide a response. Alternatively, the LLM used to generate the price indication may be a separate LLM or any other generative AI mechanism.

[0017] A price indication can be defined as the (estimated) cost of responding to a query and can include the individual costs associated with each factor that influences pricing. A price indication can also include the estimated total cost of the response. Alternatively, a price indication can be a combination of these approaches and can in particular include the individual costs of each factor that influences pricing and the estimated total cost of the response.

[0018] Allocation of a price, particularly a response price, can refer to the process of reserving from a user an amount or value, for example in a digital wallet, equal to the amount of a price indication. The allocation process does not necessarily initiate a payment by money transfer. For example, the actual payment may be made only when the allocated amount exceeds a certain threshold. One advantage of allocation is that it reduces the number of individual transactions and transaction fees when charging for response generation.

[0019] One general advantage of the present invention is that the overall performance of the service can be enhanced thanks to the revenue collected. For example, the revenue can be used to cover costs associated with hardware components, balance the overall system load, and / or provide a scalable infrastructure that can accommodate increased user demand and improve responsiveness. Furthermore, fees can be paid to copyrighted material sources and used in response generation, potentially improving response quality.

[0020] The price indication for the response provided by the LLM is established using a request for quote sent to the LLM.

[0021] In one embodiment, generating a request for a quote comprises: - Use a pricing model template and / or - Use natural language processing models and / or - Use feedback mechanisms may include The feedback mechanism is - receiving user feedback related to the response and / or response price; - storing user feedback in a history database; - analyzing user feedback, for example to identify the most frequent user complaints; - Adapting pricing requests based on analyzed user feedback Includes:

[0022] Pricing model templates are intended to provide a clear and well-defined framework for generating pricing requests. Pricing model templates are structured to instruct or guide the LLM to generate pricing models for response to user queries.

[0023] In particular, the template may instruct the LLM to generate a pricing model based on provided criteria, which may include responsive performance indicators and / or data items.

[0024] Response performance indicators may refer to primarily subjective qualitative measures, and may include, by way of example and not limitation, usefulness, feasibility, clarity and relevance, scientific value, entertainment value, potential impact on the user, or any other qualitative measure that can serve as a basis for assessing response quality and allowing price indications to be provided accordingly.

[0025] Data items may refer specifically to user data items, such as, but not limited to, age, gender, nationality, education, or any other data item that provides personal information about a user. Data items may also refer, for example, to data items extracted from or related to user browsing and search history, social media, asset purchases, etc. Data items may also include, for example, quantifiable data items, such as location, weather, date and time, or any other quantifiable data item that may be relevant to generating a price for a response. In one embodiment, some data items, particularly user data items, may originate directly from the user. For example, a user may be prompted by the platform to register for a profile or account and provide personal information.

[0026] In one embodiment, some data items, such as device information or location, are collected autonomously by the system. The system may also be provided with access to third-party cookies associated with the user that can be used to collect behavioral data items. In one embodiment, some data items are received by a payment system. For example, the payment system may store data about previous user purchases and analyze that data to gain insight into user preferences. Additionally, the system may collect data items from various digital content providers to specifically collect data not associated with the user.

[0027] According to one aspect of the present invention, the method can assign and dynamically adjust weights to responsive performance indicators based on at least one data item. An advantage of this is that the method can consider the relative importance of a responsive performance indicator relative to, for example, a user's location, educational background, or employment status. In one embodiment, the method can dynamically adjust the weights assigned to responsive performance indicators based on any combination of data items. Thus, responsive performance indicators with higher importance will be assigned a greater weight than indicators with lower importance. For example, responsive performance indicators that may not be considered significant, e.g., responsive performance indicators with weights below a certain threshold, are not included as parameters in the pricing request.

