Method for increasing a user's knowledge and respective computer program
The method addresses the inefficiencies of existing knowledge enhancement systems by using AI to generate and manage content locally, considering user goals and interests, ensuring confidentiality and ownership, and providing personalized, efficient learning experiences.
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- VISIMPLEMENT DI ALBERTO CONSANI E C SAS
- Filing Date
- 2026-01-20
- Publication Date
- 2026-07-30
AI Technical Summary
Existing knowledge enhancement methods provide unsuitable, uninteresting, or irrelevant content to users due to their reactive nature, lack of consideration for user goals and interests, and centralized architectures that compromise data ownership, portability, and confidentiality.
A method utilizing an artificial intelligence model to generate and manage information content locally, considering user goals and interests, ensuring confidentiality and ownership, and employing vector-based profiling to select relevant content for personalized learning paths.
Enables efficient, secure, and personalized learning experiences by providing relevant content, ensuring user data confidentiality and control, and facilitating portable and verifiable learning profiles.
Smart Images
Figure IMGF000001_0001 
Figure IMGF000002_0001
Abstract
Description
[0001] METHOD FOR INCREASING A USER’S KNOWLEDGE AND RESPECTIVE COMPUTER PROGRAM
[0002] Technical Field
[0003] The present invention relates to a method for increasing a user’s knowledge and to a computer program adapted to implement such method.
[0004]
[0005] Teaching solutions are known in the prior art for transmitting content to a user in order to increase their knowledge in certain areas, comprising non-technical ones. Applications and software are particularly known adapted to teaching that involve profiling the user to check their level of knowledge. For example, these solutions ask the user a series of questions relating to one or more technical areas and, based on the answers, determine the user’s level of knowledge. Once the user profile has been generated, the information material is sent to the user so that they can learn new concepts and / or knowledge from this material, thus enriching their cultural background.
[0006] Known solutions particularly search for material among various publications, identifying a multitude of information content to be transmitted to the user. This content is searched for depending on the generated profile and, therefore, on the user’s prior knowledge.
[0007] However, since known solutions only take into account the user’s existing knowledge to search for information and identify various content, it is not uncommon for the content retrieved to be unsuitable to the user because it is unsatisfactory, uninteresting, or irrelevant.
[0008] In fact, providing the user with incorrect information makes their learning experience inefficient and makes them less likely to continue learning activities. Known solutions, such as the one described in document CN 116595 188 A (DI), for example, focus on analyzing prior knowledge to recommend content that fills learning gaps. Such systems do, however, operate reactive ly, do not actively shape the user’s future goals and aspirations, and are typically based on centralized architectures that do not offer the user ownership, portability, and verifiability of their learning data.The Applicant therefore realized that the solutions available today for increasing a user’s knowledge are susceptible to various improvements.
[0009] Object of the invention
[0010] The Applicant has sought to overcome these problems by proposing a new method that involves the creation of information content using an artificial intelligence model and whereby the provision of useful content to the user is considered not only on the basis of the user’s prior knowledge but also in relation to their goals and interests.
[0011] One object of the present invention is therefore to propose a method that allows each piece of information content proposed to the user to be associated with a score indicative of its final effect on the knowledge actually acquired by the user. Another object is to provide a method that allows for more versatile, faster, and more efficient learning and knowledge growth for a user.
[0012] A further object is to provide a method the processing of which is mainly carried out locally, thus ensuring the user’s absolute confidentiality (so-called “confidentiality by default”), ensuring that all personal and profiling data are generated, stored, and processed exclusively on the user’s personal device, without ever being transmitted to external systems, unless specifically authorized. Yet another object is to provide a method that gives the user full ownership and control of their profile and / or learning path, making their profile data secure, portable, and verifiable.
[0013] The main aim of the present invention is, therefore, to devise a method that allows the aforementioned drawbacks of the prior art to be overcome within a simple, rational, easy and effective to use, as well as affordable solution.
