Methods and systems for data retrieval in decentral systems

Abstract

The invention relates to the field of sustainability, in particular to the field of sustainable production, end product re-use and end-of-life product recycling. The disclosure relates to methods, apparatuses and computer elements for retrieving data associated with output product(s) and / or input material(s) used to produce such output product(s) from a decentral network.

Description

[0001] METHODS AND SYSTEMS FOR DATA RETRIEVAL IN DECENTRAL SYSTEMS

[0002] TECHNICAL FIELD

[0003] The invention relates to the field of sustainability, in particular to the field of sustainable production, end product re-use and end-of-life product recycling. The disclosure relates to methods, apparatuses and computer elements for retrieving data associated with output product(s) and / or input material(s) used to produce such output product(s) from a decentral network.

[0004] TECHNICAL BACKGROUND

[0005] Decentral systems may allow to store data in decentral databases associated with multiple participants of the decentral system. The participants may include supply chain participants and participants of a re-use or recycling process of end-of-life products. Retrieval of specific data associated with output product(s) and / or production input(s) used to produce such output product(s) may be difficult due to defined semantics used within the decentral system to store data which may not be known by the users. Use of query data not matching the semantics used within the decentral system may result in not being able to locate and retrieve desired output product data and / or production input data. This may hamper the use of such data to improve production processes, reuse processes or recycling processes associated with such output product(s). Hence there is a need for improving retrieval of data associated with produced output product(s) and / or production input(s) from decentral systems.

[0006] SUMMARY OF THE INVENTION

[0007] Disclosed is in an aspect a method for retrieving data associated with output product(s) and / or input material(s) used to produce the output product(s) from a decentral network, the method comprising the steps of:

[0008] (a) providing search data related to the data associated with the output product(s) and / or the input material(s) to be retrieved from the decentral network;

[0009] (b) converting the provided search data into query data by a query data generator model trained to generate query data for retrieving output product data associated with the output product(s) and / or input material data associated with the input material(s) from the decentral network based on the provided search data;

[0010] (c) gathering decentral identifier(s) associated with the output product(s) from the decentral network based on the query data generated by the query data generator model;

[0011] (d) gathering output product data from the decentral network based on the gathered decentral identifier(s) associated with the output product(s) and / or gathering input material data from the decentral network based on the gathered decentral identifier(s) associated with the produced output product(s), in particular wherein access to the output product data stored in a database associated with the producer(s) of the output product(s) is controlled via the decentral identifier(s) by the producer(s) of the output product and / or access to the input material data stored in a database associated with the producer(s) of the input material(s) is controlled via the decentral identifier(s) by the producer(s) of the input material; (e) generating the data associated with the output product(s) and / or the input material(s) by transforming the gathered output product data and / or the input material data by a data transformer model trained to transform the gathered output product data and / or input material data based on the provided search data;

[0012] (f) providing the generated data associated with the output product(s) and / or the input material(s).

[0013] In another aspect disclosed is an apparatus for retrieving data associated with output product(s) and / or input material(s) used to produce the output product(s) from a decentral network, the apparatus comprising:

[0014] (a) a data providing interface configured to provide search data related to the data associated with the output product(s) and / or the input material(s) to be retrieved from the decentral network;

[0015] (b) a query data generator model configured to convert the provided input data into query data, wherein the query data generator model is trained to generate query data for retrieving output product data associated with the output product(s) and / or input material data associated with the input material(s) from the decentral network based on the provided search data;

[0016] (c) a decentral network interface configured to

[0017] • gather decentral identifier(s) associated with the output product(s) from the decentral network based on the query data generated by the query data generator model and

[0018] • gather output product data from the decentral network based on the gathered decentral identifier(s) associated with the produced output product(s) and / or to gather input material data from the decentral network based on the gathered decentral identifier(s) associated with the produced output product(s), in particular wherein access to the output product data stored in a database associated with the producer(s) of the output product(s) is controlled via the decentral identifier(s) by the producer(s) of the output product and / or access to the input material data stored in a database associated with the producer(s) of the input material(s) is controlled via the decentral identifier(s) by the producer(s) of the input material;

[0019] (d) a data transformer model configured to generate the data associated with the output product(s) and / or the input material(s), wherein the data transformer model is trained to transform the gathered output product data and / or input material data based on the provided search data;

[0020] (e) a data providing interface configured to provide the generated data associated with the output product(s) and / or the input material(s).

[0021] In yet another aspect disclosed is a method for retrieving data associated with output product(s) and / or input material(s) used to produce the output product(s) from a decentral network, the method comprising the steps of:

[0022] (a) providing search data related to the data associated with the output product(s) and / or the input material(s) to be retrieved from the decentral network; (b) generating embedding(s) from the provided search data by at least one embedding model trained to generate embedding(s) from received input data,

[0023] (c) gathering embedding(s) associated with access element(s) matching the generated embedding(s) from a database storing embedding(s) associated with access element(s) associated with output product(s) and / or input material(s)

[0024] (d) converting the provided input data and gathered embedding(s) into query data by a query data generator model trained to generate query data for retrieving output product data associated with the output product(s) and / or input material data associated with the input material(s) from the decentral network based on the provided search data;

[0025] (e) gathering decentral identifier(s) associated with the output product(s) from the decentral network based on the query data generated by the query data generator model;

[0026] (f) gathering output product data from the decentral network based on the gathered decentral identifier(s) associated with the output product(s) and / or gathering input material data from the decentral network based on the gathered decentral identifier(s) associated with the produced output product(s), in particular wherein access to the output product data stored in a database associated with the producer(s) of the output product(s) is controlled via the decentral identifier(s) by the producer(s) of the output product and / or access to the input material data stored in a database associated with the producer(s) of the input material(s) is controlled via the decentral identifier(s) by the producer(s) of the input material;

[0027] (g) generating the data associated with the output product(s) and / or the input material(s) by transforming the gathered output product data and / or the input material data by a data transformer model trained to transform the gathered output product data and / or input material data based on the provided input data;

[0028] (h) providing the generated data associated with the produced output product(s) and / or the input material(s).

[0029] In yet another aspect disclosed is an apparatus for retrieving data associated with output product(s) and / or input material(s) used to produce the output product(s) from a decentral network, the apparatus comprising:

[0030] (a) a data provider configured to provide search data related to the data associated with the output product(s) and / or the input material(s) to be retrieved from the decentral network;

[0031] (b) an embedding model configured to generate embedding(s) from the provided search data, wherein the embedding model is trained to generate embedding(s) from received input data,

[0032] (c) an embedding provider configured to gather embedding(s) associated with access element(s) matching the embedding(s) generated by the embedding model from a database storing embedding(s) associated with access element(s) associated with output product(s) and / or input material(s)

[0033] (d) a query data generator model configured to convert the provided input data and gathered embedding(s) into query data , wherein the query data generator model is trained to generate query data for retrieving output product data associated with the output product(s) and / or input material data associated with the input material(s) from the decentral network based on the provided search data;

[0034] (e) a decentral network interface configured to

[0035] • gather decentral identifier(s) associated with the produced output product(s) from the decentral network based on the query data generated by the query data generator model and

[0036] • gather output product data from the decentral network based on the gathered decentral identifier(s) associated with the produced output product(s) and / or gathering input material data based on the gathered decentral identifier(s) associated with the produced output product(s), in particular wherein access to the output product data stored in a database associated with the producer(s) of the output product(s) is controlled via the decentral identifier(s) by the producer(s) of the output product and / or access to the input material data stored in a database associated with the producer(s) of the input material(s) is controlled via the decentral identifier(s) by the producer(s) of the input material;

[0037] (f) a data transformer model configured to generate the data associated with the output product(s) and / or the input material(s), wherein the data transformer model is trained to transform the gathered output product data and / or input material data based on the provided search data;

[0038] (g) a data providing interface configured to provide the generated data associated with the output product(s) and / or the input material(s).

[0039] In yet another aspect disclosed is the use of the data associated with produced output product(s) and / or input material(s) used to produce the output product(s) retrieved from the decentral network according to the methods disclosed herein or the apparatuses disclosed herein for controlling and / or managing the production of further output product(s) and / or for controlling and / or managing the reuse and / or recycling of end-of-life product(s) and / or for monitoring the environmental impact of the input material(s), the output product(s) or production(s) producing the input material(s) or output product(s).

[0040] In yet another aspect disclosed is a computer element, in particular a computer program product or a computer readable medium, with instructions, which when executed on one or more computing node(s) are configured to carry out the steps of any of the methods disclosed herein.

[0041] In yet another aspect the present disclosure relates to a computer element with instructions, which when executed on one or more computing node(s) is configured to carry out the steps of the method(s) of the present disclosure or configured to be carried out by the apparatus(es) of the present disclosure.

[0042] Any disclosure, embodiments and examples described herein relate to the methods, the apparatuses, the uses and computer elements lined out above and below. Advantageously, the benefits provided by any of the embodiments and examples equally apply to all other embodiments and examples. Embodiments

[0043] In the following, embodiments of the present disclosure will be outlined by ways of embodiments and / or examples. It is to be understood that the present disclosure is not limited to said embodiments and / or examples.

[0044] To enable or improve re-use or recycling of used end products or components thereof, to improve production of output products and / or to enable monitoring of the environmental impact of the input material(s), the output product(s) or production(s) producing the input material(s) or output product(s), the retrieval of data of such end products, components or output products is crucial. For instance, the retrieval of such data may allow to monitor substances included in end products, components or output products, which are critical for recycling and / or re-use of such end products, components or output products. In addition, retrieval of such data may allow to determine the amount or number of recyclable or re-usable end product(s). Retrieval of such data may further allow to determine the amount of recycled material to be obtained upon performing a recycling process on such determined end products or components thereof. Hence, the amount of recycled material available in the future may be determined using such retrieved data. This may allow to control and / or manage recycling processes based on the retrieved data. Retrieval of such data may further allow to monitor the environmental impact of the input material(s), the end-products, components thereof, or the output product(s) and / or the environmental impact of production(s) producing such input material(s), endproducts, components thereof or output products. This may allow to provide a transparency on the environmental impact of the product ecosystem, allowing to steer the environmental impact based on such transparency. While such data may be available within the decentral network, use of highly defined query data to located and retrieve end product data, component data or output product data within the decentral network may hamper reliable retrieval of such data by entities involved in production, re-use and / or recycling processes.

[0045] By transforming unstructured data associated with a context, e.g. user input provided in natural language, into structured query data matching the data format used within the decentral network for locating data associated with output product(s) (such as end products, components, parts, component assemblies, chemical products, chemical intermediate products) or production input(s) thereof (such as raw materials), output product data and / or input material data matching the context included in the unstructured data may be reliably gathered under full data sovereignty of the data owners of such data (e.g. the output product producers or input material producers) without the user requiring knowledge on output product identifiers of input material identifiers allowing discovery of the decentral identifier(s) or the decentral identifier(s). This way, output product data and / or input material data may be reliably provided without violating the data sovereignty of the data owners based on context included in the user input, enabling more accurate monitoring and / or controlling of the production of further product(s), the recycling and / or re-use of the output product(s) and / or the monitoring of the environmental impact of the input materials, the output product and / or the production based on the generated response may be achieved. The users may not be required to know the data structure and keywords required to determine the decentral identifiers which are a prerequisite for gathering the output product data and / or the input material data. Instead, users may formulate their query in natural language which may then be transformed by a trained data-driven model, such as a neural network, into structured query data related to the user query, e.g. the output product(s) and / or input material(s) contained in the user query. The query data may include instruction(s) for querying the decentral network, in particular for determining decentral identifier(s) associated with output product data and / or input material data requested by the user, hence allowing to reliably retrieve the data requested by the user based on a single user input. By transforming unstructured data associated with the context into structured query data related to such context, data on output product(s) and / or input material(s) desired by the user may be reliably located and gathered via the decentral network under full control of the data owners of the respective data, hence avoiding that unauthorized users get access to such data via the computing node(s) performing the methods described herein. A trained data driven model, such as a neural network, which is parametrized on unstructured data and associated query data, may be used to translate or convert a user query containing unstructured data associated with the context into related structured query data allowing to query the decentral network for output product(s) and / or input material(s) contained within the unstructured data or natural language query of the user. This ensures, for example, that data on substances critical for re-use or recycling can be reliably identified by entities performing such re-use or recycling operation(s), hence allowing to use such data to improve re-use or recycling operations. For instance, re-use or recycling rates, quality of recyclate or the end product resulting from the re-use operation may be improved using such data on substances critical for re-use or recycling. This further ensures that amounts of recycled material available in the future may be reliably determined using such retrieved data. This may allow to control and / or manage recycling processes based on the retrieved data.

[0046] By controlling access to the output product data and / or input material data associated with the determined decentral identifier(s) by the respective data owners, the output product data and / or input material data may be provided under full control of such data owners. This way, sharing of output product data and / or input material data may be enabled based on a user input under full sovereignty of the data owners. This avoids that users can query the decentral network for output product data and / or input material data they are not authorized to access, avoiding leakage of output product data and / or input material data to unauthorized users. The decentral network may hence be queried via intermediary decentral network nodes in a reliable yet secure manner, ensuring that authorized users can access the requested data while avoiding access to the data by unauthorized users via the intermediary decentral network nodes. The intermediary decentral network nodes may be associated with an entity not being part of the production and / or use of the input material and / or output product. The intermediary decentral network nodes may be associated with an entity not being part of the production and / or use of the input material and / or output product

[0047] By transforming gathered output product data and / or input material data (e.g. output product data and / or input material data gathered based on the generated query data) based on unstructured data associated with a context (e.g. query data received from the user), the provided response to the user may be tuned to the data requested by the user. This may allow to transform the gathered data into the data requested by the user, for example by selecting data requested by the user from the gathered data. Transforming the gathered data ensures that the data returned to the user is matching the request of the user, hence avoiding the provision of superfluous data and avoiding the transfer and display of unnecessary data. In addition, this allows to ensure that production, re-use and / or recycling processes can be reliably controlled and / or monitored by avoiding the use of unrelated data (e.g. data not matching the user query) upon determination of control and / or monitoring data for such processes based on the retrieved data. This may also allow reliable monitoring of the environmental impact of the input material(s), the output product(s) and / or the production producing the input material(s) or the output product(s).

[0048] Various units, entities, nodes or other computing components may be described as “configured to” perform a task or tasks. Configured to shall recite structure meaning “having circuitry that” performs the task or tasks on operation. The units, circuits, entities, nodes or other computing components can be configured to perform the task even when the unit / circuit / component is not operating. The units, circuits, entities, nodes or other computing components that form the structure corresponding to “configured to” may include hardware circuits and / or memory storing program instructions executable to implement the operation. The units, circuits, entities, nodes or other computing components may be described as performing a task or tasks, for convenience in the description. Such descriptions shall be interpreted as including the phrase “configured to.”

[0049] In general, the methods, apparatuses, systems, computer elements, nodes or other computing components described herein may include memory, software components and hardware components. The memory can include volatile memory such as static or dynamic random-access memory and / or nonvolatile memory such as optical or magnetic disk storage, flash memory, programmable read-only memories, etc. The hardware components may include any combination of combinatoric logic circuitry, clocked storage devices such as flops, registers, latches, etc., finite state machines, memory such as static random-access memory or embedded dynamic random-access memory, custom designed circuitry, programmable logic arrays, etc.

[0050] Output product may refer to any product produced by a production plant. The production plant may be operated by a participant of a product ecosystem. The output product may be a chemical intermediate product, a chemical product, a component or part, a part assembly or an end product. Output product may refer to any product resulting from the use of produced output products, such as end-of-life products.

[0051] The input material used to produce the output product may refer to any good which is bought from suppliers and brought to a respective production plant producing the output product. The input material may include starting material used in the production process of the production plant producing the output product. The input material may correspond to an output product produced by an upstream participant of the product ecosystem). Hence, the output product of the one production plant may correspond to input material of a production plant operated by a downstream participant. The input material may comprise or be any input material entering a production producing output product(s). The input material may comprise or be any input material provided at any entry point of the production. The input material may be a raw material. The input material may be a virgin material (e.g. material not resulting from a recycling process). The input material may be recycled material. The input material may be a chemical intermediate product. The input material may be a chemical product. The input material may be a component or part. The input material may be a part assembly.