[0028] For example, price instructions provided by an LLM that are adjusted according to a user's location may take into account the cost of living in a particular location, thus providing economically more equitable access to services. In another example, price instructions that are adjusted according to a user's educational background or employment status may establish a more personalized pricing strategy, such as offering discounts to users with low or no income, thus making services more affordable. Alternatively, adapting weights for any combination of user profile data items allows for finer tuning of the desirability of indicators. In one embodiment, the weighting of responsiveness indicators is performed using a neural network.

[0029] By adapting pricing requests based on data items, customized prices can be provided to clients, potentially reducing churn and attracting new users. Additionally, given that some data items change dynamically, the method is able to capture variations in user behavior and adapt pricing requests accordingly. Another advantage is that, for example, analysis of a user's previous purchases can provide insight into the price a user is willing to pay for a particular service or product, and thus adapt pricing requests to maximize revenue. Naturally, the LLM may be required to take previous requests—the course of communication—into account in order to determine appropriate pricing instructions.

[0030] Pricing model templates can generally be based on a framework that includes placeholders, which can take any value or variable that is relevant or important to the user, such as age, location, or any other attribute, criterion, or characteristic, and can be selected to provide a more personalized pricing request.

[0031] Providing variables or values for placeholders may be performed automatically, for example, using a programming function that inserts variables or values into designated placeholders.

[0032] Alternatively, the pricing model template can be automatically adapted using a natural language processing (NLP) model, e.g., a pre-trained NLP algorithm, such as GPT, BERT (Bidirectional Encoder Representations from Transformers), or any other pre-trained model. Alternatively, the NLP model and / or algorithm may be trained on a specific dataset relevant to generating pricing requests or may rely on any type of data or rules, e.g., linguistic or statistical rules. To modify the pricing model template, an NLP model can be utilized to fill each placeholder with the most contextually appropriate parameter from a plurality of highly relevant parameters.

[0033] To improve the quality of the quote requests, the method further includes a feedback mechanism that uses client feedback to refine the quote requests.

[0034] In one embodiment, the feedback mechanism may be incorporated as a rating system, where the user is asked to rate one or various categories relating to the quality of the response to the user's query on a predetermined scale. As an example, the feedback mechanism may include criteria such as the degree to which a response is useful or important according to the user, or whether the user considers a given price indication to be appropriate for the response. Alternatively, the feedback mechanism may take the form of a survey including open-ended questions that allow the user to provide more detailed and / or personalized feedback.

[0035] In general, the feedback mechanism may be implemented in any manner suitable for receiving feedback that can be used to improve the generation of requests for quotes.

[0036] The feedback can be stored in a historical database and analyzed to identify the most frequent complaints, based on which pricing requests should be modified.

[0037] To provide the user with a price to allocate, the method determines a response price, which represents the cost allocated to providing a response to a given query or input.

[0038] In particular, determining response prices is - Calculating the total amount based on price indications; - Use price instructions to establish a response price range It may include at least one of:

[0039] After determining the response price, the method may request confirmation from the user to allocate the response price to provide the generated response.

[0040] In one embodiment, the method also employs an advertising-supported model to provide the user with a request to view the video, and if the user accepts to view the video, the method can further provide the user with several options.

[0041] In particular, the method comprises the following steps: - sending a request to the user to view the video; - determining whether the user has watched the video, and if the user has watched the video, - reducing the response price by a specified amount and allocating the reduced price; and / or - Allocating credits to users.

[0042] Credits allocated to a user may be stored, for example, in a digital wallet integrated with a payment system. In one embodiment, credits may be used to pay response prices generated for different responses to a query. Alternatively, credits may be used to purchase assets from any asset provider that may recognize credits as a valid payment currency.

[0043] This feature provides users with more flexibility regarding their payment preferences and serves as an additional reliable source of revenue for platforms utilizing this method step. Furthermore, the adoption of an ad-supported model provides technical advantages, particularly with regard to data collection and processing. Because ad-supported models routinely collect and analyze user data to improve services, one technical advantage is that the system can gain computational efficiency by directly implementing user data that has already been processed by the ad-supported model.