[0014] The aforementioned objects are achieved by the method according to claim 1. The aforementioned objects are achieved by the computer program according to claim 11.
[0015] Detailed
[0016]
[0017] This description relates to a method for increasing a user’s knowledge in accordance with the present invention.The method first comprises the phase of generating a repository of information content. Preferably, the phase of generating a repository of information content is performed in a local architecture, e.g. owned by the user, such as a smartphone, a tablet, or a computer. More preferably, the phase of generating a repository of information content is made and stored in a database of the local architecture. In detail, the repository is of the type of a vector repository. In particular, the phase of generating a repository of information content comprises the sub-phases of: - sending at least one raw content to an artificial intelligence model,
[0018] - transforming the raw content into user-usable information content by means of such artificial intelligence model, and
[0019] - collecting the transformed raw content into the repository.
[0020] According to one aspect, the artificial intelligence model is a language model. In particular, the language model is of the Large Language Model (LLM) type. The method may comprise the phase of selecting a linguistic model from a plurality of linguistic models. The linguistic model may be selected depending on respective operating parameters of the linguistic model itself. For example, the linguistic model may be selected based on a computational and / or economic cost associated with its use.
[0021] According to one aspect, the method may comprise the phase of modifying one or more contents of the repository and / or of removing one or more contents from the same repository. For example, the artificial intelligence model may modify and / or remove the contents deemed obsolete, incorrect, or otherwise no longer necessary.
[0022] In one of the possible embodiments, the phase of transforming said raw content comprises the sub-phase of summarizing the raw content by generating the information content adapted to usability for said user. In detail, the raw content is of one or more of the following types: audio, video, image, text. The information content usable by the user generated by the artificial intelligence model can be an image, a picture, a text, or a video summarizing the subject of the respective raw content. In particular, the artificial intelligence model is configured to generate, for each piece of information content, a respective characteristic vector (alsoknown as a “vector embedding”) that represents the semantic meaning thereof, which is stored in the repository together with the content itself.
[0023] In one aspect, the raw content may be a video of a lecture lasting between 30 minutes and 1 hour, which is transformed into a respective piece of information content usable by the user, of the type of a schematic diagram and / or of a text article. In other words, the information content may be a simplification of the raw content so that the user can more easily acquire its teaching. The raw content may be freely available on the web and traceable via the internet, e.g. by an artificial intelligence model or by an intelligent agent.
[0024] Furthermore, the method comprises the phase of profiling a user by generating at least a first vector and at least a second vector in a vectorial space. The first vector is representative of the user’s current and / or previous knowledge. The second vector is different from the first vector and is representative of the user’s desired and / or future knowledge. In other words, the first vector represents what the user knows and / or has previously learned, and the second vector represents what the user will learn or is interested in learning by using the information content. The set of information content selected by the method to guide the user from their current state of knowledge (represented by the first vector) to their desired state of knowledge (represented by the second vector) defines a “learning path” (also known as a “journey”), i.e., a structured sequence of actions aimed at achieving a specific goal within one or more areas of learning.
[0025] According to one aspect, the phase of profiling a user is at least partly performed by means of an algorithm of the type of an Apriori algorithm. For example, the phase of profiling the user may be performed by means of a Word2Vec model. In other words, the first vector and the second vector are generated by means of the Word2Vec model.
[0026] In one embodiment, the method can profile a multitude of different users. For example, the method can generate respective first and second vectors for each of such a multitude of users. In particular, the method involves generating vectors that are similar to each other if the respective users have similar knowledge and / or interests. Therefore, the first vector of a user may be similar or identical to thefirst vector of another user if the two users have similar levels of prior knowledge. Two users with a similar interest in increasing their knowledge may have second vectors that are equal or similar to each other.