[0052] The output product(s) may be associated with output product data. The output product data may represent a digital twin (e.g. a digital representation) of the physical entity of the output product. The output product data may represent a part of a digital twin of the physical entity of the output product. The output product data may represent one or more output product data set(s). Output product data may be stored on a dedicated storage associated with or owned by or related to the data owner of the output product data. The output product data may be collected prior to, upon and / or after production of the output product. The output product data may include production data associated with the production of the output product, output product composition data, output product property data, environmental attributes associated with the output product, output product material safety data or a combination thereof. The output product data may further include one or more decentral identifier(s). The output product data may include one or more output product data set(s). An output product data set may include a part of the output product data and one or more decentral identifier(s). The one or more decentral identifier(s) may include a decentral digital twin identifier and a decentral data set identifier. This may allow to uniquely identify a given output product data set within the output product data. The output product data may include one or more authentication mechanisms associated with the decentral identifier(s) and the data contained therein. The output product data may relate to one or more authorization mechanisms associated with the decentral identifier(s) and the data contained therein. The one or more authorization mechanisms may include authorization rules determining if access to output product data or a part thereof is granted.

[0053] The composition data may relate to the material composition of the output product. The composition data may specify the material composition of the output product. For example, the composition data may specify at least part of the chemical composition of the output product. The environmental attribute data may include emission data, recyclate content, biobased content and / or renewable content data and / or biodegradability data.

[0054] Property data of the output product may include a measured chemical and / or physical property. Property data of the output product may be determined from collected data and / or the use of the produced output product. The collected data may be used to determine at least one physical and / or chemical property of the produced output product. For instance, the at least one physical and / or chemical property may be determined from sensor data. Data associated with the use of the output product may be collected via at least one identifier associated with the output product. The data may be collected during and / or after use of the output product. Collected data may include at least one measured physical and / or chemical property of the used output product. The measured physical and / or chemical property may include the chemical and / or physical properties. The data may be collected with a suitable sensor configured to measure the chemical and / or physical property. The chemical property may be a property of the output product that becomes evident during, or after, a chemical reaction. Hence, the chemical property may be any quality that can be established only by changing the chemical identity of the output product. Examples of chemical properties include heat of combustion, enthalpy of formation, toxicity, chemical stability in a given environment, flammability, oxidation state(s), ability to corrode, combustibility, acidity and basicity, chemical composition, recyclate content used for producing or manufacturing the output product, bio-based content used for producing or manufacturing the output product, renewable content used for producing or manufacturing the output product, emission data, such as carbon footprint data, and / or pH value. The physical property may be any property of the output product that is measurable. Hence, the value of a physical property describes a state of the output product. Examples of physical properties include absorption, brittleness, boiling point, capacitance, color, concentration, density, ductility, distribution, efficacy, elasticity, electric charge, electrical conductivity, electrical impedance, electric potential, flow rate, fluidity, hardness, heat capacity, inductance, intrinsic impedance, luminance, luminescence, luster, mass, melting point, opacity, permeability, permittivity, plasticity, pressure, radiance, resistivity, reflectivity, refractive index, solubility, specific heat, strength, stiffness, temperature, tension, thermal conductivity, thermal resistance, viscosity, volume and / or wave impedance. The measured at least one physical and / or chemical property may be obtained by sensors configured to measure such property. The sensor may be included in a measuring device. The sensor may correspond to the measuring device.

[0055] The input material(s) may be associated with input material data. The input material data may represent a digital twin (e.g. a digital representation) of the physical entity of the input material. The input material data may represent a part of a digital twin of the physical entity of the input material. The input material data may represent one or more input material data set(s). Input material data may be stored on a dedicated storage associated with or owned by or related to the data owner of the input material data. The input material data may be collected prior to, upon and / or after production of the input material. The input material data may include data related to a property of the input material, data related to the use of the input material and / or production data. Such data may include the data previously described in the context of the output product. The input material data may further include one or more decentral identifier(s). The input material data may include one or more input material data set(s). An input material data set may include a part of the input material data and one or more decentral identifier(s). The one or more decentral identifier(s) may include a decentral digital twin identifier and a decentral data set identifier. This may allow to uniquely identify a given input material data set within the input material data. The input material data may include one or more authentication mechanisms associated with the decentral identifier(s) and the data contained therein. The input material data may relate to one or more authorization mechanisms associated with the decentral identifier(s) and the data contained therein. The one or more authorization mechanisms may include authorization rules determining if access to input material data or a part thereof is granted.

[0056] The output product data and / or the input material data may have a defined semantic structure. The defined semantic structure may be obtained by applying a semantic model to collected data. The semantic model may contain a semantic description including a data structure and / or properties, such as data types, possible or allowable ranges and / or values and / or physical unit(s) of parameters described by values. Use of a semantic model ensures that the output product data and input material data contain data having a defined semantic structure, allowing to exchange such data in a standardized and harmonized way.

[0057] The output product and the input material(s) may be associated with one or more access element(s). The access element may be a data set configured to provide access to the output product data or input material data, respectively. The access element may be a digital representation of the output product data or the input material data, respectively. The access element may include one or more decentral identifier(s) and access data. The decentral identifier(s) may include decentral identifier(s) associated with the output product data, such as the decentral digital twin identifier of the output product and decentral data set identifier(s). The decentral identifier(s) may include decentral identifier(s) associated with the input material data, such as the decentral digital twin identifier of the input material and decentral data set identifier(s). The access data may include a representation for accessing the output product data or input material data, respectively. The access data may include a locator or pointer, such as am url or uri, to a dedicated storage, such as a dedicated storage address, associated with the data owner of the output product data or input material data, respectively. The pointer or locator may point directly to the dedicated storage. The pointer or locator may point to a data providing network node associated with the dedicated storage. The access element may include one or more authentication mechanisms associated with the decentral identifier(s) and the access data. The access element may be associated with one or more authentication mechanisms associated with the decentral identifier(s) and the access data. The access element may be provided to a decentral registry storing access elements. The decentral registry may be associated with a data providing network node. This may allow to control access to such registry and access to access element(s) stored in such registry via the data providing network node. Access to the decentral registry may be controlled by the data owner of the output product data or the input material data the access elements stored in such registry are associated with. The decentral registry may be associated with a participant of the product ecosystem. The decentral registry may be associated with the data owner of the output product data or the input material data, respectively. The decentral registry may be part of the decentral network but may not be associated with a particular participant of the production chain, e.g. may be regarded as infrastructure node of the decentral network.

[0058] The output product(s) and input material(s) may be part of a product ecosystem. The product ecosystem may include chemical products. The product ecosystem may include production chains to produce output product(s). The product ecosystem may include processing chains to process end-of- life products. Processing chains may include recycling chains to recycle at least part of end-of-life products or component(s) thereof. Processing chains may include re-use chains to re-use end-of-life product(s). The product ecosystem may include various participants, such as raw input material producers, chemical product producers, chemical product users, end-product producers, end-product users, EOL product collectors and recyclers. The product ecosystem may allow to use recycled materials resulting from recycling of end-of-life end products to produce new products, such as chemical products. The product ecosystem may be associated with the production and / or re-use and / or recycling of physical products.

[0059] The participants of the product ecosystem may be connected via the decentral network. The decentral network may include one or more decentral network node(s) configured to perform data transactions. The decentral network node(s) may be associated with participants of the product ecosystem. The data transactions may be based on a transaction protocol including authentication and / or authorization mechanism(s). Based on the authentication and / or authorization mechanism(s) a peer- to-peer communication channel between decentral network node(s) of the decentral network may be established. The one or more authentication mechanism(s) may be associated with or linked to the decentral identifier(s). The one or more authentication mechanism(s) associated with the decentral identifier(s) may be provided to decentral network node(s). The one or more authentication mechanism(s) associated with the decentral identifier(s) may be accessible by decentral network node(s). The decentral configuration allows for more efficient use of computing resources and strengthens control with respect to data access by each data owner of the decentral network.

[0060] Data providing network node(s) may be configured to provide access to data stored in a dedicated storage associated with the respective data providing network node(s). The data stored in the dedicated storage may include the output product data or input material data, respectively. The data providing network node(s) may be configured to provide access to such data upon request by data consuming network node(s). The access to such data may be under control of the data providing network node associated with the respective data. The data providing network node may be configured to authenticate the data consuming network node(s) and / or to authorize access to such data by the data consuming network node(s).

[0061] Data consuming network node(s) may be configured to request access to data stored in a dedicated storage associated with the data providing network node(s). Data consuming network node(s) may be configured to determine access elements associated with output products and / or input materials, for example by querying the decentral network, in particular decentral registries storing access elements using query data. The data consuming network node may be associated with a participant of the product ecosystem.

[0062] The decentral identifier may comprise any unique identifier(s) uniquely associated with the producer of the output product, and the respective output product. The decentral identifier may comprise any unique identifier(s) uniquely associated with the producer of the input material, and the respective input material. The decentral identifier may include one or more Universally Unique Identifier(s) (UUID) or one or more Digital Identifier(s) (DID(s)). The decentral identifier may be issued by a central or decentral identity issuer. The decentral identifier may include authentication information. Via the decentral identifier and its unique association with the with the producer of the output product and the output product, access to the output product data may be controlled by the producer of the output product. Via the decentral identifier and its unique association with the with the producer of the input material and the input material, access to the input material data may be controlled by the producer of the input material. This contrasts with central authority schemes, where identifiers are provided by such central authority and access to data is controlled by such central authority. Decentral in this context refers to the usage of the identifier as controlled by the data owner. The decentral identifier may be discoverable for decentral participant nodes registered within the decentral network. Discovery of the decentral identifier may be controlled by the data owner of data associated with the decentral identifier. For instance, discovery of the decentral identifier(s) associated with the output product data may be controlled by data owner(s) of the output product data, e.g. output product producers. Likewise, discovery of the decentral identifier(s) associated with the input material data may be controlled by the data owner(s) of the input material data, e.g. the input material producers.

[0063] The data owner may be an entity having access to the output product data and controlling access by data consuming services of the decentral network to the output product data or parts thereof. The data owner may be an entity having access to the input material data and controlling access by data consuming services of the decentral network to the input material data or parts thereof. The data owner may be the output product producer or the input material producer. Via the decentral identifier(s) and its / their unique association with the data owner and output product data or input material data access to the output product data or input material data may be controlled by the data owner. The output product data may be accessible for the data owner. The data owner may hence directly or indirectly own the output product data. The input material data may be accessible for the data owner. The data owner may hence directly or indirectly own the input material data. The output product data or input material data may be stored in a data base of or associated with the data owner. The output product data or input material data may be stored in a data base accessible by or under control of the data owner. The data owner may control access to the chemical product passport via the data providing service of the data owner. The data owner may control access to the output product data or input material data, respectively. The output product data or input material data may be associated with the data owner. The data owner may be the owner of the output product data or the output product data owner. The data owner may be the owner of the input material data or the input material data owner.

[0064] Retrieving data associated with output product(s) and / or input material(s) used to produce the output product(s) from a decentral network may include locating data provider(s) associated with respective output product data and / or input material data associated with such output product(s) and / or input material(s) and accessing such data from the located data provider(s). The output product data may be related to the retrieved data. The output product may include the retrieved data. The input material data may be related to the retrieved data. The input material data may include the retrieved data.

[0065] The query data generator model may include at least one data-driven model. The data driven model may include at least one network architecture with at least one input layer, one or more hidden layers and at least one output layer. The query data generator model may be based on a sequential and / or parallel neural network. One or more network architecture(s) may be connected to be performed sequentially and / or parallel. At least one layer of the query data generator model may include an input layer or channel, one or more hidden layers or channels and an output layer or channel. The query data generator model may include parameters, such as kernels, weights, biases, functional relationships, operators, constraints or the like, which are trained based on training data set(s). The query data generator model configured to generate query data may receive search data, such as user input, at the input layer and generate the query data.

[0066] The data transformer model may include at least one data-driven model. The data driven model may include at least one neural network architecture with at least one input layer, one or more hidden layers and at least one output layer. The data transformer model may be based on a sequential and / or parallel neural network. One or more neural network architecture(s) may be connected to be performed sequentially and / or parallel. At least one layer of the data transformer model may include an input layer or channel, one or more hidden layers or channels and an output layer or channel. The data transformer model may include parameters, such as kernels, weights, biases, functional relationships, operators, constraints or the like, which are trained based on training data set(s). The data transformer model may be connected to one or more databases storing output product data and / or input material data, for example via APIs. The data transformer model may be trained to generate query data to query the one or more databases for output product data or parts thereof and / or input material data or parts thereof matching the search data received from the user. The query data may be generated by sequentially generating querying the database(s) to determine the data structure and content of the database(s). The data transformer model may further be trained to provide the database response from the generated query data in a given data scheme. The data transformer model may include one or more algorithms or routines. The one or more algorithms and / or routines may be connected to the neural network architecture. The one or more algorithms and / or routines may allow to determine whether output generated by the neural network architecture fulfils given criteria. This may allow to determine when the query data matching the search data received from the user has been determined and an output may be generated. The data transformer model configured to transform output product data and / or input material data may receive input data including the search data at the input layer and may generate the transformed data. Data transformation may include selecting a part of the gathered output data and / or input material data, converting gathered output product data and / or input material data into a different data scheme and / or performing at least one arithmetic operation on the respective data. Arithmetic operations may include addition, subtraction, multiplication and division. For instance, arithmetic operation(s) may be used to convert gathered or selected output product data / input material data into a unit contained in the search data received from the user.

[0067] In an embodiment, the data associated with the output product(s) and / or input material(s) may be related to or associated with the gathered output product data and / or input material data. The data associated with the output product(s) and / or input material(s) may correspond to transformed output product data and / or input material data.

[0068] In an embodiment, search data includes unstructured data associated with a context. The search data may be provided by a user. The search data may correspond to a prompt provided by the user. The unstructured data may include unstructured text data. The unstructured text data may relate to the data associated with output product(s) and / or input material(s) desired by the user. The unstructured data may include output product type(s) and / or input material type(s), property / properties of the output product type(s) and / or input material type(s), manufacturer type(s), manufacturer name(s), output product name(s), input material name(s), output product identifier(s), input material identifier(s) or a combination thereof. The unstructured text data may represent one or more sentences in a natural language. The context may be associated with the output product(s) and / or input material(s). The context may be associated with the data associated with the output product(s) and / or input material(s). The context may be associated with instruction(s) related to the output product(s) and / or input material(s). The context may be associated with instruction(s) related to data associated with the output product(s) and / or input material(s). The instruction(s) may include output product type(s) and / or input material type(s) and / or properties of output product type(s) and / or input material type(s).

[0069] In an embodiment, the query data may include structured data. The structured data may be related to structured data included in access element(s) associated with the output product(s) and / orthe input material(s). The structured data may include value(s), key(s) and / or key-value pair(s) included in the access element(s). The value(s), key(s) and / or key-value pair(s) may relate to the output product(s) and / or the input material(s). The value(s), key(s) and / or key-value pair(s) may be included in access element(s) associated with the output product(s) and / or the input material(s).

[0070] In an embodiment, the query data generator includes at least one data-driven model parametrized according to training data set(s) including unstructured data and associated query data. The associated query data may include metadata associated with such query data. The metadata may include a feature vector generated by an encoder based on unstructured input data associated with the given context. The metadata may include nearest feature vector(s) obtained from a vector database based on embeddings of the input data. The metadata may include one or more label(s). The labels may signify aspect(s) associated with or related to output product(s) and / or input material(s). The aspect(s) may signify or relate to different types previously described. The labels may signify aspect(s) corresponding to or being associated with output product(s) or input material(s). The labels may signify aspect(s) corresponding to or being associated with output product data or input material data. In addition, the label(s) may signify query data associated with or related to the respective aspect. For instance, the label(s) may signify aspect(s) and associated query data.

[0071] In an embodiment, the query data generator model trained to generate the query data generates at least one classifier discriminating between one or more output product types, one or more manufacturer types, one or more input material types, one or more output product type properties and / or one or more input material type properties. Generating the query data may include generating at least one feature vector (or representation) from search data and matching the feature vector (or representation) with one or more reference template vector(s) for different types, such as output product types, input material types, manufacturer types, output product property types, input material property types. The at least one feature vector (or representation) may be generated by providing search data, such as unstructured data associated with a context, to a neural network trained on training dataset(s) of unstructured data associated with a defined context and associated query data, generating an output and transforming the output to the feature vector(s). The feature vector (or representation) may include a representation of the aspect(s) included in the search data, such as the unstructured data associated with a context.

[0072] In an embodiment, the query data generator model trained to generate the query data is trained to extract sequences associated with the output product(s), the input material(s), properties of output product(s) and / or properties of input material(s) and to generate the at least one classifier based on the extracted sequences. The extracted sequences may correspond to aspect(s) previously described.

[0073] In an embodiment, the query data relates to the data associated with the output product(s) and / or the input material(s). The query data may relate to output product data and / or input material data which may in turn relate to the data associated with the output product(s) and / or input materials. For instance, the query data may include key-value pair(s) included in access elements associated with output product data or a part thereof and / or input material data or a part thereof. The key-value pair(s) may hence be used to gather such output product data and / or input material data or a part thereof associated with such access element(s).