[0044] At least some of the above given problems are also solved by a computer readable medium storing instructions that, when executed by at least one processor, cause the at least one processor to perform one of the methods / embodiments as described above.

[0045] The object of the present invention is further solved by a system for providing a response to a user query, the system being an integrated control system comprising: - receiving a user query from a user; - generating a bid request based on a user query and / or user-attached parameters; - sending a pricing request to the large-scale language model and requesting the large-scale language model to generate a price indication, in particular a price; - receiving price indications generated by a large-scale language model; - determining a response price based on the price indication; - requesting confirmation from the user for allocating a response price; - sending a response to the user related to the user query; An integrated control system adapted to:

[0046] Similar or identical benefits arise with respect to this system as those described with respect to the method above.

[0047] Further embodiments are discussed in the dependent claims.

[0048] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. [Brief explanation of the drawings]

[0049] [Figure 1] Shown are users, an integrated control system, a large-scale language model server, and a payment system connected through the Internet. [Figure 2] 1 shows an exemplary diagram of a method for providing a response to a user query according to the present invention; [Figure 3] 1 illustrates components of an integrated control system used to implement a method for providing a response to a user query. [Figure 4] 1 illustrates an exemplary pricing request utilized to generate price indications in response to a user query. DETAILED DESCRIPTION OF THE INVENTION

[0050] In the following description, the same reference signs may be used for the same parts and for (different) parts with the same effect.

[0051] 1 shows a user 10 connected to an integrated control system 20 and a payment system 40 for making preferably micropayments over the Internet 1. The integrated control system 20 comprises a user interface that provides the user 10 with seamless access to the integrated control system 20. The user 10 interacts with the integrated control system 20 through a client device, such as, but not limited to, a mobile device, a PC, or any other hardware device or component compatible with the integrated control system 20. The integrated control system 20 is also communicatively connected to a large-scale language model server 30 that hosts a large-scale language model 31 and a payment system 40.

[0052] The integrated control system 20 may act as a gateway to the large scale language model server 30 and the payment system 40. Alternatively, the payment system 40 may communicate directly with the user 10.

[0053] The integrated control system 20 receives queries from the users 10, processes the queries, and forwards the queries to the large scale language models 31 hosted on the large scale language model servers 30. Accordingly, the integrated control system 20 receives responses from the large scale language model servers 30, the responses being generated by the large scale language models 31 and relating to the queries.

[0054] The integrated control system 20 may require the user to register an account for access. As part of the registration process, the integrated control system 20 may require the user to provide personal information, which may refer to one or more data items such as age, banking information, education status, etc. The personal information may be collected, for example, using a questionnaire. Additionally, the integrated control system 20 may be able to autonomously gather information, for example, by accessing third-party cookies, by resolving IP and / or MAC addresses, by analyzing the type of software the user is using (which operating system is installed on the user's 10 client device), or by any other means of autonomously gathering information. The integrated control system 20 may also implement advanced security features, such as two-factor or biometric authentication, to provide secure access to functionality. Alternatively, the user may register with the payment system 40, and the integrated control system 20 may retrieve the user's credentials and / or personal information from the payment system 40.

[0055] Whether the user is registered or not, the integrated control system 20 assigns a unique identification number (ID) to the user. The ID number is then stored in a general database 25 (see FIG. 3) along with other data, such as user credentials and personal information. The ID number is also provided to a history database 24 (see FIG. 3), which serves as a detailed historical repository of user activity, particularly interactions with the integrated control system 20 and / or payment system 40. The ID number can be used as a common identifier to retrieve information about the user from the history database 24 and the general database 25.

[0056] FIG. 2 illustrates an exemplary method according to the present invention.

[0057] In an initial step, the integrated control system 20 receives a user query Q in the form of text, audio, and / or video from the user 10 via a user interface.