[0027] In one of the embodiments of the invention, the method comprises the phase of clustering users. This phase can be performed based on the first vector and / or on the second vector. In detail, the clustering phase is performed by means of a clustering algorithm adapted to group similar users into one or more groups. According to one aspect, the method comprises the phase of creating a community of interest accessible to the users and / or of which users can be part. For example, users who are similar to each other can interact with each other in such a community of interest. Such a community of interest can be implemented at least partly by means of the clustering algorithm. The community of interest may be of the type of a chat room, a social network, and / or a forum, i.e., a web discussion platform where different users may talk via written messages on topics of interest to them.
[0028] The method also comprises the following phases of:
[0029] - retrieving information content from the repository depending on the first vector and on the second vector, and
[0030] - transmitting the retrieved content to the user.
[0031] Preferably, the phase of retrieving information content from the repository comprises the sub-phase of performing a semantic similarity calculation between the first vector and the second vector.
[0032] Preferably, the phase of retrieving information content from the repository is performed by means of the processor of the user’s device.
[0033] According to one aspect, the method comprises the phase of associating each content of the repository with a respective characteristic vector within the vectorial space. In particular, the method comprises the phase of associating each information content to the user with a respective score representative of its effect on the user’s knowledge. The score is associated with the respective characteristic vector of each content. In detail, the phase of associating each content in therepository with a respective characteristic vector is performed using the Word2Vec model.
[0034] The characteristic vector of each content is structurally similar to the first vector and to the second vector of the user. In detail, the characteristic vector of the content is representative of its subject and of its ability to increase the user’s knowledge.
[0035] In one embodiment, the phase of retrieving content from the repository comprises the sub-phases of:
[0036] - comparing the first vector and / or the second vector with one or more characteristic vectors of respective information content,
[0037] - identifying at least one useful piece of information content as a function of this comparison, wherein the useful piece of information content identified is adapted to increase the user’s knowledge.
[0038] The method involves identifying the information content in such a way that the respective characteristic vector can be added to the user’s first vector to give the user’s second vector as a result. In other words, the second vector is equal to the vectorial sum of the characteristic vector of the useful information content and the first vector.
[0039] In one embodiment, the phase of identifying at least one useful piece of information content comprises the sub-phase of performing a similarity calculation between the first vector and the characteristic vector of information content in the repository and / or between the second vector and such characteristic vector. This similarity calculation allows a semantic search to be performed based on the conceptual proximity between the user’s needs and objectives (represented by their vectors) and the content available in the repository (represented by their characteristic vectors).
[0040] According to one aspect, the phase of transmitting the retrieved information content to the user comprises the sub-phase of transmitting only one of the information contents to the user. The method involves sending a single piece of information content, rather than a multitude of contents, to the user, selectedspecifically for that user. In other words, the method reduces the likelihood that uninteresting and irrelevant content will be mistakenly provided to the user. The phase of transmitting the information content to the user comprises displaying such information content on a screen and / or a video terminal. The phase of transmitting the information content may be performed by means of a personal device of the user, such as e.g. a smartphone, a tablet, or a computer. The phase of transmitting the information content may also be performed by means of software and / or a web platform.
[0041] In one embodiment, the phase of comparing the first vector and the second vector with one or more characteristic vectors of respective information content of the repository comprises the sub-phase of performing a vectorial search in the content repository. Advantageously, the vectorial search allows for increased efficiency and speed of searching for information content while decreasing the possibility of error in selecting the correct content for the user.
[0042] According to one aspect, the repository comprises a first, a second, and a third group of content. The content of the repository can therefore be divided into three separate groups. Raw content transformed by means of the artificial intelligence is collected in the first group of content. It cannot be ruled out that two or more pieces of information content, e.g., content from the first group and content from the second group, may have a similar or identical teaching subject.
[0043] In one embodiment, the repository generating phase comprises the sub-phase of collecting a second group of information content created by means of at least one qualified operator. In detail, the method comprises creating information content by means of a qualified operator. In particular, such information content created by means of the operator is collected in the second group.
[0044] This qualified operator is identified in the technical field of the present invention by the acronym SME, i.e., Subject-Matter Expert, and / or as a knowledgeable person. The qualified operator may be, e.g., a university professor who is an expert in a particular technical field. In detail, the qualified operator has proven knowledge of the subject matter involved.Advantageously, the information content created by qualified operators is of high quality and reliability.