[0074] In an embodiment, the data associated with the output product(s) is related to access element(s) configured to provide access to the output product data, wherein the query data includes key-value pair(s) contained within such access element(s) and being related to the data associated with the output product(s). The key-value pair(s) contained within such access element(s) may be associated with the output product data associated with such output product(s). The key-value pair(s) may allow to determine access data pointing to one or more part(s) of the output product data.

[0075] In an embodiment, the data associated with the input material(s) is related to access element(s) configured to provide access to the input material data, wherein the query data includes key value pair(s) contained within such access element(s) and being related to the data associated with the input material(s).

[0076] In an embodiment, the output product data is gathered based on the decentral identifier(s) associated with the produced output product(s) and the query data and / or wherein the input material data is gathered based on decentral identifier(s) associated with the produced output product(s) and at least a part of the query data. The decentral identifier(s) may be determined by providing at least part of the query data to data providing node(s) associated with decentral registries storing access element(s). The at least a part of the query data may be provided by a decentral data consuming network node to the data providing node(s). The decentral identifier(s) may be determined by determining a data providing node(s) and providing at least a part of the query data to such data providing node(s). The data providing node(s) may be configured to query the associated decentral registry / registries based on the query data. The data providing node(s) may be configured to determine matching access element(s) by matching the query data, such as key-value pairs included in the query data, to access element data including key-value pairs and stored in the decentral registry / registries. Upon determining matching data, data providing node(s) may be configured to provide decentral identifier(s) included in access element(s) containing matching query data to the data consuming node querying the decentral network. The data providing node(s) may be determined based on the query data, such as key-value pair(s) included in the query data. The query data to be provided to the data providing node(s) may be determined based relationships signified by data included in the query data. For instance, a part of the query data may signify the output product(s) and a further part may signify input material(s). The part of the query data signifying the output product may be used to gather decentral identifier(s) associated with the output product(s) from the decentral network. Query data signifying input material(s) may trigger gathering of relationship data as described in the following.

[0077] In an embodiment, the input material data is gathered based on relationship data set(s) signifying relationship(s) between the output product(s) and the input material(s) used to produce the output product(s). The relationship data set may include the one or more decentral identifier(s) associated with the output product and one or more decentral identifier(s) associated with the input material(s). The relationship between the output product and the input material(s) may be specified by denoting the decentral identifier associated with the respective output product as parent decentral identifier and the decentral identifier(s) associated with the input materials as child decentral identifier(s). The relationship data set(s) may be accessed via access data contained in an access element associated with the respective output product. By recursive determination of child relationship identifier(s) of respective parent identifier(s), the bill of material tree of an output product may be determined. This may allow to determine the decentral identifier(s) of all input materials, such as raw materials, chemical intermediate products, chemical products and discrete products used to produce the respective output product. Using said decentral identifier(s), respective input material data may be gathered from respective data providing network node(s). The determined decentral identifier(s) may then be used to gather the associated input material data. Such recursive determination may be needed in case the output product data gathered using the decentral identifier associated with the output product does not contain the data requested by the user.

[0078] In an embodiment, gathering the output product data includes:

[0079] • gathering access element(s) associated with the produced output product based on the gathered decentral identifier(s) associated with the output product(s),

[0080] • gathering the output product data based on access data included in the access element and the query data.

[0081] The gathered access element(s) may be parsed to determine access data. The access data may include the decentral identifier associated with the output product, the decentral identifier(s) associated with output product data set(s) and an address of the data proving node associated with the dedicated storage storing the output product data set(s). The access element may include access data per output product data set. The access data may be determined by matching the access element data to the generated query data. For instance, the access data may be determined by parsing the access element data and matching the parsed access element data to the query data. The key value pair(s) included in the access element data may be matched to key-value pairs included in the query data to determine the access data. The determined access data may be used to request output product data or a part thereof from the data providing node identified in the access data. The request may include the decentral identifiers included in the access data. The output product data or the part thereof may be requested by a data consuming node.

[0082] In an embodiment, the data transformer model includes at least one data-driven model parametrized according to training data set(s) including unstructured data and associated data transformation scheme(s). The unstructured data may be associated with a context. The data transformation scheme may include database queries (e.g. database query data). The transformation scheme may further include database table(s) associated with the database queries. The queries may be based on SQL query language. The database tables may be SQL. The database queries may be based on NoSQL query language.

[0083] In an embodiment, the data transformer model trained to transform the gathered output product data and / or input material data determines database query data to retrieve output product data and / or the input material data from a database storing gathered output product data and / or input material data based on the provided user input data. The retrieved data may correspond to the data associated with the output product(s) and / or input material(s). The data transformer model may further convert the retrieved data. For instance, the data transformer model may be in communication with a calculator model configured to convert data. The calculator model may be configured to convert the data based on input received from the data transformer model. The input received from the transformer model may include at least a part of the retrieved data from the database as well as the required conversion. Conversions may include arithmetic conversions to convert one unit associated with the retrieved data to another unit associated with the search data. The result of the conversion may be returned from conversion model to data transformer model. The database query data may be determined based on the provided search data and instructions related to the determination of the database query data. The instructions may include unstructured data associated with a context. The context may be associated with the determination of the database query data. For instance, the context may include or relate to database queries usable to determine database tables and / or database content and / or algorithmic conversions required to convert one unit into another unit. The unstructured data may include unstructured text data. The unstructured text data may signify the instructions related to the determination of the database query data. The unstructured text data correspond to text in natural language. The database query data may be determined by the trained data transformer model based on input data including the search query and the instructions. The input data may be generated by merging the search data with predefined instructions. The trained data transformer model may determine the database query data by repeatedly generating query data and evaluating the receive database response with respect to the input data. The trained data transformer model may include an output parser model configured to parse the output produced by the trained data-driven model(s) of the data transformer model to determine whether the output of the trained data-driven model(s) is the data associated with the output product(s) and / or input material(s) matching the search data. For instance, the output of the trained data-driven model(s) may include an indicator whether the output is a final output or an intermediate output. The output parser model may be configured to determine such indicator and to initiate generation of further query data or provision of the final output. Use of predefined instructions may allow the trained data transformer to accurately determine database query data based on the provided search data by repeatedly generating query data based on the instructions and evaluating the received database response.

[0084] In an embodiment, the method further includes a step of storing the gathered output product data and / or input material data in a database. The gathered output product data and / or input material data may be stored in the database prior to transforming the gathered output product data and / or input material data. The database may be connected to the data consuming node gathering the output product data and / or input material data. The database may serve as a data sink for output product data and / or input material data gathered from the decentral network, allowing to collect the gathered data prior to transforming the gathered data. This may allow to collect the data in an asynchronous manner and / or within different time periods and avoids that only a fraction of the gathered output data and / or input material data is transformed and provided to the user.

[0085] In an embodiment, the embedding model is parametrized on training data set(s) including plurality of input vectors and corresponding output vectors. The embedding model may generate embedding(s) from the search data, such as the unstructured data associated with a context. The embedding model may generate embeddings from access element(s) associated with output product(s) and / or access element(s) associated with input material(s). Such embeddings may be stored within a database, such as a vector database. The embedding model may generate embeddings from data associated with access element(s), such as documentation including mandatory key-value pairs to be included in access elements for given output product types and / or input material types. Such embeddings may be stored within a database, such as a vector database.

[0086] In an embodiment, matching the embedding(s) generated from the search data with embeddings stored in the database may include determining nearest embedding vector(s) and providing the nearest embedding vector(s). The provided nearest embedding vector(s) may represent the context contained in the search data. This may allow to avoid fine-tuning of the query data generator model with new training data to ensure sufficient accuracy of the generated query data for new output product(s) and / or input materials and associated access element data.

[0087] BRIEF DESCRIPTION OF THE DRAWINGS

[0088] In the following, the present disclosure is further described with reference to the enclosed figures. The same reference numbers in the drawings and this disclosure are intended to refer to the same or like elements, components, and / or parts.

[0089] FIG. 1 illustrates a first example of a participant network of a product ecosystem associated with a decentral peer-to-peer network for exchange of data associated with input materials, chemical product(s), discrete product(s), end product(s) and recycled material(s). FIG. 2 illustrates a system for providing output product data for access via a decentral network in accordance with an exemplary embodiment of the present invention.

[0090] FIG. 3A illustrates a first example of a digital access element including a decentral identifier and access data.

[0091] FIG. 3B illustrates a second example of a digital access element including a decentral identifier and access data.

[0092] FIG. 4 illustrates an example system for determining access elements allowing access to data associated with output product(s) and data associated with production input(s) used to produce such output product(s) via decentral network in accordance with an exemplary embodiment of the present invention.

[0093] FIG. 5 illustrates a block diagram of an example implementation of a neural network engine for generating query data to retrieve data associated with product(s) or input material(s) used to produce the product(s) from a decentral network and to generate response data from the retrieved data in accordance with an exemplary embodiment of the present invention.

[0094] FIG. 6 illustrates a block diagram of an example implementation of a data query generator included in the neural network engine in accordance with an exemplary embodiment of the present invention.

[0095] FIG. 7 illustrates a block diagram of an example implementation of a search result compositor included in the neural network engine in accordance with an exemplary embodiment of the present invention.

[0096] FIG. 8A illustrates an example of a training process of the data query generator included in the neural network engine in accordance with an exemplary embodiment of the present invention.

[0097] FIG. 8B illustrates an example of a neural network system and a training process thereof included in the search result compositor of the neural network engine in accordance with an exemplary embodiment of the present invention.

[0098] FIG. 9 illustrates an example system for gathering or retrieving data associated with product(s) or input material(s) used to produce the product(s) via decentral network in accordance with an exemplary embodiment of the present invention. FIG. 10 illustrates a flow chart of an example method for gathering or retrieving data associated with product(s) and / or input material(s) used to produce the product(s) from a decentral network in accordance with an exemplary embodiment of the present invention.

[0099] FIG. 11 illustrates a flow chart of a further example method for retrieving data associated with product(s) and / or input material(s) used to produce the product(s) from a decentral network in accordance with an exemplary embodiment of the present invention.

[0100] FIG. 12 illustrates an aspect of the methods of FIG. 10 and FIG. 11 accordance with one embodiment of the present invention.

[0101] FIG. 13 illustrates a sequence diagram of a method for gathering or retrieving data associated with product(s) or input material(s) used to produce the product(s) via decentral network in accordance with an exemplary embodiment of the present invention.

[0102] FIGs. 14A, 14B illustrate graphical user interfaces displayed by the consumer app described in the context of FIG. 13.

[0103] FIG. 15 illustrates an example system for retrieving or gathering production input data based on a decentral identifier(s) associated with output product(s) produced from such production input(s) in accordance with an exemplary embodiment of the present invention.

[0104] FIG. 16 illustrates an example method for retrieving or gathering production input data based on a decentral identifier(s) associated with output product(s) produced from such production input(s) in accordance with an exemplary embodiment of the present invention.

[0105] FIG. 17 illustrates an embodiment of training an embedding layer.

[0106] DETAILED DESCRIPTION

[0107] FIG. 1 illustrates a first example of a participant network of a product ecosystem associated with a decentral peer-to-peer network for exchange of data associated with input materials, chemical product(s), discrete product(s), end product(s) and recycled material(s). Such data may include production data, property data, use data, environmental attribute data or a combination thereof. Production data may include data associated with the production of the input materials, chemical product(s), discrete product(s), end product(s) and recycled material(s). Property data may include measured physical and / or chemical properties and / or physical and / or chemical properties determined from measured data. Use data may include data associated with the use of the input materials, chemical product(s), discrete product(s) and end product(s). The use data may be acquired during use. Environmental attribute data may include emission data, recycled content data, biobased content data, biodegradability data, renewable content data or a combination thereof.

[0108] The decentral network environment may include a decentral participant network 136. The decentral participant network 136 may include one or more decentral network participants 102 to 114. The decentral network participants may be part of a product ecosystem including chemical products. The product ecosystem may include production chains to produce an end-product. The product ecosystem may include recycling chains to recycle at least part of an end-of-life product resulting from the use of the end product. The product ecosystem may include a raw input material producer 104, a chemical product producer 102, a chemical product user 106, an end-product producer 108, an end-product user 110 an EOL product collector 112 and a recycler 114. The decentral participant network 136 may be a chemical supply chain. The product ecosystem may allow to use materials resulting from recycling of end-of-life products to produce new products, such as chemical products. The product ecosystem may be associated with the production and / or recycling of physical products. The product may be a chemical product, an intermediate chemical product, a component, a component assembly, an end product, an end-of-life product or a recycled product.

[0109] The participant(s) of the decentral participant network 136 may be associated with the production the product and / or recycling of the product. The decentral network participant 102 to 114 may refer to a manufacturer of physical products, such as input material producer 104, chemical product producer 102, chemical product user 106, end-product producer 108, a user of physical goods, such as endproduct user 110, and / or a participant of a recycling chain associated with the physical product, such as EOL product collector 112 and recycler 114. The decentral network participant may be associated with a decentral participant identifier. The decentral participant identifier may uniquely identify the decentral network participant within the decentral participant network 136.

[0110] The participant(s) of the decentral participant network 136 may be connected via material flow 142. The material flow 142 may correspond to the flow of product from one participant of the decentral participant network 136 to the downstream participant of the decentral participant network 136. The material flow 142 may refer to a continuous or a discontinuous flow of product. The flow of product may include any means of transportation suitable to transport the product from a participant to the downstream participant. The means of transportation may include pipes, containers, barrels, packages. The material flow 142 may be associated with raw materials 118 used to produce the chemical product, such as virgin raw materials. The raw materials may be provided to chemical product producer 102 for producing chemical product(s) and / or intermediate chemical product(s) (not shown). The material flow 142 may be associated with chemical product(s) 120. The chemical product(s) 120 may be provided to chemical product user 106 for producing discrete product(s). In contrast to chemical production, the discrete products being produced are distinct units sold as individual products. The material flow 142 may be associated with recycled material 116. The recycled material 116 may be provided to chemical product producer 102 for the production of chemical product(s). At least part of the participants of the decentral participant network 136 may be associated with decentral participant network nodes 122 to 134. The decentral participant nodes 122 to 134 may be under control of the respective decentral participant associated with the respective decentral participant node. The decentral participant nodes 122 to 134 may form decentral network 140. The decentral network 140 may be a peer-to-peer communication network. The decentral network 140 may be configured to perform data transactions 138. The data transactions 138 may be based on a transaction protocol including authentication and / or authorization mechanism(s). Based on the authentication and / or authorization mechanism(s) a peer-to-peer communication between decentral network nodes 122 to 134 associated with decentral network participants 102 to 114 may be established. The one or more authentication mechanism(s) may be associated with or linked to a decentral identifier as described in the context of FIG. 9. The one or more authentication mechanism(s) associated with the decentral identifier may be accessible by the decentral participant nodes as described in the context of FIG. 9. The decentral configuration allows for more efficient use of computing resources and strengthens control by the data owners of the decentral network.

[0111] Data transactions between decentral network participant nodes may be based on a decentral identifier associated with respective data to be accessed, for example as described in the context of FIG. 9 to FIG. 13. The decentral identifier may be uniquely associated with the physical entity of a produced output product and associated data. The decentral identifier may be uniquely associated with the physical entity of an input material and associated data. The decentral identifier may uniquely identify the respective output product or input material within the decentral network. The decentral identifier may be associated with further decentral identifier(s), such as decentral identifier(s) of production input(s) used to produce the input material or output product. This may allow to track the production input(s). The decentral identifier may be included in a digital access element associated with the product, for example as described in the context of FIG. 3A and FIG. 3B.

[0112] The data flow 138 (e.g. transactions) between decentral network participant nodes may be directly or indirectly associated with the material flow 142 between the decentral network participants. For instance, data flow 138 may be directly associated with material flow 142 if data associated with an input material provided from the input material producer 104 to the chemical product producer 102 is accessed by decentral participant node 124 associated with said chemical product producer 102. For instance, data flow 138 may be indirectly associated with material flow 142 if data associated with a chemical product produced by chemical product producer 102 is accessed by decentral participant node 134 associated with recycler 114.

[0113] The decentral participant nodes 122 to 134 may be decentral computing nodes. The decentral computing node may be any device or system that includes at least one physical and tangible processor, and a physical and tangible memory capable of having thereon computer-executable instructions that are executed by a processor. The memory may take any form and depends on the nature and form of the computing node. At least part of the decentral participant nodes 122 to 134 may be decentral data providing network nodes. At least part of the participant nodes 122 to 134 may be decentral data consuming network nodes. A participant of the decentral participant network 136 may be associated with a decentral data providing network node and / or a decentral data consuming network node depending on whether data is provided to downstream participants and / or consumed from upstream participants. For instance, input material producer 104 may be associated with a decentral data providing network node configured to provide input material data to a downstream participant (e.g. chemical product producer 102) for example as described in the context of FIG. 2. In addition to or alternatively, chemical product producer 102 may be associated with a decentral data consuming network node configured to access data associated with a recycled input material produced by an upstream participant (e.g. recycler 114).