[0058] The integrated control system 20 stores the user query Q in its history database 24 and analyzes the user query Q to extract information therefrom. For example, the integrated control system 20 may be capable of executing a natural language processing model stored in memory 22 (see FIG. 3 ), which performs a semantic analysis on the user query Q to identify personal information, user preferences, or trends in the user's requests to the large-scale language model 31. Furthermore, the semantic analysis may be used to identify the underlying reason for or motivation behind the user query Q and adapt the pricing request PQ for the user query Q according to the user's particular intent.

[0059] In one embodiment, the natural language processing model is provided as part of integrated control system 20. In another embodiment, the natural language processing model is implemented in a separate component that is communicatively coupled to integrated control system 20.

[0060] In one embodiment, the integrated control system 20 sends a request to the large scale language model 31 instructing the large scale language model 31 to perform a (semantic) analysis on the user query Q and return the processed information to the integrated control system 20. In each embodiment, the extracted information is stored in the history database 24 to build a more detailed data set for the user.

[0061] The integrated control system 20 then generates a pricing request PQ, typically in text form.

[0062] Upon generating the pricing request PQ, the integrated control system 20 sends the pricing request PQ to the large-scale language model 31 (see FIG. 2), which generates / initializes a pricing model as instructed by the pricing request PQ.

[0063] In one embodiment, the pricing request PQ includes a user query Q, and the large-scale language model 31 generates a pricing model using information contained in the user query Q. The information in the pricing request PQ includes—in addition to the user query Q—criteria on the basis of which the large-scale language model 31 generates the pricing model. The criteria may include qualitative measures, such as complexity, impact, personalization level, etc., and / or data about the user, such as age, nationality, location, etc.

[0064] The large-scale language model 31 interprets the user query Q and the information sought by the user query Q, and evaluates the user query Q based on criteria provided in the pricing request PQ. For example, the pricing request PQ may request / instruct the large-scale language model 31 to generate pricing models for the user query Q "What is the capital of the UK?", "Evaluate complexity and usefulness (=performance indicators)", and "Consider that the user is a student (=background information = data item)."

[0065] In a next step, the integrated control system 20 receives the pricing model - a descriptive text describing the pricing model - and a price indication generated by the large-scale language model 31 for the user query Q, the price indication being in particular a price P, for example 10 euro cents.

[0066] In one embodiment, integrated control system 20 receives prices P in the form of individual prices for each criterion included in the pricing model. Continuing with the example above, the pricing model may establish the following prices P for user query Q "What is the capital of the UK?" Complexity: 1 euro cent Utility: 2 euro cents

[0067] The integrated control system 20 processes the received price P and determines a response price RP. In one embodiment, the integrated control system 20 determines the response price RP by calculating the sum of the individual prices, or may simply pass on the price provided by the large-scale language model 31, for example, 10 euro cents as described above.

[0068] After generating the response price RP, the accumulation control system 20 transmits the response price RP to the user 10 and requests confirmation C to allocate the response price RP. For example, the accumulation control system 20 may request confirmation C and transmit the response price RP through a modal window overlaid on existing content, the modal window including the response price RP and interactive buttons such as "accept" or "reject" allocation of the response price RP.

[0069] In a next step, according to one embodiment of the present invention, the integration control system 20 receives a confirmation C from the user to allocate the response price RP. If the user rejects the allocation, for example by clicking a "Reject" button, the integration control system 20 may display a message acknowledging the rejection and requesting the user to provide a new query. If the user accepts the allocation, the integration control system 20 asks the payment system 40 to allocate the response price RP by sending an allocation request RAll to the payment system 40. The allocation request RAll may include the response price RP and / or the user's ID and is constructed to prompt the payment system 40 to initiate the allocation. For example, the allocation request RAll may be an HTTP request sent to a dedicated API endpoint of the payment system 40. The allocation request RAll may be provided in any format that is acceptable to the payment system 40 and meets the requirements of the payment system 40.

[0070] After the integration control system 20 sends an allocation request RAll to the payment system 40, the payment system 40 allocates all response prices RP from the users.