[0045] According to one aspect, the repository generating phase may comprise the subphase of collecting a third group of information content generated by one or more additional sources. In detail, the method comprises the phase of locating and / or retrieving information content present in the state of the art. Such located content is collected in the third group.
[0046] The information content of the third group can be found, e.g., on the web, online, in scientific publications, in specialized manuals or through similar means. The third group of content may also comprise content created by the user, e.g., during their learning and knowledge improvement process.
[0047] According to one aspect, each piece of information content collected in the repository is associated with one or more sources, e.g. containing the name of the author or authors of such content and / or an identification code for a related publication. Advantageously, citing the sources of the information content provides the user with the opportunity to further explore the topic.
[0048] According to one aspect, the method comprises the phase of identifying a plurality of learning areas. The learning areas may be associated with aspects of the user’s personal life. For example, the learning areas may be associated with one or more of personal finance, medicine, sports, or the like.
[0049] In one embodiment, there are eight learning areas, each associated with one of the following aspects: mindset, health, body, capabilities, goals, personal finance, emotions, and relationships Mind, Brain, Body, Capabilities, Mission, Wealth, Emotions, Relations).
[0050] The dimension of the vectorial space depends on the learning areas. Specifically, the first vector and the second vector represent the user’s current and desired knowledge, respectively, in at least one learning area. In other words, the user’s level of knowledge is assessed and measured in each of the learning areas. The first vector and the second vector can represent the user’s knowledge in all learning areas.In one embodiment with eight learning areas, the first vector and the second vector can be vectors of dimensions 1x8, if generated according to a single learning area, or of 8x8, if generated according to all learning areas. One of the goals of the method is therefore to help the user transform the knowledge acquired into concrete actions repeated over time, definable as “new habits”. A new habit is an observable and measurable behavior that the user adopts regularly following the completion of one or more learning paths, and its adoption and maintenance are a key indicator of the method’s success.
[0051] In detail, the phase of generating the first vector comprises the sub-phase of scoring the user based on their level of knowledge in at least one learning area. The phase of generating the second vector comprises the sub-phase of scoring the user based on the desired level of knowledge in at least one learning area.
[0052] According to one aspect, the phase of generating the first vector and / or the phase of generating the second vector comprises the phase of scoring the user for each learning area and for each relationship and combination between all learning areas.
[0053] In one embodiment, the first vector and / or the second vector has dimensions 1x64 or 64x64, where 64 is the number of possible combinations between all eight areas of learning. In detail, the first vector and the second vector are representative of the score assigned to the user for each learning area and for each possible combination thereof.
[0054] The method comprises the phase of checking learning or the learning path achieved / completed by the user of the transmitted information content. The learning checking phase comprises the sub-phases of subjecting the user to a test, receiving answers to that test, and assessing the answers received.
[0055] In a preferred embodiment, the phase of checking the user’s learning path may comprise the phase of calculating a proficiency index based on the observation of the user’s actions / habits and of measuring a plurality of key performance indicators (“KPIs”) of behavioral nature. These indicators may include, by way of example yet non limited to: the return on investment (“ROI”) perceived by the user in a learning area, the rate of retention of a new habit, the number ofsuccessfully completed learning paths, or the frequency with which the user applies the acquired knowledge in real life measured, e.g., using the data coming from the device’s sensors.
[0056] According to one aspect, the method comprises the phase of updating the first vector and generating a respective first vector updated based on one or more of: learning check, content characteristic vector, content-associated score. In detail, the first updated vector may be equal to the second vector if the user has fully learned the subject matter of the information content transmitted to them and / or the learning check is substantially positive. The first updated vector is structurally similar to the first vector and / or to the second vector and therefore will not be further described.