[0114] The decentral network 140 may include further decentral network nodes. The further decentral network nodes may be decentral infrastructure service nodes (not shown in FIG. 1). The decentral infrastructure service nodes may not be associated with a participant of the product ecosystem. The decentral infrastructure service nodes may provide services for decentral participant nodes 122 to 134, such as verifying the identity of the decentral network participant nodes 122 to 134 prior to performing a data exchange. The decentral network participant nodes 122 to 134 may be associated with or include certificate(s), such as X.509 certificate(s). The certificate(s) may be associated with decentral infrastructure service node(s) including e.g. a certificate issuing service and / or a dynamic provisioning service providing dynamic attribute tokens (e.g. OAuth Access Tokens). This way the decentral network participant nodes 116 to 124 possess a unique identifier embedded in a X.509 certificate that identifies the respective decentral network participant node 122 to 134. The information required to verify the certificate may be provided via an authentication registry associated with the certificate issuing service and / or a dynamic provisioning service. For instance, in the IDSA Reference Architecture Model, Version 3.0 of April 2019, a decentral data providing network node associated with a data owner, a Certification Authority (CA), a Dynamic Attribute Provisioning Service (DAPS) and a decentral data consuming network node associated with a data consumer are used to verify the identity prior to performing a data exchange (not shown).

[0115] FIG. 2 illustrates a system for providing digital twin(s) or a part thereof of output products for access via a decentral network in accordance with an exemplary embodiment of the present invention. The output product(s) may be associated with respective digital twin(s). The digital twin may be a digital representation of a physical entity of the output product with a defined semantic description of said physical entity of the output product. The digital twin of the physical entity is hence a digital version of said physical entity. Once created, the digital twin may be used to represent the physical entity in a digital representation of a real-world system. The digital twin may be uniquely linked to the physical output product via digital identifier(s), such as decentral identifier(s). The digital twin may be created such that it is identical in form and behavior of the corresponding physical entity of the output product. Additionally, the digital twin may mirror the properties of the physical entity during its lifetime. For example, sensors may capture real-time (or near real-time) data, such as transport data or use data, from the physical entity of the output product to relay it back to a remote digital twin. Sensor may include hard sensors and / or soft sensors. The digital twin may then be updated to maintain its correspondence to the physical entity of the output product. Hence, the digital twin may at any time represent the current state of the physical entity. The digital twin may contain one or more data sets (also referred to as aspects or assets of the digital twin or passports of the output products). Each data set may be associated with an identifier (aspect or asset identifier). This allows to uniquely identify each data set contained in the digital twin. The digital twin may further contain a digital identifier of the output product.

[0116] The decentral network may correspond to decentral network 140 illustrated in the context of FIG. 1. The output product may be a chemical product. Chemical products may include chemical intermediate products, chemical end products and recycled material. The output product may be a discrete product. Discrete products may include discrete components, discrete parts, discrete part assemblies and discrete end products. Production input(s) (also denoted as input material(s) hereinafter) may include chemical materials. Chemical materials may include raw materials (e.g. virgin materials), recycled materials and chemical intermediate product(s). Production input(s) may include discrete material(s). Discrete material(s) may include components, parts or part assemblies.

[0117] A production 212 may produce one or more output product(s) 214 from one or more production input(s) 210. The one or more production input(s) 210 may be received from input material producer 104 (see FIG. 1). The one more production input(s) 210 may be received from recycler 114 (see FIG. 1). The one more production input(s) 210 may be received from chemical product producer 102 (see FIG. 1). The one more production input(s) 210 may be received from chemical product user 106 (see FIG. 1). The production 212 may be operated by an operating system. The operating system may include or be associated with data provider backend 222.

[0118] Data related to the production of the output product 214 may be collected by product data collector 202. Data related to the production of the produced output product 214 may include output product data. The output product data may include production data associated with the production of the output product, output product composition data, output product property data, environmental attributes associated with the output product, output product material safety data or a combination thereof. The composition data may relate to the material composition of the output product. The composition data may specify the material composition of the output product. For example, the composition data may specify at least part of the chemical composition of the output product. The property data may relate to or include a property of the output product. The property output product may include a performance property, a chemical property, such as flammability, toxicity, acidity, reactivity, heat of combustion or the like, and / or a physical property such as density, color, hardness, melting points, boiling points, electrical conductivity or the like. The environmental attribute data may include emission data, recyclate content, biobased content and / or renewable content data and / or biodegradability data. The output product data may include measured data and / or data determined from measured data and / or from input material data associated with production input(s) 210. The output product data may be collected prior to, during and / or after production of the output product 214. Data related to the production of the produced output product 214 may further include input material data associated with the production input(s) 210. The input material data may be collected via a decentral network 140 using data consuming network node 126A associated with the entity operating the production 212 (in this example chemical product user 106) from a decentral data providing network node associated with the producer of the production input(s) 210 (in this example chemical product producer 102) (see also FIG. 1). The input material data may be generated in association with one or more network nodes, e.g. in relation to data generating node(s), associated with the respective production input producer. The input material data may relate to different stages of the production chain of the output product. The input material data may relate to the production of production input(s) 210 used to produce the output product. The input material data may be specific to any production input(s) 210 used to produce the output product or to the produced output product 214. The input material data may be specific to individual entities or batches of the production input(s) 210 or to the produced output product 214. Collecting input material data may include gathering decentral identifiers(s) associated with one or more production input(s) 210 used to produce the output product 214. Such decentral identifier(s) may be collected by reading an identifier element, such as a bar code or QR code, connected to the production input(s) 210. The identifier element may encode a digital input material identifier which may be used to gather input material data, for example using the principle illustrated in FIG. 4. Access to the input material may be controlled by the input material producer via the decentral identifier(s). The input material data may be collected by product data collector 202 of data provider backend 222. The input material data may be collected prior to, during and / or after production of the output product 214.

[0119] The data related to the production may be stored in one or more databases associated with production 212. The data related to the production may be stored in one or more databases associated with the operating system operating production 212. At least part of the collected data related to the production of the output product, such as the output product data, may be owned by the entity operating production 212. The data related to the production may be collected by the entity operating production 212. For instance, the operating system associated with production 212 may be configured to collect such data. The data related to the production may be collected on behalf of the entity operating the production 212. The data related to the production may be stored on a data storage. The data storage may be owned by or associated with or accessible for the entity operating production 212. The data storage may be associated with or accessible by data provider backend 222.

[0120] The data related to the production may be collected from one or more databases associated with production 212 by product data collector 202. The data related to the production may be collected in response to a received request. The request may be received from DT generator 216. The request may include data related to the produced output product. Data related to the produced output product 214 may include a digital output product identifier, such as a batch ID, a LOT number, an order number or a combination thereof. In response to the request, data related to the production may be collected by product data collector 202 from one or more databases storing such data. The data may be collected based on the data related to the produced output product contained in the received request. The collected data may be provided to DT generator 216. One or more decentral identifier(s) associated with the produced output product 214 may be provided by ID provider 204. The one or more decentral identifier(s) may be provided in response to a received request. The request may be received from DT generator 216. The request may include data related to the produced output product 214 as previously described. The one or more decentral identifier(s) may be uniquely associated with the physical entity of the output product. The one or more decentral identifier(s) may comprise any unique identifier uniquely associated with the output product(s), the production participant(s) of the production chain producing the output product(s) and / or the data related to the production of the output product. The decentral identifier(s) may include or relate to one or more Universally Unique Identifier(s) (UUID(s)) or one or more Digital Identifier(s) (DID(s)). Via the decentral identifier(s) and the unique association with the output product, the producer of the output product and / or data related to the production of the output product, the access to the digital twin data or a part thereof may be controlled by the producer of the output product. This contrasts with central authority schemes, where identifiers are provided by such central authority and access to such data is controlled by such central authority. Decentral in this context refers to the usage of the identifier in implementation as controlled by the data owner. The decentral identifier may be discoverable for decentral participant nodes registered within the decentral network. Discovery of the decentral identifier may be controlled by the data owner of data associated with the decentral identifier.

[0121] ID provider 204 may provide the one or more decentral identifier(s) to DT generator 216. DT generator 216 may be configured to generate digital twins associated with produced output product(s) 214. DT generator 216 may be configured to apply one or more semantic models (also denoted as aspect model(s)) to the data proved by product data collector 202. The semantic model(s) may be stored in a database accessible by DT generator 216. DT generator 216 may be configured to gather the semantic model(s) from such database based on data associated with such semantic model(s). Such data may be included in a request to generate a digital twin received by DT generator 216. Data associated with semantic model(s) may include a semantic model ID, semantic model name or a combination thereof. Application of a semantic model to the data provided by product data collector 202 may result in generation of a data set having a defined semantic structure (e.g. an aspect, asset or passport). The data set may be associated with decentral identifier(s) provided ID provider 204. The data set may include decentral identifier(s) provided by ID provider 204. The decentral identifier(s) may include a digital twin identifier and / or a data set identifier (e.g. aspect or asset identifier). The data set may include a specific part of the data related to the production as provided by product data collector 202. For instance the data set may include specific production data or specific composition data or specific environmental attribute data or material safety data. The generated digital twin for a given product may include the decentral digital twin identifier provided by ID provider 204 and the data sets generated forthat given product. The digital twin may further include or be associated with the data set identifier(s) associated with said data set(s). The generated digital twin may signify at least a part, in particular all, data set(s) generated by DT generator 216 for a given output product. The generated digital twin, e.g. the generated data set(s), may be provided to a database associated with data provider backend 222. The database 220 may be accessible for data consumer nodes, such as node 134 associated with recycler 114 via data provider node 126B associated with the data owner of the data stored in DT storage 220, e.g. the entity operating production 212, such as chemical product user 106. Consumer node(s) may hence access the data set(s) stored in DT storage 220 under control of the provider node 126B associated with the data owner of the data stored in DT storage 220. The generated digital twin may be stored in the database 220 for access by data consumer(s), such as output product users or consumers. Access to database 220 may be controlled by the data owner of the digital twins stored in such database based on the decentral identifier(s) included in the digital twin or parts thereof.

[0122] In addition to the digital twin, data provider backend 222 may be configured to generate access element(s) associated with the produced output product(s) 214 and the digital twin(s) generated by DT generator 216. The access elements may be generated by access element generator 208. Exemplary access elements generated by access element generator 208 are illustrated in FIG. 3A and FIG. 3B. The access element may include access data for one or more aspect(s).The access element generator 208 may be configured to generate one access element per data set generated by DT generator 216. The access element generator 208 may be configured to generate one access element per digital twin generated by DT generator 216. The generated access element(s) may include one or more decentral identifier(s) and access data associated with one or more data set(s) generated by DT generator 216. The decentral identifier(s) included in the access element may relate to or correspond with the decentral identifier(s) associated with or included in the data set(s) (e.g. the decentral digital twin identifier and data set identifier(s)). Hence, the access element may be associated via the decentral identifier(s) with the physical entity of the output product.

[0123] The access data may be associated with the dedicated storage storing the digital twin, e.g. DT storage 220. The access data may relate to the dedicated storage storing the digital twin, e.g. DT storage 220. The access data may point to the dedicated storage storing the digital twin, e.g. DT storage 220. The access data may be a representation for accessing the digital twin or a part thereof. The access data may include a locator or pointer locating or pointing to the provider node associated with the dedicated storage storing the digital twin, e.g. to provider node 126B associated with DT storage 220. The access data may include a locator or pointer locating or pointing to the dedicated storage storing the digital twin, e.g. to DT storage 220. The access data may include a locator or pointer locating or pointing to the storage location of data set(s) stored DT storage 220. The access data may include one or more digital link(s) pointing to the dedicated storage storing the digital twin, e.g. DT storage 220. The access data may include a locator or pointer, such as am url or uri, to a dedicated storage address of the dedicated storage storing the digital twin, e.g. DT storage 220. The access data may be associated with or include a decentral participant identifier associated with the entity operating the provider node. The access data may be associated with or include the decentral identifier(s). The access data may include a locator or pointer, such as an url or uri, of the provider node associated with the dedicated storage storing the digital twin, a decentral participant identifier associated with the entity operating the provider node and the decentral identifier(s) (e.g. the decentral digital twin identifier and data set identifier(s)). The access data may comprise at least one interface to a data providing network node. The access data may include at least one interface to a data consuming network node. The access data may include an endpoint for data exchange or sharing (resource endpoint) or an endpoint for service interaction (service endpoint), that may be uniquely identified via the data transaction protocol.

[0124] The access element may further include one or more authentication mechanism(s) associated with the decentral identifier(s) and the access data. The one or more authentication mechanism(s) may be associated with or linked to decentral identifier(s). The one or more authentication mechanism(s) associated with the decentral identifier(s) may be accessible by the decentral participant network node(s). The access element may further relate to authorization information linked to the decentral identifier(s). The authorization information may be associated with the decentral identifier(s) and the access data. The authorization information may include access rules depending on the data set to be accessed and the role of the accessing data consuming network node.

[0125] Access element generator 208 may be configured to provide the generate access element(s) to a decentral registry, such as decentral registry 218. The decentral registry may be configured to store access elements. The decentral registry may be accessible by consumer nodes of the decentral network, such as decentral network 140, via a provider node, such as node 126B associated with such registry. This may allow to control access to such registry by consumer nodes, such as node 134, via the provider node associated with such registry. This may allow to control access to such registry by the data owner of the digital twins associated with the access elements stored in the decentral registry 218. The access element(s) stored in decentral registry 218 may be searched by consumer nodes using query data, for example as described in the context of FIG. 4 to FIG. 7.

[0126] What has been described in the context of FIG. 2 in relation to the generation of digital twins and access elements for produced output products is likewise applicable for the generation of digital twins and access elements for production input(s).

[0127] FIG. 3A illustrates a first example of an access element including a decentral identifier and access data. The access element may be a digital access element. The access element may be generated by access element generator 208 of FIG. 2. The access element may provide access information to access the digital twin or parts thereof of an output product or a production input associated with such access element. The access element may be stored within a decentral identity infrastructure, such as a decentral registry (see for example 218 of FIG. 2). The decentral identity infrastructure may allow retrieval of the access element using decentral identifier(s) associated with such access element.

[0128] The decentral identifier may include Decentralized Identifier (DID). The decentral identifier-based access element may in this case be a DID document 304 associated with the DID. Besides the DID document 304 serving as access element, FIG. 3A shows a DID owner data element 302 including decentral identifier-based owner data. Generally, the decentral identifier-based owner data may include the decentral identifier associated with a subject such as an output product or production input and may include one or more authentication mechanism(s). The decentral identifierbased owner data 302 may include owner data that is electronically owned and controlled by the DID owner. In this context electronically owned may refer to data that is stored in an owner repository or wallet. Such data may be securely stored and / or managed on an organizational server or client device. The decentral identifier-based owner data 302 may include a DID, a private key and a public key. The DID owner may own and control the DID that represents an identity associated with the DID subject, a private key and public key pair that are associated with the DID. DID may be understood as an identifier and authentication information associated with or uniquely linked to the identifier.

[0129] The DID subject may be a raw material, a basic substance, a chemical product, a component, an end product or a recycled material. The DID subject may be a machine, a system, or a device used for producing the raw material, the basic substance, the chemical product, the component, the end product, the recycled material or a collection of such machine(s), device(s) and / or system(s). The DID owner may be a supply chain participant or a manufacturer such as a chemical manufacturer producing chemicals. The DID owner may be an upstream participant of chemical product user 106 such as a supplier that supplies production input(s) 210. The DID owner may be a downstream participant of the chemical product user 106 such as end-product producer 108. The DID owner may be any participant of the product ecosystem including raw chemical product supplier, intermediate chemical product manufacturer, intermediate part manufacturer, component manufacturer, end product manufacturer, end product user, EOL collector and / or sorter or recycler.

[0130] The DID may be any identifier that is associated with the DID subject and / or the DID owner. Preferably, the identifier is unique to the DID subject and / or DID owner. The identifier may be unique at least within the scope in which the DID is anticipated to be in use. The identifier may be a locally or globally unique identifier for the raw material, a basic substance, a chemical product, a component, an end product, the recycled material or a collection thereof; the machine, the system, or the device used for producing the raw material, a basic substance, a chemical product, a component, an end product, the recycled material, or the collection of such machine(s), device(s) and / or system(s); the chemical manufacturer producing chemicals, the upstream participant of the chemical manufacturer, the downstream participant of the chemical manufacturer or a collection thereof; any participant of the material ecosystem including raw chemical product supplier, intermediate chemical product manufacturer, intermediate part manufacturer, component manufacturer, component assembly manufacturer, end product manufacturer, end product user, EOL collector, recycler or a collection thereof.