[0071] In one embodiment, the payment system 40 initiates a payment transaction from a user only if the allocated response price RP exceeds a certain threshold. If the response price RP is below the threshold, the payment system 40 can store the response price RP, for example, in a digital wallet (dedicated database), and add the response price RP to the response prices RP provided to different user queries Q. A payment transaction is initiated when the cumulative amount exceeds the threshold.

[0072] After the payment system 40 allocates all the response prices RP from the users, the accumulation control system 20 receives an allocation confirmation CA11 from the payment system 40. For example, the accumulation control system 20 can receive a notification by the payment system 40 each time the payment system 40 performs an allocation, the notification including the user's ID.

[0073] Upon receiving the allocation confirmation CAll, the integrated control system 20 sends a request AQ to the large scale language model 31 providing an answer A to the user query Q.

[0074] When integrated control system 20 receives response A from large-scale language model 31, in one embodiment, it simply forwards response A to user 10.

[0075] 3 shows the individual components of integrated control system 20. In the described embodiment, integrated control system 20 includes a processor 21 configured to execute instructions stored in memory 22, a communication interface 23 that enables integrated control system 20 to communicatively interact with users 10, a large-scale language model 31 hosted on a large-scale language model server 30, and a payment system 40. Additionally, integrated control system 20 includes a history database 24 and a general database 25 for storing data from the client devices of users 10, the large-scale language model 31, and the payment system 40.

[0076] The memory 22 stores an instruction set that implements a process control algorithm 27. When executed by the processor 21, the process control algorithm 27 comprises a method such as that described in relation to FIG.

[0077] In one embodiment of the present invention, the historical database 24 and the general database 25 are dedicated blocks of the integrated control system 20 that store data utilized in generating the price requests PQ.

[0078] In one embodiment, the history database 24 and the general database 25 can be cloud databases, and one of the steps of the process control algorithm 27 is to send and receive user data from the cloud databases, for example using an API.

[0079] In one embodiment, the history database 24 also receives and stores data from the payment system 40, the data relating to previous purchases made by the user. The data may be received from the payment system 40, for example, after every purchase made by the user.

[0080] FIG. 4 shows an exemplary pricing request PQ for generating a pricing model according to a large-scale language model 31.

[0081] In this example, the pricing request PQ includes a user query Q and guides the large-scale language model 31 to generate a pricing model based on the user query Q and qualitative measures, namely, entertainment value, practical value, complexity, personalization, and user impact (also referred to herein as performance indicators).

[0082] An exemplary bid request PQ is provided below: "You will be a market researcher specializing in pricing digital content. Your goal is to develop a cent-by-cent pricing model for the responses provided by ChatGPT to user queries or prompts. Please provide your analysis and pricing suggestions based on the following categories: - Entertainment value: the degree to which consumers enjoy the content; - Practical value: the obvious benefit or usefulness that the answer provides; - Complexity: whether the answer addresses a simple or complex topic; - Personalization: Are answers tailored to individual consumer needs or questions? - Impact: Consider the potential overall effect of the answer on the consumer's life or decision-making process task: For each answer ChatGPT provides to a particular query or prompt, it applies the pricing criteria described above. Provides detailed breakdown and rationale for pricing to the cent for each category. When a pricing model such as the one requested is developed, the pricing model is applied to the query "What is the climate like in Antarctica?"