[0057] According to a further embodiment, the phase of updating the first vector comprises the sub-phase of acquiring, with the user’s explicit consent, data relating to their well-being from sources on their personal device. Such data may be acquired automatically from device sensors (e.g., accelerometer, gyroscope, etc.) and / or from system programming interfaces (“APIs”) such as HealthKit for iOS devices or Health Connect for Android devices. The acquired data, such as e.g. the number of steps, hours of sleep, heart rate, or physical activity data, is processed locally on the user’s device to dynamically enrich and update the first vector, thus providing an objective and contextualized representation of the user’s current status within the learning areas.
[0058] In one embodiment, the method comprises the phase of generating a token associated with the user. The phase of generating a token comprises the sub-phase of associating the token with one or more of: the first vector, the second vector, the first updated vector, the learning, the learning path. Preferably, the token comprises a cryptographic certificate locally stored on the user’s device.
[0059] In one embodiment, before generating the token, a phase of depersonalizing the user-related data is performed locally on the user’s device. This involves storing, even temporarily, the profiling data on the device itself to allow it to be processed in a secure environment controlled by the user, before any data is transmitted externally (so-called “secure enclave”). The token may be of the type of NFTtoken and / or smart contract. In detail, the method involves depersonalizing the user-related data and associating such depersonalized data with the token.
[0060] In one embodiment, the method comprises the phase of depersonalizing and aggregating data relating to a multitude of users. The token can be associated with the data relating to a multitude of users.
[0061] In one embodiment, the token may be sent and stored in one or more remote registries, such as a blockchain.
[0062] According to one aspect, the method comprises the phase of generating and displaying to the user a graph based on one or more of: the first vector, the second vector, the first updated vector. In detail, the graph is representative of the user’s profiling and / or of the user-related data.
[0063] The graph is of the radar graph type. The radar graph may have a number of axes or points equal to the number of learning areas. In one embodiment, the graph is of the knowledge graph type adapted to display information representative of the user’s level of knowledge in each learning area.
[0064] In one embodiment, the method comprises the phase of receiving an input from the user and of using such input to perform the phase of profiling the user. Specifically, the phase of generating the first vector and / or the phase of generating the second vector are performed based on said input. This input of the user may represent a user request. The input may represent the user’s interest in improving their skills in a particular learning area.
[0065] The phase of receiving an input from the user can therefore be performed before or during the phase of profiling the user so that the input can be used in the phase of profiling the user themselves.
[0066] According to one aspect, the method may comprise the phase of assessing the reliability and / or truthfulness of an information content of the first group in the repository. In detail, the phase of assessing the reliability and / or truthfulness may comprise the sub-phase of comparing a piece of information content from the first group with a piece of information content from the second and / or third group. For example, the information content of the first group may be compared with a pieceof information content from the second group belonging to the same learning area and created by a qualified operator.
[0067] One or more of the phases of the described method are performed by means of an artificial intelligence model and / or by means of an intelligent agent. Examples of intelligent agents that can be used in the present invention are known in the technical field as Embedding Agents, Dispatching Agents, RAG Agents, and Feedback Agents. For example, updating a user profile and / or retrieving information content can be performed by means of an intelligent agent.
[0068] According to one aspect, the method comprises the phase of preparing a plurality of intelligent agents, wherein each intelligent agent is configured to perform one or more respective phases of the method of the invention. In other words, different intelligent agents are configured to perform different phases of the method. In detail, the phase of receiving an input from the user can be performed by means of an intelligent agent adapted to transform a request in text, image, or audio format received from the user into a respective vectorial subject that can be used for user profiling and for generating the first and / or second vector.
[0069] According to one aspect, the phase of selecting the linguistic model from a plurality of linguistic models can be performed by a respective intelligent agent adapted to compare two or more linguistic models with each other.
[0070] In one embodiment, the method comprises the phase of recording one or more of: - a start date for learning and / or developing the user’s knowledge,
[0071] - a date of last update of the user’s profiling,
[0072] - a date of last transmission of a piece of information content,
[0073] - an economic value associated with the user,
[0074] - at least one number representing the user’s interactions with the information content, preferably associated with one or more of these learning areas.