[0131] The DID may be any identifier that is associated with the DID subject and the DID owner. Preferably, the DID is unique to the DID subject and / or DID owner. The DID may be unique at least within the scope in which the DID is anticipated to be in use. The DID may be a locally or globally unique identifier for any of the above mentioned possible DID subjects. The DID may also be a Uniform Resource Identifier (URI) such as a Uniform Resource Locator (URL). Moreover, the DID may be an Internationalized Resource Identifier (IRI). The DID may be a Uniform Resource Identifier (URI) such as a Uniform Resource Locator (URL). The DID may be an Internationalized Resource Identifier (IRI). The DID may be a random string of numbers and letters for increased security. In one embodiment, the DID may be a string of 128 letters and numbers e.g. according to the scheme did:method name: method specific-did such as did:example:ebfeb1f712ebc6f1 c276e12ec21 . The DID may be decentralized ID independent of a centralized, third party management system and under the control of the DID owner.

[0132] The access element as DID document data 304 may be associated with the DID, i.e. the DID included in the decentral identifier-based owner data 302. Accordingly, the access element may include a reference to the DID, which is associated with the DID subject that is described by the DID document 304. The DID document 304 may also include an authentication information such as the public key. The public key may be used by third-party entities that are given permission by the DID owner / subject to access information and data owned by the DID owner / subject. The public key may also be used for verifying that the DID owner, in fact, owns or controls the DID. The DID document may include authentication information, authorization information e.g. to authorize third party entities to read the DID document or some part of the DID document e.g. without giving the third party the right to prove ownership of the DID.

[0133] The access element 304 may include access data that digitally links to the digital twin of the output product or production input the digital access element is associated with, e.g. by way of service endpoints. A service endpoint may include a network address at which a service operates on behalf of the DID owner. In particular, the service endpoints may refer to services, such as decentral data providing network node(s), of the DID owner that give access to digital twin data, e.g. data set(s) included in the digital twin. Such services may include services to read or analyze data contained in such data set(s).

[0134] The access element 304 may include further identifiers, such as data set identifier(s) associated with data set(s) included in the digital twin data. The access element 304 may include various other information such metadata specifying when the digital access element was created, when it was last modified and / or when it expires.

[0135] The DID and access element 304 may be associated with a data registry node such as a centralized data service system or a decentralized data service system 306 , e.g. a distributed ledger or blockchain or a decentralized file system. The distributed ledger or blockchain may be used to store a representation of the DID that points to the digital access element 304. A representation of the DID may be stored on distributed computing nodes of the distributed ledger or blockchain 306. For example, DID hash may be stored on multiple computing nodes of the distributed ledger and point to the location of the digital access element 304. In some embodiments, the access element 304 may be stored on the distributed ledger 306. Each of the computing nodes may store a copy of the distributed ledger 306. In this way, each DID hash can be stored redundantly, thereby allowing for an increased data safety. DIDs associated with a plurality of different access element 304 may be included in the distributed ledger 306.

[0136] In some embodiments, the access element 304 may be stored on the distributed ledger 306, i.e. either additionally or alternatively to the associated DID representation being stored on the distributed ledger 306. In other embodiments, the access element 304 may be stored in a data storage (not illustrated) that is associated with the distributed ledger or blockchain or decentralized file system.

[0137] The distributed ledger or blockchain 306 may be any decentralized, distributed network that includes various computing nodes that are in communication with each other. For example, the distributed ledger 306 may include a first distributed computing node, a second distributed computing node, a third distributed computing node, and any number of additional distributed computing nodes (not shown). The distributed ledger or blockchain 306 may include known technology stacks like Bitcoin (see e.g. Bitcoin documentation of November 11 , 2022 published https: / / en.bitcoin.it / wiki / Protocol_documentation), Ethereum (see e.g. Ethereum documentation of August 15, 2022 published on https: / / ethereum.org / en / developers / docs / ), Solana (see e.g. Solana documentation of November 11 , 2022 published on https: / / spl.solana.com / ), Polygon (see e.g. Polygon documentation of November 11 , 2022 published on https : / / wiki. polygon. technology / ) or other implementations with varying degree of data transactions performed on the distributed ledger. The description of the example framework is only for illustrative purposes and shall not be considered limiting.

[0138] FIG. 3B illustrates a second example of an access element including a decentral identifier and access data. The access element may be a digital access element. The access element may be generated by access element generator 208 of FIG. 2. The access element may provide access information to access the digital twin or parts thereof of an output product or a production input associated with such access element. The access element may be stored within a decentral identity infrastructure, such as a decentral registry (see for example 218 of FIG. 2). The decentral identity infrastructure may allow retrieval of the access element using decentral identifier(s) associated with such access element.

[0139] The decentral identifier 310 may include one or more Universally Unique Identifier(s) (UUID(s)). The decentral identifier may include a first decentral identifier and a second decentral identifier. The first and the second decentral identifier may be different from each other. The first decentral identifier may signify the decentral digital twin identifier associated with the digital twin of the output product or production input, respectively, while the second decentral identifier may signify a decentral identifier associated with a data set included in said digital twin. Combination of the first and the second identifier allows to uniquely identify a given data set within a digital twin including a plurality of different data sets or assets.

[0140] The UUID(s) may be unique at least within the scope in which the UUID(s) are anticipated to be in use. The UUID(s) may be a locally or globally unique identifier for a raw material, a basic substance, a chemical product, a component, a component assembly, an end product or a recycled material. In one embodiment, the UUID may be a string of 128 letters and numbers e.g. according to the scheme [0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{12}. The UUID(s) may be decentralized ID(s) independent of a centralized, third party management system and under the control of the data owner owning the digital twin data associated with the UUID(s). The decentral identifier-based digital access element 308 may in this case be a JSON data structure including the decentral identifiers 310. The JSON data structure may include one or more key-value pairs. The JSON data structure may include one or more arrays. At least a part of the arrays may include one or more object(s). The object(s) may include one or more key-value pair(s).

[0141] The access element 308 may include access data 312 that digitally links to data set(s) included in the digital twin the access element is associated with, e.g. by way of service endpoints as described in the context of FIG. 3A. A service endpoint may include a network address at which a service operates on behalf of the data owner. In particular, the service endpoints may refer to services, such as decentral data providing network node(s), of the data owner that give access to digital twin data associated with the output product or production input(s) the access element is associated with. Such services may include services to read or analyze data contained in such data set(s).

[0142] The access element 308 may include various other information such metadata specifying when the digital access element was created, when it was last modified and / or when it expires (not shown in FIG. 3B).

[0143] The access element 308 may be stored in a data registry node such as a decentral registry (not shown, see for example decentral registry 218 in FIG. 2). The access element may be accessible for data consuming nodes via data providing nodes associated with the registry as described in the context of FIG. 2.

[0144] FIG. 4 illustrates an example decentral system for determining access elements enabling access to data associated with output product(s) and / or data associated with production input(s) used to produce such output product(s) via decentral network in accordance with an exemplary embodiment of the present invention. The output product(s) may be produced by a production from one or more production input(s) (e.g. input material(s)) as described in the context of FIG. 2. A digital twin may be generated for each physical entity of produced output product or production input (which corresponds to an output product of an upstream participant of the product ecosystem) as described in the context of FIG. 2. The digital twins may include decentral identifier(s) and one or more data set(s) and may be stored in a database (see FIG. 2). An access element may be generated for or each physical entity of produced output product or production input as described in the context of FIG. 2. The access elements may include decentral identifier(s) and access data may be stored in a decentral registry (see FIG. 2). A decentral data providing network node may be associated with the database and registry to allow access to such access elements and data set(s) by decentral data consuming network nodes (see FIG. 2). The access to such access elements and data set(s) may be controlled by the decentral data providing network node.

[0145] The decentral system may be used to implement parts of the sequence diagram illustrated in FIG. 13. The decentral system may be decentral system 140 described in the context of FIG. 1 . The decentral system may be used to query the decentral network for decentral identifier(s) associated with output product(s) and production input(s). The decentral system may be used to determine access element(s) associated with such decentral identifier(s).

[0146] The participants of the decentral network may be associated with decentral participant identifiers. Each participant of the decentral network may be associated with one or more decentral participant identifier(s). The decentral participant identifier may comprise any identifier uniquely associated with a participant of the decentral network and / or with a production site of the participant of the decentral network. The decentral participant identifier may include letters and / or numbers. The decentral participant identifier may include one or more Universally Unique Identifier(s) (UUID(s)) and / or one or more Decentralized Identifier(s) (DID(s)). The decentral participant identifier may be associated with or may include a verifiable claim or credential. The verifiable claim may be issued by a central or decentral identity issuer making one or more claims about a subject, such as a consumer entity being a trustworthy participant of the decentral network. For instance, the issuer may make a claim about a consumer (e.g. the customer entity) the DID as decentral participant identifier is associated with. The verifiable claim may include those claim(s) as well as proof instructions to prove that claim(s) have not been tampered with and were indeed issued by the claims issuer. The verifiable claim may also include duration information metadata that defines a period of time that the verifiable claim is valid for use or that defines a specific number of times that the verifiable claim is authorized for use. The verifiable claim may also include a DID of the claims issuer and / or the subject, such as a consumer entity. The verifiable claim may be signed by the claims issuer. The claims issuer may provide the verifiable claim to a claims holder, such as the consumer entity, for presentation to any relying party that relies upon the veracity of those claims, such as a decentral data provider. The signature of the verifiable claim may be validated with a public key associated with the claims issuer to determine that the customer entity is a trusted entity within the decentral network. The verifiable credential may be presented by the decentral data consuming network node and may be used by the decentral data providing network node to verify that the decentral participant associated with the decentral data consuming network node is a trusted entity within the decentral network prior to providing access to the digital twin, hence ensuring that the digital twin can be exchanged in a secure and controlled manner within the decentral network.

[0147] The decentral network 140 illustrated in FIG. 4 may include different environments, such as an infrastructure environment 406, a data consumer environment 402 and a data provider environment 1 404. The data consumer environment 402 may be associated with a participant of the product ecosystem, such as the product ecosystem illustrated in FIG. 1 . The consumer environment 402 may be associated with a participant not being part of the product ecosystem. The provider environment 1 404 may be associated with a data owner of the output product data (e.g. an output product producer) or the input material data (e.g. an input material producer). The environments may include one or more node(s) or unit(s). One or more applications, such as consumer app 408, may run on these node(s). The unit(s) may include data driven model(s), such as neural network engine 410. The data-driven model(s) may be connected to data processing unit(s). The data processing unit(s) may be configured to process data provided by the data-driven model(s) and / or to process data provided by consumer node 134. The data processing unit(s) may be configured to provide processed data to the data-driven model(s) and / or to the consumer node 134. A part of the node(s) may be connected via peer-to-peer communication channels to allow data transfer between the nodes, as described in the context of FIG. 1 and FIG. 2. The decentral network 140 may include more or less environments than illustrated in FIG. 4.

[0148] Data consumer environment 402 may include a node running a consumer app 408, a neural network engine 410 and consumer node 134 (e.g. decentral data consuming network node 134). The neural network engine 410 may include one or more data-driven model(s) (see FIG. 5 to FIG. 7). The data- driven model(s) may be trained data-driven model(s). Training may be performed as described in the context of FIG. 5 and FIG. 8A to FIG. 8B. The data-driven model(s) may include a transformer architectures. The transformer architectures may include an encoder and / or a decoder, for example as illustrated in FIG. 8A and FIG. 8B. The neural network engine 410 may include a data manager 520 (see FIG. 5). Data manager 520 may be configured to manage data received from query data generator 504 of neural network engine 410 and / or consumer node 134. Managing data may include processing data. Managing data may include generating request(s) based on data received from query data generator 504 and / or consumer node 134. The requests may be sent to consumer node 134 and / or node(s) included in the infrastructure environment 406. Managing data may include parsing data received from consumer node 134. For instance, data manager 520 may be configured to parse access element(s) received from consumer node 134 based on the provided query data to determine data signifying a relationship between the produced product and production input(s) used to produce the product. Such data may be used by data manager 520 to generate a request to gather production input data based on such determined data signifying such relationships, for example as described in the context of FIG. 15 and FIG. 16.

[0149] The consumer app 408 may be configured to communicate with neural network engine 410. The consumer app 408 may be configured to perform authentication process(es) with neural network engine 410. The consumer app 408 may be configured to perform authentication process(es) with consumer node 134, For instance, consumer app 408 may access a decentral IAM network node (not shown) and decentral IAM network node may be configured to provide an access token to consumer app 408 upon successful authentication. This access token may be used by consumer app 408 to access neural network engine 410 and / or consumer node 134. This increases security and avoids that neural network engine 410 and / or consumer node 134 is accessed and data from the decentral network is gathered by unauthorized parties. The consumer app 408 may be connected to neural network engine 410. The consumer app 408 may be connected to neural network engine 410 and consumer node 134. The consumer app 408 may include a frontend and a backend server. The frontend may run on a user device, such as a smartphone or a desktop computer. The frontend may display a graphical user interface, such as illustrated in FIG. 14A and FIG. 14B. The backend server may provide data to the frontend, for example data used to render the graphical user interface and / or data displayed within the graphical user interface. The backend server may communicate with neural network engine 410. Neural network engine 410 may be configured to process data received from consumer app 408, such as input data. Neural network engine 410 may be configured to generate query data based on the received input data. The generated query data may be provided via data manager 520 to consumer node 134 to query the decentral network for output product data and / or input material data related to the query data. Neural network engine 410 may be configured to generate search result data including data associated with the output product or production input(s) related to the received input data. Neural network engine 410 may be configured to generate search result data from the gathered data set(s). Neural network engine 410 may be configured to provide the generated search result data to consumer app 408.

[0150] Consumer node 134 may be configured to communicate with neural network engine 410. An authentication step may be necessary prior to establishing communication. Consumer node 134 may be configured to receive data from and to transfer data to neural network engine 410 after successful authentication. Consumer node 134 may be configured to gather data stored within the decentral network 140, such as endpoint(s) of provider node(s), access element(s) stored within decentral registry 218 and / or output product data / input material data stored with a dedicated storage associated with the provider node 126 of the output product data owner / input material data owner. Data consumer environment 402 may be associated with a participant of the product ecosystem, such as product ecosystem illustrated in FIG. 1. For instance, data consumer environment 402 may be associated with a recycler 114 receiving an EOL product or a component thereof from an EOL product collector 112.

[0151] Infrastructure environment 406 may include endpoint resolver 416, participant ID resolver 418 and provider node resolver 420. Endpoint resolver 416 may be configured to provide endpoints of participant ID resolver 418 and provider node resolver 420. Endpoint resolver 416 may store such endpoint(s), for example within a database associated with or accessible by endpoint resolver 416. Participant ID resolver 418 may be configured to provide decentral participant identifier(s) based on received key(s) and / or value(s). Participant ID resolver 418 may store a mapping table mapping decentral participant identifier(s) to respective key(s) and / or value(s). The mapping table may be stored in a database associated with or accessible by participant ID resolver 418. Participant ID resolver 418 may be configured to provide multiple decentral participant identifier(s) if no key(s) and / or value(s) are received with the request from model unit 910. Value(s) may include digital identifier(s), such as output product identifier(s) or input material identifier(s). Value(s) may include values related to physical identifier elements, such as serial number(s). Value(s) may include decentral identifier(s) associated with output product(s) and input material(s). Provider node resolver 420 may be configured to provide endpoint(s) of provider node(s) based on received decentral participant identifier(s). Provider node resolver 420 may store a mapping table mapping decentral participant identifier(s) to endpoints of associated provider nodes. A decentral participant identifier may be mapped per endpoint. One decentral participant identifier may map to one or more endpoint(s) of associated provider node(s), for example if an entity associated with such decentral participant identifier operates more than one provider node. Infrastructure environment 406 may not be associated with a participant of the product ecosystem, such as the product ecosystem illustrated in FIG. 1.

[0152] Data provider environment 1 404 may include provider node 126, decentral registry 218 and a data provider backend 222. Data provider environment 1 404 may be associated with a data owner, such as a manufacturer producing output product(s) or production input(s) (see FIG. 2). Decentral registry 218 may be configured to store access element(s) associated with output product(s) or production input(s) and to provide such access element(s) to data provider 126. The data provider backend 222 may be configured to generate digital twin(s) of output product(s) or production input(s) as described in the context of FIG. 2. Data provider backend 222 may be configured to generate and store access elements associated with such digital twin(s) as described in the context of FIG. 2. Provider backend 1 912 may be configured to provide key(s) and / or value(s) and associated decentral participant identifier(s) to participant ID resolver 418. In response to receiving such data from data provider environment 1 404, participant ID resolver 418 may generate mapping data mapping the decentral participant identifier to the provided key(s) and / or value(s). Participant ID resolver 418 may store the generated mapping data within a mapping table as previously described. Data provider environment 1 404 may further include a digital twin storage storing digital twins or parts thereof of such output products or production inputs (not shown, see DT storage 220 of FIG. 2 and DT storage 916, 922 of FIG. 9). The digital twin storage may be a dedicated storage associated with the data owner. The data owner may own such dedicated storage. The data owner may have access to such dedicated storage. Provider node 126 may be configured to provide data, such as access elements stored in decentral registry 218 and data sets (e.g. aspects or assets) stored in a digital twin storage. Provider node 126 may be configured to perform authentication and authorization step(s) prior to providing data, for example as described with reference to FIG. 13 later on.