[0083] In response to an exemplary pricing request PQ, the large-scale language model 31 generates a pricing model and applies the pricing model to the user query Q. The pricing model is summarized as follows: "Entertainment value: Entertainment High (e.g., creative, humorous, engaging): 15 cents Entertainment Medium (e.g., informative and engaging): 10 cents Entertainment Low (e.g., purely informational): 5 cents Practical value: High practical value (e.g., practical advice, solutions): 20 cents Medium practical value (e.g., informative but not very practical): 10 cents Low practical value (e.g., general information): 5 cents Complexity: High complexity (e.g., detailed explanation of a complex topic): 15 cents Medium complexity (e.g., covers a wide range of topics with some depth): 10 cents Low complexity (e.g., easy, straightforward information): 5 cents Personalization: Highly personalized (e.g., adapted to the user's specific needs): 20 cents Medium personalization (addresses the query but is not very tailored): 10 cents Low personalization (e.g., general information): 5 cents Impact: High impact (e.g., potentially life-changing advice): 25 cents Medium impact (e.g., influences decision-making): 15 cents Low impact (e.g., general information with minimal impact): 5 cents Applying the pricing model to the query "What is the climate like in Antarctica?": Entertainment Value: Low (0.5 cents) – query is factually focused Practical Value: Medium (10 cents) - Provides valuable information about the climate Complexity: Medium (10 cents) - Requires some detail to describe the climate Personalization: Low (0.5 cents) - Answers are not very tailored to individual needs Impact: Low (5 cents) - General information with minimal impact.

[0084] In one embodiment, the pricing request PQ does not include the user query Q, and after sending the pricing request PQ, the integrated control system 20 sends the user query Q to the large-scale language model 31. In one embodiment, the integrated control system 20 may send only a summary of the original user query Q, for example, "Assuming you were asked about... please provide a pricing model taking... into account."

[0085] In one embodiment, the integrated control system 20 hosts a large language model 31. In particular, the integrated control system 20 stores the large language model 31 in memory 22.

[0086] In one embodiment, the integrated control system 20 can send a response price RP and ask the user to watch the video. For example, the integrated control system 20 can implement an additional interactive button in the modal window that provides the response price RP and a request for confirmation C, e.g., the additional button is "Watch Video."

[0087] In one embodiment, if the user chooses to watch the video by clicking the "Watch Video" button, the integrated control system 20 executes dedicated method steps from the process control algorithm 27 to reduce the response price RP by a specific amount and then requests confirmation C to allocate the reduced price. For example, after reducing the response price RP, the integrated control system 20 may update the modal window to include the reduced price and "Accept" and "Reject" buttons to confirm C the allocation of the reduced price.

[0088] In one embodiment, if the user chooses to watch the video, the integration control system 20 allocates credits to the user and also requests confirmation C to allocate a response price RP. The allocated credits may be stored in the integration system 20, for example in the history database 24, or alternatively may be sent to the payment system 40 and stored in a digital wallet. The credits can be used by the user to pay for other responses or to purchase assets from asset providers that accept that payment method or payment currency.

[0089] In one embodiment, the confirmation to allocate C provided by the user is received directly by payment system 40. Payment system 40 then performs an All step of allocating the response price RP or alternatively the reduced price, and a CAll step of confirming the allocation to aggregation system 20.

[0090] In one embodiment, integrated control system 20 receives a pricing model and price instructions along with receiving a response A to user query Q. For example, pricing request PQ may further include instructions to large scale language model 31 to generate response A to user query Q. Thus, method step AQ of requesting response A to user query Q and the subsequent method step of receiving response A are excluded.

[0091] At this point it should be noted that all of the above-mentioned parts are claimed to be relevant to the present invention when considered alone and in any combination with the details shown in the drawings, in particular: [Explanation of symbols]

[0092] Reference sign: 1. Internet 10 users 20 Integrated Control System 21 processors 22 Memory 23 Communication Interface 24 History Database 25 General Database 27 Process Control Algorithms 30 LLM Servers 31 Large-scale Language Models (LLM) 40 Payment Systems Q Query PQ Pricing Request P Price RP Response Price C. Confirmation A Response AQ How to request a response to a user query RAll allocation requests CAll allocation confirmation All allocation

Claims

1. 1. A computer-implemented method for providing a response, comprising: receiving a user query (Q) from a user (10) by an integrated control system (20); generating a bid request (PQ) based on said user query (Q) and / or parameters attached to said user (10); sending said pricing request (PQ) to a large-scale language model (31), wherein said integrated control system (20) requests said large-scale language model (31) to generate a price indication, in particular a price (P); receiving, by the integrated control system (20), the price indications generated by the large-scale language model (31); determining a response price (RP) based on the price indication; requesting confirmation from said user (10) for allocating said response price (RP); receiving, by said integrated control system (20), a confirmation (C) for allocating said response price (RP); allocating said response price (RP), preferably using a payment system (40); sending a response (A) related to said user query (Q) to said user (10); 1. A computer-implemented method comprising:

2. The method of claim 1 , wherein the confirmation request for allocating the response price (RP) includes the response price (RP).