[0075] The present invention also relates to a computer program comprising instructions which, when run by a computer, cause the computer to perform the phases of the method. The invention also relates to a computer-readable storage medium on which such a computer program is stored.In one preferred embodiment, the method provides that the data processing, the vectorial profile generation, and / or the semantic similarity calculation take place exclusively within the user’s device, without any transmission of personal data to the outside, to remote servers or distributed processing systems for aggregation purposes (so-called “local secure fusion” architecture). This architecture may comprise the use of a local database, e.g. based on SQL technology, residing in the device’s memory. Preferably, the database is configured to store both the characteristic vectors of the content and the vectors representing the user profile, preferably in the form of large-sized binary objects (“BLOBs”), thus ensuring confidentiality by design.
Claims
CLAIMS1) Method for increasing a user’s knowledge, comprising the phases of:- generating a repository of information content,- profiling a user by generating:- a first vector in a vectorial space, said first vector being representative of the current knowledge of said user,- a second vector in said vectorial space, said second vector being different from said first vector and representative of the desired knowledge of said user,- retrieving information content from said repository depending on the first vector and on the second vector,- transmitting said retrieved content to said user, andwherein the phase of generating a repository of information content comprises the sub-phases of:- sending at least a raw content to an artificial intelligence model,- transforming said raw content into user-usable information content by means of said artificial intelligence model,- collecting said raw content transformed into said repository.2) Method according to the preceding claim, comprising the phase of:- associating each content of said repository with a respective characteristic vector within said vectorial space, and / or- associating each information content to the user with a respective score representative of its effect on the user’s knowledge.3) Method according to the preceding claim, wherein the phase of retrieving content from said repository comprises the sub-phases of:- comparing the first vector and the second vector with one or more characteristic vectors of respective information content,- identifying at least one useful information content as a function of said comparison, said useful information content being adapted to increase the knowledge of said user.4) Method according to the preceding claim, wherein the phase of transmitting the retrieved information content to the user comprises the sub-phase of transmitting only one of said information content to the user, and / or the phase of comparing the first vector and the second vector with one or more characteristic vectors of respective information content comprises the sub-phase of performing a vectorial search in said content repository.5) Method according to any one of the preceding claims, wherein:- said raw transformed content is collected into a first group of content, and the phase of generating said repository comprises the sub-phase of collecting: - a second group of information content produced by means of at least one qualified operator, and- a third group of information content generated by one or more additional sources.6) Method according to any one of the preceding claims, wherein the phase of transforming said raw content comprises the sub-phase of summarizing the object of said raw content by generating the information content usable by the user.7) Method according to any one of the preceding claims, comprising the phase of identifying a plurality of learning areas, one dimension of said vectorial space being a function of said learning areas, wherein said first vector and said second vector are representative of the current and desired knowledge, respectively, of said user in at least one learning area.8) Method according to any one of the preceding claims, wherein:- generating the first vector comprises the sub-phase of:- scoring the user as a function of his / her level of knowledge in at least one learning area; and / or- generating the second vector comprises the sub-phase of:- scoring the user as a function of his / her desired level of knowledge in at least one learning area.9) Method according to any one of the preceding claims, comprising one or more of the phases of:- checking a learning by the user of said transmitted information content, - updating said first vector and generating a respective first vector updated as a function of one or more of: learning check, characteristic vector, content score, - generating a token on blockchain associated with the user, wherein the phase of generating a token comprises the sub-phase of associating said token with one or more of: the first vector, the second vector, the first updated vector,- generating and displaying to the user a graph as a function of one or more of: the first vector, the second vector, the first updated vector, said graph being representative of the user’s profiling and / or of the user-related data.10) Method according to any one of the preceding claims, wherein one or more of said phases are carried out by means of an artificial intelligence model and / or by means of an intelligent agent.11) Computer program comprising instructions which, when run by a computer, cause the computer to carry out the phases of the method according to any one of the preceding claims.