[0153] Consumer app 408 may be configured to initiate the illustrated sequence. Consumer app 408 may initiate the illustrated sequence by providing a user input for gathering or retrieving output product data or input material data (see also FIG. 10 and FIG. 11). The user input may include unstructured data associated with a context. The context may relate to data associated with produced output products or production input(s) used to produce such output products. The context may be associated with instruction(s) related to the output product(s) and / or input material(s). The context may be associated with instruction(s) related to data associated with the output product(s) and / or input material(s). The instruction(s) may include output product type(s) and / or input material type(s) and / or properties of output product type(s) and / or input material type(s). The user input may correspond to a prompt. With reference to FIG. 14A, consumer app 408 may display a graphical user interface 1402 and may initiate the illustrate sequence upon detecting a user input 1406. The user input may be provided to neural network engine 410. With reference to FIG. 5 and FIG. 6, query data generator 504 of neural network engine 410 may generate query data, for example as described in the context of FIG. 5, FIG. 6 and FIG. 10. Neural network engine 410 may send a request to provide endpoints of participant ID resolver 418 and provider node resolver 420 to endpoint resolver 416. The request may be generated and send by data manager 520 of neural network engine 410. The request may be authenticated. Upon successful authentication, endpoint resolver 416 may provide the requested endpoint data to neural network engine 410. The endpoint data may be provided to data manager 520 of neural network engine 410. The endpoint data may include a URI, such as a URL, of the participant ID resolver 418 and provider node resolver 420, respectively.

[0154] Upon receiving the endpoint data from endpoint resolver 416, neural network engine 410 may be configured to request decentral participant identifier(s) from participant ID resolver 418. The request may be generated by data manager 520. The request may include a part of the query data generated by neural network engine 410 and received endpoint data. Such query data may include a key, a value or a key-value pair(s). The request may be authenticated. Upon successful authentication, participant ID resolver 418 may provide one or more decentral participant identifier(s) matching the data contained in the received request. The decentral participant identifier(s) may be provided to neural network engine 410. The decentral participant identifier(s) may be provided to data manager 520 of neural network engine 410.

[0155] Upon receiving the decentral participant identifier(s) from participant ID resolver 418, neural network engine 410 may be configured to request endpoint(s) of provider node(s) associated with said decentral participant identifier(s) from provider node resolver 420. The request may be generated by data manager 520 of neural network engine 410. The request may include one or more decentral participant identifier(s) received from participant ID resolver 418. The request may be authenticated. Upon successful authentication, provider node resolver 420 may provide one or more endpoint(s) associated with the received decentral participant identifier(s). Provider node resolver 420 may provide more than one endpoint per decentral participant identifier. The endpoint(s) may be URI(s), such as URL(s) of respective provider node(s). In another embodiment (not shown), neural network engine 410, in particular data manager 520 of neural network engine 410, may be configured to send a request for such endpoint(s) to consumer node 134 and consumer node 134 may be configured to request such endpoint(s) from provider node resolver 420. The request may include at least a part of the decentral participant identifiers included in the request received from neural network engine 410.

[0156] Neural network engine 410 may be configured to generate a request for decentral identifier(s) associated with the query data. The request may be generated by data manager 520 of neural network engine 410. The request may include the query data and endpoint(s) received from provider node resolver 420. The request may be provided to consumer node 134.

[0157] Consumer node 134 may be configured to request decentral identifier(s) from provider node(s) associated with the endpoint(s) received from provider node resolver 420. The request may be generated in response to receiving a request from neural network engine 410. The request may be generated upon receiving endpoint(s) from provider node resolver 420. A request may be sent per endpoint. With reference to FIG. 9, a request for decentral identifier(s) may be send to provider node 1 126 of data provider environment 1 404 and provider node 2 126 of data provider environment 2 906. The request may be authenticated. Such authentication may be based on data related to an authentication mechanism. The authentication mechanism may be based on certificate(s) and / or token(s), for example a device certificate (X.509v3), a TLS connection certificate (X.509v3) and a ‘Dynamic Attribute Token’ (OAuth Access Token), associated with the respective decentral participant nodes, e.g. consumer node 134 and provider node 126. If authentication fails, no data may be provided by respective data provider(s).

[0158] With continued reference to FIG. 9, the provider node(s) receiving the request (e.g. provider node 1 126 and provider node 2 126) may be configured to query associated decentral registry(ies) 218, 918 to determine whether the associated registry includes decentral identifier(s) related to the query data contained in the received request from the consumer node 134. Such query may be performed if the authentication is valid. Provider node(s) not having determined decentral identifier(s) related to the query data may send a respective response to consumer node 134. Provider node(s) 126 having determined decentral identifier(s) related to the query data may initiate contract negotiations with consumer node 134. Provider node 126 may provide an electronic contract to consumer node 134. The electronic contract may include one or more authorization rule(s) associated with the decentral identifier. The electronic contract may be provided to consumer app 408. Consumer app 408 may display the contents of the electronic contract and may request signature of the electronic contract by the user. This allows the user to determine access and usage conditions associated with the desired data. If the user signs the electronic contract, consumer app 408 may provide data being indicative of the signature, such as a token, to consumer node 134. If the user does not sign the electronic contract, consumer app 408 may likewise forward data being indicative of declining the contract to consumer node 134. Consumer node 134 may forward this data to provider node 126. Upon declining the contract, provider node 126 may terminate the connection and may not provide any data. Use of the electronic contract ensures that the consumer node 134 and further systems, such as consumer app 408 and / or neural network engine 410, handling the data are complying to at least one policy associated with the data.

[0159] Provider node(s) having determined decentral identifier(s) related to the query data may provide such decentral identifier(s) to consumer node 134. Such identifier(s) may be provided upon successful contract negotiation. Consumer node 134 may provide received decentral identifier(s) to neural network engine 410. Consumer node 134 may provide received decentral identifier(s) to data manager 520 of neural network engine 410.

[0160] With continued reference to FIG. 9 and upon receiving decentral identifier(s), neural network engine 410 may be configured to generate a request to gather access element(s) associated with at least a part of the received decentral identifier(s). The request may be generated by data manager 520 of neural network engine 410. The request may include respective decentral identifier(s). The request may be provided to consumer node 134. Consumer node 134 may be configured to request respective access element(s) from provider node(s) providing decentral identifier(s) included in the received request from neural network engine 410. The request to respective provider node(s) may include the decentral identifier(s) and the decentral participant identifier. Upon receiving the request, provider node(s) 126 may gather access element(s) from associated decentral registries 218, 918 based on the decentral identifier(s) contained in the received request. The gathered access element(s) may then be provided to consumer node 134. Consumer node 134 may provide the received access element(s) to neural network engine 410. Consumer node 134 may provide the received access element(s) to data manager 520 of neural network engine 410.

[0161] With continued reference to FIG. 9, neural network engine 410 may further be configured to generate a request to gather decentral identifier(s) associated with production input(s) based on the query data. For instance, if the query data includes an indication that data on production input(s) is to be retrieved and / or gathered, neural network engine 410 may be configured to generate such a request. The request may be generated by data manager 520 of neural network engine 410. Generating the request may include parsing the access elements received from consumer node 134 to determine access data signifying a relationship between the produced output product(s) and production input(s) used to produce the output product(s). The request may include the decentral identifier(s) of output products for which decentral identifier(s) of production input(s) are to be determined and the determined access data. The request may be provided to consumer node 134 or a node configured to gather production input data, for example as described in the context of FIG. 15 and FIG. 16.

[0162] With continued reference to FIG. 9, neural network engine 410, such as data manager 520 of neural network engine 410, may be configured to parse the received access element data. Neural network engine 410 may be configured to determine access data associated with data set(s) matching the query data. Data manager 520 of neural network engine 410 may be configured to determine access data associated with data set(s) matching the query data. For instance, the query data may include an indication about what data on an output product or production input is required. Neural network engine 410, in particular data manager 520 of neural network engine 410, may match at least a part of the query data to access element data to identify access data included in the access element data and matching the query data. The query data may include instructions to determine matching access data. Neural network engine 410 may be configured to generate a request to gather digital twin data based on the determined access data. The request may be generated by data manager 520 of neural network engine 410. The request may include the access data. The request may be generated per data set to be gathered. The request may be generated per access element received from consumer node 134. The request may be forwarded to consumer node 134. Consumer node 134 may generate a request to gather said data set(s) from respective provider node(s). The request may include the identifiers included in the access data received from neural network engine 410 and the decentral participant identifier associated with consumer node 134. The request may be provided to respective provider node(s) 126 identified in the determined access data. Consumer node 134 and provider node(s) 126 may be authenticating as previously described.

[0163] With continued reference to FIG. 9, provider node(s) 126 may initiate contract negotiations with consumer node 134 as previously described. Contract negotiations may be initiated upon successful authentication. Provider node(s) 126 may gather requested aspect(s) from digital twin storage 220, 922 storing digital twin data (e.g. data set(s) or aspect(s)) based on the decentral digital twin identifier and associated data set identifier(s). The data set(s) may be gathered upon successful contract negotiations. The peer-to-peer communication channel may be terminated and no data set(s) may be provided if no electronic contract is signed. The gathered data set(s) may be provided to consumer node 134. Consumer node 134 may forward the received data set(s) to neural network engine 410.

[0164] Neural network engine 410 may be configured to store the received data set(s) (e.g. the received aspect(s)) in a data storage (not shown). The data set(s) may be stored according to rule(s) stipulated by the signed electronic contract. With continued reference to FIG. 9, neural network engine 410 may be configured to transform the gathered data set(s). The gathered data set(s) may be transformed by search result compositor 508 to generate a uniform data set, for example as described in the context of FIG. 5, FIG. 7 and FIG. 10. The uniform data set generated by search result compositor 508 may be provided as search result data to consumer app 408. The uniform data set may signify a list of output product(s) or production input(s) fulfilling criteria defined by the user in the input data. Consumer app 408 may display the received uniform data set within a graphical user interface, for example as shown in FIG. 14B.

[0165] By using a neural network engine including data-driven model(s), in particular a trained data driven model(s), unstructured user input associated with a produced output product or a production input thereof may be transformed into structured query data required to query a decentral network storing such data. This allows to translate user input, such as unstructured data associated with or encoding a given context, into query data having a defined data structure matching the data to queried. The unstructured data and context are hence used by the trained model to determine defined data structures associated with such context and allowing to determine access elements matching the unstructured data and context of the user input. By using the neural network engine data set(s) gathered from various data providing nodes can be transformed into a uniform data set based on the generated query data. This allows to convert the data gathered from the decentral network based on the query data generated by the neural network engine into a data set containing only the data matching to or being related to the input data provided by the user hence avoiding provision of additional (e.g. not requested) data or provisioning of all gathered data set(s). The system allows to efficiently retrieve or gathered structured data associated with produced output products and production input(s) thereof from a decentral network based on unstructured data associated with or encoding a context, e.g. associated with or encoding a request for specific data on produced output product(s) or production input(s) thereof, without requiring the possession of physical entities of input materials and / or output products and / or without requiring knowledge on decentral identifiers associated with requested output product data and / or input material data.

[0166] FIG. 5 illustrates a block diagram of an example implementation of a neural network engine for generating query data to retrieve data associated with product(s) or input material(s) used to produce the product(s) from a decentral network and to generate response data from the retrieved data in accordance with an exemplary embodiment of the present invention. The neural network engine may correspond to neural network engine 410 described in the context of FIG. 4. The neural network engine may be configured to generate structured query data based on unstructured user input, for example as described in the context of FIG. 6, FIG. 10 and FIG. 11 . The user input may be associated with a context as described in relation to FIG. 4. The neural network engine may further be configured to transform the gathered or retrieved output product data (e.g. data associated with produced output product(s)) and / or production input data (e.g. data associated with input material(s) used to produce such output product(s)) to a uniform data set based on the query data, for example as described in the context of FIG. 7 and FIG. 10.

[0167] FIG. 5 illustrates example couplings, signals and data that are passed among various modules, such as query data generator 504, data manager 520 and search result compositor 508. FIG. 5 illustrates example couplings, signals and data that are passed among various modules, such as the query data generator 504, data manager 520, search result compositor 508 and human-in-the-Loop ML module 510. FIG. 5 illustrates example couplings, signals and data that are passed among various modules, such as the query data generator 504, data manager 520, search result compositor 508, vector database 506, embedding model 518 and human-in-the-Loop ML module 510. The neural network engine 410 advantageously learns to generalize and scale its various modules, such as query data generator 504 and search result compositor 508, from a particular domain to another by having at least a part of the modules pre-trained on the most general domain. The pre-trained models may be pre-trained, e.g. parametrized, using training data set containing unlabeled unstructured data of a general domain. The general domain may include fundamental linguistic, visual, world, and commonsense knowledge, which are domain-agnostic. These pre-trained modules may be fine-tuned or semantically conditioned to domain specific knowledge using domain specific training data, for example as described in the context of FIG. 8A and FIG. 8B. The domain specific training data may be structured labelled training data. In addition or alternatively, the accuracy of the pre-trained models may be improved by adding context specific data to the received search data, for example as described in the context of FIG. 11 .

[0168] Query data generator 504 may be coupled for communication and interaction with the consumer App 408, search result compositor 508 and the data manager 520. Query data generator 504 may further be coupled for communication and interaction with human-in-the-Loop ML module 510 and / or vector database 506. Query data generator 504 may receive search data, e.g. user input, from consumer App 408. Search data may include unstructured data. The unstructured data may be associated with a context as described in relation to FIG. 4. The unstructured data may be associated with produced output product(s) or production input(s) used to produce such output product(s). Query data generator 504 may be coupled to human-in-the-Loop ML module 510 to send requests for training data and / or to receive training data from human-in-the-Loop ML module 510. Query data generator 504 may be coupled to vector database 506 to receive nearest vector(s) stored in such vector database 506 based on embeddings of the user input. This may allow to provide additional context related to the user input to query data generator 504, hence improving the accuracy of the query data generated by query data generator 504 in response to a received user input while avoiding continuous fine-tuning of query data generator 504 by labelled training data set(s) to ensure accurate generation of query data if the data stored within the decentral network is updated (e.g. data for new product(s) and / or production input(s) is stored and / or stored data is amended). Query data generator 504 may parse input data associated with output product(s) and / or production input(s) thereof (e.g. unstructured user queries) and may convert such input data to query data (e.g. structured data including key(s), value(s) and / or key value pair(s) associated with output product(s) and / or production input(s) thereof).

[0169] Referring now also to FIG. 6, query data generator 504 may include a speech module recognition 604 and a neural semantic parser 608. Query data generator 504 may further include embedding model 606 and vector database 506. Any search data including audio signal(s) related to speech may be routed to the speech module recognition 604 which may process the audio signal(s) and may convert it to unstructured data signifying a text. The unstructured data may be provided to embedding model 606 or neural semantic parser 608.

[0170] Embedding model 606 may be configured to generate embeddings from the received search data. Generating embeddings may include generating vector(s) representing each sentence within the search data. The vector(s) may be generated by one hot encoding each word within each sentence and concatenating the one-hot vectors for each word. The vector(s) may be generated by encode each word using a unique number to generate dense vector(s). The vector(s) may represent dense vector(s) of floating point values. The values may represent trainable parameters (e.g. weights learned by embedding model 606 during training). The embedding model 606 may be different from embedding model 518. Embedding model 606 may be trained according to the method described in the context of FIG. 17. The embeddings generated by embedding model 1602 may be provided to vector database 506. Vector database 506 may be configured to conduct a "nearest neighbor" search to identifying stored vectors that most closely match the embeddings generated by embedding model 606. The nearest vector(s) may be provided as relevant context to the neural semantic parser 608. Use of additional context from a vector database in addition to input data received from a user as input for neural semantic parser 608 may allow to improve the accuracy of output (e.g. feature vector) of the neural semantic parser 608 by reducing hallucination. Since the embeddings stored in vector database 506 may be updated with much less effort than required to fine-tune neural semantic parser 608 with new labelled training data, use of such vector database 506 may allow to ensure a high accuracy of the output produced by trained neural semantic parser 608 without requiring constant fine-tuning of the trained neural semantic parser 608 on updated training data.