3. generating the pricing request (PQ) Use a pricing model template and / or Use natural language processing models and / or Use feedback mechanisms Including, The feedback mechanism is receiving user feedback related to said response (A) and / or said response price (RP); storing said user feedback in a history database (24); analyzing the user feedback, for example to identify the most frequent user complaints; Adapting the pricing request (PQ) based on the analyzed user feedback.

3. The method of claim 1 or 2, comprising:

4. A method according to any one of claims 1 to 3, in particular claim 3, wherein the pricing model template comprises at least one placeholder, to which at least one response performance indicator or at least one user profile data item is assigned.

5. 5. The method of claim 1, wherein at least one response performance indicator is provided to the large-scale language model (31), for example as part of the pricing request (PQ), and the pricing request (PQ) causes the large-scale language model (31) to generate the price instruction based on the at least one response performance indicator.

6. 6. The method of claim 1, wherein at least one user profile data item is provided to the large-scale language model (31), for example as part of the price request (PQ), and the price request (PQ) causes the large-scale language model (31) to generate the price indication based on the at least one user profile data item.

7. assigning and / or (dynamically) adjusting, by said integrated control system (20), a weight for at least one responsive performance indicator based on at least one user profile data item; transferring said weights to a large-scale language model (31) for use in generating said price indication; The method according to any one of claims 1 to 6, comprising:

8. Method according to any of claims 1 to 7, in particular claim 7, wherein said assigning and / or (dynamically) adjusting weights is performed using a neural network.

9. Determining the response price (RP) comprises: calculating a total amount based on said price instructions; establishing a response price range using said price indications; The method according to any one of claims 1 to 8, comprising at least one of:

10. Allocating the response price (RP) comprises: sending, by said integrated control system (20), a request for allocation (RAll) to said payment system (40); receiving, by said integrated control system (20), said allocation confirmation (Call) from said payment system (40) and / or said user (10); The method according to any one of claims 1 to 9, comprising:

11. Allocating the response price (RP) comprises: In particular, after receiving an allocation request (RAll) / the allocation request (RAll) from the integrated control system (20), sending a request from the payment system (40) to the user (10) to allocate the response price (RP); receiving, by said payment system (40), from said user (10) a confirmation (C) for allocating said response price (RP); allocating said response price (RP) by said payment system (40); The method according to any one of claims 1 to 10, in particular claim 1, comprising:

12. sending a request to the user (10) to watch a video; determining whether the user (10) has watched the video, and if the user has watched the video, reducing the Response Price (RP) by a specified amount and allocating the reduced price; and / or Allocating credits to said users (10). The method according to any one of claims 1 to 11, in particular claim 1, comprising:

13. The method according to any of claims 1 to 12, in particular claim 11, wherein said payment system (40) comprises a digital wallet for allocating said credits.

14. A computer readable medium (22) storing instructions that, when executed by at least one processor (21), cause said at least one processor (21) to implement the method of any one of claims 1 to 13.

15. A system for providing a response to a user query (Q), said system comprising an integrated control system (20), said integrated control system (20) comprising: receiving a user query (Q) from a user (10); generating a bid request (PQ) based on said user query (Q) and / or parameters attached to said user (10); sending said pricing request (PQ) to a large-scale language model (31) and requesting said large-scale language model (31) to generate a price indication, in particular a price (P); receiving the price indication generated by the large-scale language model (31); determining a response price (RP) based on the price indication; requesting confirmation from said user (10) for allocating said response price (RP); sending a response (A) related to said user query (Q) to said user (10); A system configured to:

Images