[0171] Neural semantic parser 608 may be configured to output query data based on search data received from consumer App 408. Neural semantic parser 608 may be configured to output query data based on search data received from consumer App 408 and nearest vector(s) received from vector database 506. Neural semantic parser 608 may include a transformer architecture. Neural semantic parser 608 may include an encoder (see FIG. 8A). The encoder may output a feature vector based on input (e.g. search data received from consumer App 408 and optionally nearest vector(s)). The feature vector may be used as input for a classifier. The classifier may output one or more classification(s) (e.g. the query data). The classifier may further output a confidence score tied to the outputted classification(s). The classifier may further output a confidence score per outputted classification. Neural semantic parser 608 may be trained according to the method described in the context of FIG. 8A. Referring back to FIG. 5, data manager 520 may be configured to manage data received from query data generator 504, such as query data, and / or data received from consumer node 134, such as decentral identifier(s) of output product(s), access element(s) associated with output product(s), decentral identifier(s) of production input(s) and / or access element(s) associated with such production input(s). Data manager 520 may be configured to generate request(s) based on the received data, for example as described in the context of FIG. 4 and FIG. 9. Data manager 520 may be configured to generate the request(s) based on a relationship represented by the data included in the generated query data. For instance, data manager 520 may be configured to determine the part of the query data, such as key-value pair(s), associated with the output product(s) and to generate a request to gather decentral identifier(s) associated with output product(s) based on such determined query data parts. Data manager 520 may parse the data received from consumer node 134 to generate request(s). For instance, data manager 520 may be configured to parse the access element data to determine access data for accessing respective output product data. Data manager 520 may be configured to initiate determination of decentral identifier(s) associated with input material(s) and gathering of input material data based on the received decentral identifier(s) associated with the output product(s) and the part of the query data associated with input material(s), such as key-value pair(s) associated with input material(s). Data manager 520 may be configured to link query data associated with input data with gathered product data / production input data received from consumer node 134. Data manager 520 may be configured to link query data associated with input data with gathered product data / production input data received from consumer node 134 and with nearest vector(s) received from vector database 506. Data manager 520 may be configured to generate input data for search result compositor 508. The input data may include the search data received from consumer App 408 and product data / input production data gathered by consumer node 134 based on the query data. The input data may include the search data, nearest vector(s) and product data / input production data gathered by consumer node 134 based on the query data. The input data may include the search data received from consumer App 408 and nearest vector(s).

[0172] Search result compositor 508 may be coupled for communication and interaction with consumer App 408 and storage 522. Search result compositor 508 may receive search data from consumer App 408. Search result compositor 508 may be configured to query storage 522 based on the received search data from consumer App 408. Search result compositor 508 may be coupled to human-in-the-Loop ML module 510 to send requests for training data and / or to receive training data from human-in-the- Loop ML module 510. Search result compositor 508 may be coupled to vector database 506 to receive nearest vector(s) stored in such vector database 506 based on embeddings of the user input. This may allow to provide additional context related to the user input to search result compositor 508 as previously described. Search result compositor 508 may generate query data to query storage 522 based on the input data received from consumer App 408 and may return the data received from storage 522 in a natural language format to consumer App 408.

[0173] Referring now also to FIG. 7, search result compositor 508 may include neural semantic converter 716. Search result compositor 508 may further include embedding model 606 and vector database 506. Embedding model 606 may be configured to generate embeddings from the received search data as previously described. The embeddings may be provided to vector database 506. Vector database 506 may be configured to conduct a "nearest neighbor" search to identifying stored vectors that most closely match the embeddings generated by embedding model 606. The nearest vector(s) may be provided as relevant context to the neural semantic converter 716 as previously described.

[0174] Neural semantic converter 716 may be configured to output output product data and / or production input data gathered from storage 522 based on database query data generated by neural semantic converter 716 in response to search data received from consumer App 408. This process may be triggered by consumer node 134 indicating that all data set(s) have been gathered and stored in storage 522. Neural semantic converter 716 may be configured to output uniform data based on search data received from consumer App 408, the nearest vector(s) received from vector database 506 and the product data and / or production input data gathered from storage 522. Neural semantic converter 716 may include a transformer architecture.

[0175] Neural semantic converter 716 may include a neural network trained to generate database query data to query the storage 522. The neural network may be trained to generate SQL query data based on the user input. The neural network may be trained to extract parameters from the input data to generate the database query data. For instance, the neural network may be trained to extract output product(s) and / or production input(s) and / or properties thereof and / or restrictions on such properties from the input data. The neural network may be trained to generate to generate the output by sequentially generating query data to query the contents of storage 522. For instance, the neural network may be trained to generate database query data based on data received from storage 522 in response to previously generated query data. The neural network may be trained to generate query data to query storage 522 for all tables contained therein. Upon receiving the tables, the neural network may be configured to generate query data to query the schema and example rows of tables included in storage 522. The neural network may be trained to determine the tables for which the schema is to be queried based on the input data. Neural semantic converter 716 may include the structure illustrated in FIG. 8B.

[0176] Referring back to FIG. 5, the human-in-the-Loop ML module 510 may be coupled to provide teaching actions and / or instances to the neural network engine 410. Human-in-the-Loop ML module 510 is particularly advantageous because it implements a holistic human-in-the-loop machine learning paradigm, including both machine teaching and active learning. The machine teaching Al paradigm combines the power of an intelligent human in the loop, as the teacher, with an Al system that learns to improve over time through efficiently interacting with its teacher. The teacher in the loop has some basic understanding of the capabilities and prior learnings of the model that it is interacting with and is meant to provide some minimal supervision over the work of the neural network engine 410, as well as provide feedback on errors and mistakes. Through this efficient human-in-the-loop paradigm, the neural network engine 410 may be fine-tuned on specific domains with minimal training instances and in a short period of time. The human-in-the-Loop ML module 510 may be coupled to receive labelled training data (e.g. training data annotated by humans) from any number of humans that have different roles and expertise. For example, those roles may include an average user, a crowd worker, trained crowd worker, domain expert, a model expert, or various other human users. Any one of these humans may be considered teachers as it has been described above. The human-in-the-Loop ML module 510 may uses various statistical algorithms for vetting and training data set(s) as they get collected, for further ensuring the quality and accuracy. The human-in-the-Loop ML module 510 may allow supervised learning, semisupervised learning, or unsupervised learning for building, training and re-training the data-driven model(s) included in neural network engine 410 based on the type of data available and the particular machine learning technology used for implementation. The human-in-the-Loop ML module 510 may include various other components such as a deployment module, an evaluation module, a generalization module, a collection module and an instantiation module to implement the process described below for continually improving the operation and accuracy of the neural network engine 410. The human-in-the-Loop ML module 510 is particularly advantageous because it provides the ability to take human feedback into account to improve the operation of the neural network engine 410. In some implementations, the human-in-the-Loop ML module 510 may be used to control behavior by taking feedback into account and have guarantees for generating (or not generating) a particular output given a particular input.

[0177] Human-in-the-Loop ML module 510 may train a ML model and may deploy that model in the real world for evaluation. Training may be done, for example, as described in the context of FIG. 8A and FIG. 8B. For example, the Al model may be deployed as part of the query data generator 504 or search result compositor 508. As the Al model operates in a specified setting, the human-in-the-Loop ML module 510 may evaluate the model’s accuracy and collects failure cases. Based on the evaluation and collected failure cases, the human-in-the-Loop ML module 510 may determine a teaching set. These teaching sets may include labelled data for training actions. Using the teaching set(s), the human-in-the-Loop ML module 510 may instantiate specific teaching actions and / or instances and may provide them to neural network engine 410 for training.

[0178] Embedding model 518 may be coupled to provide embeddings to the vector database 506. The embeddings may be generated by embedding model 518 from data associated with access elements. Data associated with access elements may include structured data associated with output product(s) or production input(s). The structured data may include key value pairs. The key-value pairs may be associated with access data. The access data may be associated with a given data set included in a digital twin, for example as described in the context of FIG. 2. The key value pairs may be associated with data related to the output product or the production input thereof. Data related to the output product may include an output product type. Data related to production input(s) used to produce the output product may include a production input type. Data associated with access elements may include standardization documents defining key value pairs of access elements for a given output product type. The embeddings may be generated from data associated with access elements as previously described. The embedding model 518 may be trained as described in the context of FIG. 17. Storage 522 may be coupled to consumer node 134. Storage 522 may be configured to store output product data set(s) and / or production input data set(s) gathered by consumer node 134. This may allow to collect the data gathered by consumer node 134 since data may not be provided to consumer node 134 at a single time point but rather during a period of time. Storage 522 may be coupled to search result compositor 508. Storage 522 may be queried by search result compositor 508 based on input data received from consumer App 408, for example as described in the context of FIG. 7 and FIG. 8B.

[0179] FIG. 8A illustrates an example of a training process of the data query generator included in the neural network engine in accordance with an exemplary embodiment of the present invention. The query data generator may be query data generator 504 described in the context of FIG. 5 and FIG. 6. The query data generator may include one or more neural network(s).

[0180] The neural network may be based on an encoder - classifier architecture. The neural network may include at least one encoder and at least one classifier. Encoder and classifier may be implemented on System on a Chip (SoC) Integrated Circuit (IC). The query data generator 504 unit may be configured to train the encoder and / or the classifier and / or execute the trained encoder and / or the trained classifier. The query data generator 504 may include a GPU-enabled computer processor. Training the neural network on the GPU-enabled computer processor may output weights or kernels that are described using floating-point numbers. In such embodiments, the floating-point operating parameters may be converted to integer number representations to be used by query data generator 504.

[0181] In the neural network, nodes are connected to one another via one or more edges. A neural network can include an input layer, an output layer, and one or more intermediate layers. Individual nodes can process their respective inputs according to a predefined function, and provide an output to a subsequent layer, or, in some cases, a previous layer. The inputs to a given node can be multiplied by a corresponding weight value for an edge between the input and the node. In addition, nodes can have individual bias values that are also used to produce outputs. Various training procedures can be applied to learn the edge weights and / or bias values. The term “parameters” when used without a modifier is used herein to refer to learnable values such as edge weights and bias values that can be learned by training a machine learning model, such as a neural network.

[0182] The neural network structure can have different layers that perform different specific functions. For example, one or more layers of nodes can collectively perform a specific operation, such as pooling, encoding, or convolution operations. For the purposes of this document, the term “layer” refers to a group of nodes that share inputs and outputs, e.g., to or from external sources or other layers in the network. The term “operation” refers to a function that can be performed by one or more layers of nodes. The term “model structure” refers to an overall architecture of a layered model, including the number of layers, the connectivity of the layers, and the type of operations performed by individual layers. The term “neural network structure” refers to the model structure of a neural network. The term “trained model” and / or “tuned model” refers to a model structure together with parameters for the model structure that have been trained or tuned. Note that two trained models can share the same model structure and yet have different values for the parameters, e.g., if the two models are trained on different training data or if there are underlying stochastic processes in the training process.

[0183] The encoder and / or classifier may include at least one convolutional neural network (CNN). The encoder and / or classifier may include other types of networks or functions such as FFT, wavelets, deep learning, like CNN, energy models, normalizing flows, a recurrent neural network (RNN), a gated recurrent unit (GRU), a long short-term memory (LSTM) recurrent neural network, vision transformers, or transformers used for natural language processing, autoregressive image modelling, Autoregressive Image Modeling, Normalizing Flows, Deep Autoencoders, Deep Energy-Based Models, Vision Transformers. Supervised or unsupervised schemes may be applicable to generate representations. The encoder and classifier may be based on the same or different networks or functions. The encoder architecture may differ from the classifier architecture. The encoder and classifier may have the same architecture.

[0184] The training process may include providing unstructured data associated with a defined context as input. The given context may be associated with or relate to output product(s) and / or production input(s) used to produce such output product(s). The given context may be associated with or relate to output product data and / or production input data. The output product data and / or the production input data may correspond to data on output product(s) and / or production input(s) to be retrieved from a decentral network. The input may include labelled unstructured data including metadata associated with search query data. The metadata may include a feature vector generated by the encoder based on unstructured input data associated with the given context. The metadata may include nearest feature vector(s) obtained from a vector database based on embeddings of the input data as described in the context of FIG. 4 to FIG. 7. The unstructured data may correspond to sentences. Each sentence may include one or more label(s). The labels may signify aspect(s) associated with or related to output product(s) and / or production input(s). The labels may signify aspect(s) corresponding to or being associated with output product(s) or production input(s). The labels may signify aspect(s) corresponding to or being associated with output product data or production input data. In addition, the label(s) may signify search query data associated with or related to the respective aspect. For instance, the label(s) may signify aspect(s) and associated search query data. The query data may include key(s), value(s) and / or key value pair(s) associated with or included in access element(s) associated with output product(s) and / or production input(s). Input may include a plurality of training data set(s) associated with different output products and / or production input(s). The output product(s) and / or production input(s) associated with the training data set(s) may be associated with varying search query data, such as key(s), value(s) and / or key value pair(s) associated with or included in access element(s) associated with the different output product(s) and / or production input(s). The query data associated with the access element(s) for output product(s) and / or production input(s) associated with the training data set(s) may be known. The known query data and known aspect(s) may be provided to the training process as label(s). For training the input may be provided to the aspect generation model 806. The aspect generation model 806, such as an encoder of, may define aspects in the input as feature vectors in a feature space. For example, the encoder may define aspects related to output product(s) and / or production input(s) as feature vectors in the feature space. The feature vectors may include feature vectors representing the aspect(s) present in the input in a feature space. A feature space may be an N- dimensional feature space. A feature vector may be an n-dimensional vector of numerical values that define features of the aspect for a region in the feature space that corresponds to a region in the input. The number of dimensions may depend on the embodiment. For example, in the case of aspects, the feature vectors may refer to representations of the aspect(s) present within the unstructured data, such as one or more sentence(s) or question(s).

[0185] The feature vectors generated by aspect generation model 806 may be provided to the classifier 810. The classifier 810 may generate a classification using the feature vectors. Classification of feature vectors may include mapping the feature vectors to one or more classifiers by a classification network to determine output data from the input.

[0186] Output generated by the classifier may include query data, such as key value pair(s), related to or associated with the input. For training, the query data and aspect(s) of the input may be known and may be correlated with the output produced by the classifier. As the classifier classifies feature vectors received from aspect generation model 806, the classifier may provide one or more classifications for aspect(s) included in the input based on the correlation between known query data and feature vectors. The classification outputted by the classifier 810 may include key value pair(s) associated with access element(s) of output product(s) or production input(s). The key value pair(s) may allow to query decentral registries storing such access element(s) for access element(s) and / or decentral identifier(s) associated with access element(s) matching the key value pair(s).

[0187] By training, the query data associated with the context of the input may be determined by correlating classification(s) with metadata or labels. For example, metadata may provide known embeddings of aspects present within the unstructured data defining the query data determined by the classification process. Correlating classification(s) with metadata or label(s) may include the classifier assessing differences between classification(s) and metadata or label(s). The neural network may thus perform error function analysis on the differences between the classification(s) for given metadata or label(s) and refine the classification process until the classification process accurately determines the metadata or the label(s). Thus, the query data generator 504 may be trained by the labelled training data to accurately determine query data.

[0188] The operation on the input by the trained query data generator 504 may include classification to determine query data or any other process involving a neural network module operating on unstructured data to classify the unstructured data.

[0189] FIG. 8B illustrates an example of a neural network system and a training process thereof included in the search result compositor of the neural network engine in accordance with an exemplary embodiment of the present invention. The search result compositor may be search result compositor 508 described in the context of FIG. 5 and FIG. 7. The search result compositor may include one or more neural network(s).

[0190] The system may include an Agent 836 including a transformer 842 for generating database query data and output 828, a tool list for querying storage 522 and output parser 852 to monitor whether the response generated by transformer 842 is a final response to be provided as output 828 or an intermediate response requiring further actions by engine 840.

[0191] The neural network may be based on an encoder - decoder architecture. The neural network may include at least one encoder and at least one decoder. Encoder and decoder may be implemented on SoC structures as described in FIG. 8A. A neural network processing unit may be configured to train the encoder and / or decoder and / or execute the trained encoder and / or trained decoder. The neural network processing units may include a GPU-enabled computer processor. Training neural network on the GPU-enabled computer processor may output weights or kernels that are described using floating-point numbers. In such embodiments, the floating-point operating parameters may be converted to integer number representations to be used on neural network processing unit.

[0192] The encoder and / or decoder may include at least one convolutional neural network (CNN). The encoder and / or decoder may include other types of networks or functions such as FFT, wavelets, deep learning, like CNN, energy models, normalizing flows, a recurrent neural network (RNN), a gated recurrent unit (GRU), a long short-term memory (LSTM) recurrent neural network, vision transformers, or transformers used for natural language processing, autoregressive image modelling, Autoregressive Image Modeling, Normalizing Flows, Deep Autoencoders, Deep Energy-Based Models, Vision Transformers. Supervised or unsupervised schemes may be applicable to generate representation. The encoder and the decoder may be based on the same or different networks or functions. The encoder architecture may differ to the decoder architecture. The encoder and decoder may have the same architecture.

[0193] The training process of the encoder - decoder architecture may include providing a collection of unstructured data (e.g. unstructured text data) and corresponding data transformation scheme. The data transformation scheme may include database queries (e.g. query data). The transformation scheme may further include database table(s) associated with the database queries. The queries may be based on SQL query language. The database tables may be SQL. The database queries may be based on NoSQL query language.

[0194] Input may include a plurality of training data sets with different unstructured text data. The unstructured text data may include various unstructured text data types. The unstructured text data types may include unstructured data signifying key words, such as greater than, lower than, equal, less than, etc.

[0195] For training the input may be provided to the encoder. The encoder may define features in the unstructured data as feature vectors in a feature space. The feature vectors may include feature vectors representing the unstructured data features in a feature space. A feature space may be an N- dimensional feature space. A feature vector may be an n-dimensional vector of numerical values that define features of the unstructured data. The number of dimensions may depend on the embodiment.

[0196] The feature vectors generated by the encoder may be provided to decoder. The decoder may decode the representations or feature vectors. The decoder may include a neural network, such as a recurrent neural network (RNN), a gated recurrent unit (GRU), a long short-term memory (LSTM) recurrent neural network or any architecture referred to above. Decoding the representations or feature vectors may include mapping the feature vectors via a regression network to determine (e.g., extract) query data from the input. Query data generated by the decoder may include information of the input. For training, the query data matching the question may be known and may be correlated with the decoded feature vectors. As decoder operates on feature vectors in the feature space, the decoder may provide one or more predictions for the database query data based on the correlation between known database query data and feature vectors.

[0197] By training, database query data to perform a database query storing the gathered output product data set(s) and / or the production input data set(s) may be determined by correlating decoded feature vectors with known database query data. Correlating decoded feature vectors with known database query data may include the decoder assessing differences between decoded feature vectors and the database query data. The encoder-decoder architecture may thus perform error function analysis on the differences between the decoded feature vectors for given database query data and refine the feature vector decoding process until the feature vector decoding process accurately determines the database query data. Thus, the decoder-encoder architecture may be trained by the training data to accurately determine database query data to query storage 522.

[0198] The prompt template 838 may be used to generate a prompt provided to transformer 842 based on search data received from consumer App 408. The prompt template 838 may refer to the blueprint of the prompt for which an output should be generated. The prompt template may include a static part and a dynamic part. The static part may define the tools that may be used to generate an answer based on the input data. The static part may include instructions and / or information for transformer 842 in handling the received input. Instructions may include the required output data and / or required output data format and / or limitations concerning the use of tools and the construction of the final answer. Information may include a description of tool(s) and / or required input to tool(s) and / or error handling procedures. Use of such static part may allow the transformer 842 to decide on which tool(s) to use to determine the output from the received input (e.g. structured data generated based on the prompt template 838 from the received search data from consumer App 408). Use of the static part may allow the transformer 842 to determine which subsequent tool to use upon receiving a response from a previously selected tool. Hence, the static part may allow transformer 842 to determine the sequence in which tools need to be used (e.g. a chain of actions to be performed or a chain of tools to be called). For instance, the transformer 842 may determine the sequence to call tools 844, e.g. to generate input data for such tools, from the received input. Tools 844 may be associated with or related to function(s) that are executed based on output generated by transformer 842. Function(s) may include function(s) used to query storage 522. Function(s) may include function(s) to query the database and / or to query the database scheme and / or the database tables.

[0199] Output parser 852 may be configured to analyze each output the language model generated by using a tool and to decide whether the next tool is to be executed or whether a final answer has been generated. The output parser 852 may further be configured to parse the output of transformer 842 into a final format. The final format may be defined by prompt template 838.

[0200] Use of Agent 836 may allow to provide an answer to the user query (e.g. the unstructured data received from consumer App 408) by gathering data from storage 522 based on the received user query. This may allow to tailor the output provided to the user to the input received from the user. Moreover, this may allow to tailor the provided data associated with the output product(s) and / or production input(s) to the search received from the user.

[0201] FIG. 10 illustrates a flow chart of an example method for retrieving or gathering data associated with output product(s) and / or input material(s) used to produce the output product(s) from a decentral network in accordance with an exemplary embodiment of the present invention. The method may be implemented by a consumer environment of the decentral network, such as data consumer environment 402 described in the context of FIG. 4 and FIG. 9. The method may be implemented by neural network engine 410 including query data generator 504 and search result compositor 508.

[0202] The output product may be an output product having been produced by a production, for example as described in the context of FIG. 2. The output product may be produced from one or more production input(s). Production input(s) may refer to input material(s) used within at least one production step associated with the production of the output product. Production input(s) used by one participant of the product ecosystem (see FIG. 1) may correspond to output product(s) produced by an upstream participant of the product ecosystem (see FIG. 1). The output product may be a chemical product as described in the context of FIG. 2. The output product may be a discrete product as described in the context of FIG. 2.

[0203] With reference to FIG. 5 to FIG. 7 and FIG. 13, search data from a user indicating a search for data associated with produced output products or production input(s) thereof may be received (see block 1002). The search data may be received via consumer app 408. The user may input the search data in a graphical user interface, such as graphical user interface 1402 displayed in FIG. 14A. The graphical user interface 1402 may be display by consumer app 408. The graphical user interface 1402 may be displayed by the frontend of consumer app 408. The input data 1406 may be typed by the user in a text field displayed within graphical user interface 1402. Upon detecting a user input on the send button 1404, the search data 1406 entered by the user may be send by consumer app 408 to neural network engine 410. The search data may be sent via a backend server of consumer app 408 to neural network engine 410. The search data may include unstructured data. The unstructured data may be associated with a context. The context may be associated with the output product(s) and / or the production input(s) thereof. The context may be associated with a property of the output product or the production input(s) thereof. The context may encode the data associated with the output product and / or production input(s) to be gathered or retrieved from the decentral network. In addition or alternatively, the context may encode property data of the output product(s) or production input(s) to be retrieved or gathered from the decentral network.

[0204] With continued reference to FIG. 5 to FIG. 7 and FIG. 13, query data may be generated by inputting the received search data into neural network engine 410 (see block 1004). The neural network engine 410 may include query data generator 504. A data-driven model may be instantiated and executed on query data generator 504. In particular, the received search data may be provided to query data generator 504 of neural network engine 410 for executing a data-driven model configured to generate query data.

[0205] The data driven-model may be trained to generate query data from unstructured data. The data-driven model may be trained to generate query data from unstructured data associated with a context. The data-driven model to be instantiated and executed on query data generator 504 is described in more detail in the context of FIG. 8A. The data-driven model may be parametrized according to a training data set including search data and associated label(s), for example as described in the context of FIG. 8A. The label(s) may relate to aspect(s) and associated key value pairs associated with or included in access elements stored in decentral registries of the decentral network. The key value pairs may have a defined semantic structure. The defined semantic structure may be defined by the data included in the access elements. The key value pairs may be associated with output product types, manufacturer types, production input types, output product type properties or production input product type properties. The training data set may include search data and an associated label(s). The trained data-driven model may generate one or more classifier(s) discriminating between key value pairs of output product types, key value pairs of manufacturer types, key value pairs of production input types, key value pairs of output product type properties or key value pairs of production input product type properties. The trained data driven model may generate an output product classifier discriminating between key value pairs of output product types, and / or a manufacturer classifier discriminating between key value pairs of manufacturer types and / or a production input classifier discriminating between key value pairs of production input types and / or an output product property classifier discriminating between key value pairs of output product type properties and / or a production input type property identifier discriminating between key value pairs of production input product types properties. The trained data-driven model may hence allow to transform search data including unstructured data associated with a context into structured data used within the decentral network and being associated with the same context.

[0206] A compiler executed by a CPU present within query data generator 504 may determine a task list for the query data generator 504. The data-driven model to be executed may include one or more neural network(s). The neural network(s) may include network layers or sub-layers that are instantiated or implemented as a series of tasks executed by the query data generator 504. For example, the data- driven model may include network layers (or sub-layers) including feed forward layers and normalization layers. The neural network(s) may be instantiated by the query data generator 504. To do so the neural network(s) may be converted into a task list to become executable by the query data generator 504. The neural network(s) may be converted by the CPU to the task list. The task list includes a linear link-list defining a sequence of tasks including tasks per feed forward layer and / or normalization layer. Each task may be associated with a task descriptor that defines a configuration of the query data generator 504 to execute the task. Each task may correspond with a single network layer of the neural network(s), a portion of a network layer of the neural network(s), or multiple network layers of the neural network(s). Based on the task list generated by the CPU and provided to the query data generator 504, the query data generator 504 instantiates the neural network(s) by executing the tasks of the task list under the control of a neural task manager.

[0207] The neural task manager may receive a task list from a compiler executed by the CPU, store tasks in its task queues, choose a task to perform, and send instructions to other components of the neural processor circuit for performing the chosen task. The neural task manager may include one or more task queues. Each task queue may be coupled to the CPU and the task arbiter. Each task queue may receive from the CPU a reference to a task list of tasks that when executed by the neural processor circuit instantiates the neural network(s). The reference stored in each task queue may include a set of pointers and counters pointing to the task list of the task descriptors in the system memory. Each task queue may be further associated with a priority parameter that defines the relative priority of the task queues. The task descriptor of a task may specify a configuration of the neural processor circuit for executing the task.

[0208] Decentral identifier(s) associated with the one or more produced product(s) may be gathered via the decentral network based on the generated query data (see block 1006). Hence, gathering of the decentral identifier(s) may not require the possession of physical entities of input materials and / or output products comprising physical identification elements encoding identifiers associated with such decentral identifiers. The decentral identifier(s) may be required to gather the associated output product data and / or input material data (e.g. without knowledge of the decentral identifier(s), no output product data and / or input material data can be gathered via the decentral network). The query data may be used to query the decentral network for decentral identifier(s) associated with data, in particular with access elements, matching the query data. With continued reference to FIG. 13 neural network engine 410 may be configured to determine end point(s) associated with provider node(s) 126 based on the query data, for example as described in the context of FIG. 4 and FIG. 9. Based on the determined end point(s), neural network engine 410 may generate a request to gather decentral identifier(s). The request may include the query data and gathered endpoint(s). The request may be provided to consumer node 134. Consumer node 134 may request, based on the received request, decentral identifier(s) from provider node(s) 126 associated with endpoint(s) included in the data received from neural network engine 410, for example as described in the context of FIG. 4 and FIG. 9. Authentication may be performed between consumer node 134 and provider node(s) 126, for example as described in the context of FIG. 4 and FIG. 9. Provider node(s) 126 may query, based on the received query data, associated decentral registry 218 for decentral identifier(s) matching the query data. Matching of the query data to access element data may include determining similarity score(s) between the query data and the access element data. The matching may be performed by provider nodes receiving a request to provide decentral identifier(s) based on query data from a consumer node. Decentral identifier(s) matching the query data may be provided by respective provider node(s) 126 to the requesting consumer node 134, for example as described in the context of FIG. 4 and FIG. 9. Decentral identifier(s) may be provided from consumer node 134 to data manager 520.

[0209] With continued reference to FIG. 13, access element(s) associated with said gathered decentral identifier(s) may be gathered, for example as described in the context of FIG. 4 and FIG. 9. The access elements may include access data indicating the decentral data providing node(s) storing the output product data and / or input production data related to or matching the input data received in block 1002. The gathered access elements may be provided from consumer node 134 to neural network engine 410.

[0210] With reference to FIG. 12, it may be determined whether output product data is to be gathered. For instance, data manager 520 may be configured to determine whether to gather output product data based on the received query data and the access element(s). Data manager 520 may parse the received access element data and match the query data to the parsed access element data to determine whether output product data need(s) to be gathered. For instance, data manager 520 may parse the access element data to determine whether access data matching the query data is contained in the access element data. If no output product data is to be gathered, data manager 520 may initiate gathering of input material data based on the decentral identifier(s) of the output product data. The input material data may be gathered, for example, as illustrated in FIG. 15 and FIG. 16. If output product data is to be gathered, data manager 520 may generate request(s) to gather the output product data based on the gathered access elements. Output product data may include such data set(s) associated with access data matching the query data. With continued reference to FIG. 13, model unit 410 may be configured to parse the access element data to determine the appropriate access data based on the query data. The access element data may be matched to the query data to determine the access data associated with or related to the user input. For instance, the access element data may be parsed to determine access data for environmental attribute data by matching the query data including environmental attribute data with a key-value pair included in the access element data and signifying environmental attribute data. The appropriate access data may include access data of data set(s) matching the query data. Matching the query data may include determining similarity score(s) between the query data and the access element data as previously described. Neural network engine 410 may generate a request to gather data associated with output product(s). The request may include the determined access data associated with or related to the user input. The access data may include a pointer to the provider node(s) 126 associated with DT storages storing the output product data and decentral identifier(s). The decentral identifier(s) may be associated with the output product(s). The decentral identifier(s) may be associated with a data set signifying the output product data. The request may be provided to consumer node 134. With continued reference to FIG. 13, consumer node 134 may gather output product data (e.g. data set(s)) based on the request received from neural network engine 410. Consumer node 134 may gather the output product data as described in the context of FIG. 4 and FIG. 9. Consumer node 134 may request access to output product data at respective provider node(s) 126. The request may include the decentral identifier(s). The respective provider node(s) may be determined by consumer node 134 based on the access data included in the gathered access element(s). Authentication may be performed between consumer node 134 and provider node(s) 126 as described in the context of FIG. 4 and FIG. 9. Provider node 126 may initiate contract negotiations with consumer node 134 as described in the context of FIG. 4 and FIG. 9. Contract negations may be initiated upon successful authentication. Authorization and authentication may be based on the decentral identifier(s) and decentral participant identifier(s) related to the consumer node 134. This way, access to the data set(s) may be controlled by the data owner of the data set(s), hence avoiding unauthorized access to such data set(s) by a user requesting data from the decentral network. Respective provider node(s) 126 may gather data set(s) (e.g. asset(s)) from associated databases, such as database 220, based on data provided by consumer node 134, e.g. based on decentral identifier(s) provided by consumer node 134. Provider node(s) 124 may query associated databases, such as database 220, based on the decentral identifier(s) received from consumer node 134. Data set(s) matching the decentral identifier(s) may be gathered by provider node(s) 124. Authorization rule(s) associated with such data set(s) may be applied prior to providing such data set(s) to consumer node 134. Authorization rule(s) associated with such data set(s) may be provided along with such data set(s) to consumer node 134. Consumer node 134 may provide the received data set(s) to neural network engine 410. Consumer node 134 may store gathered data set(s) in a data storage (see FIG. 5).

[0211] Referring back to FIG. 12, it may be determined whether production input data may be gathered. Data manager 520 may be configured to determine whether production input data may be gathered based on the query data. For instance data manager 520 may parse the query data to determine whether the query data includes key-value pair(s) associated with production input(s). If production input data may be gathered, data manager 520 may be configured to initiate gathering of production input data. Production input data associated with production input(s) of the output product(s) may be gathered based on gathered decentral identifier(s) of the output product via the decentral network. Production input data may be gathered based on the query data. For instance, this step may be performed if the query data includes key value pairs associated with production input types and / or production input product type properties. In another instance, this step may be performed if the query data includes a production input classifier discriminating between key value pairs of production input types and / or a production input type property identifier discriminating between key value pairs of production input product type properties. The decentral identifier(s) associated with the production input(s) may be gathered via a node configured to resolve relationship(s) between output product(s) and production input(s) used to produce such output product(s). The decentral identifier(s) may be gathered via relationship node 1504 as illustrated in FIG. 15. Relationship node 1504 may be part of infrastructure environment 406 (see FIG. 4), e.g. may not be associated with a participant of the product ecosystem (see FIG. 1). Relationship node 1504 may be part of a decentral data provider environment, such as data provider environment 1 404, data provider environment 2 906 (see FIG. 4, FIG. 9). For instance, relationship node 1504 may be part of the decentral data provider environment associated with the producer of the output product.

[0212] With reference to FIG. 15 and FIG. 16, neural network engine 410 may be configured to request decentral identifier(s) associated with production input(s) based on gathered decentral identifier(s) associated with such output product(s). The request may include decentral identifier(s) associated with the output product(s) (afterwards denoted as parent identifier(s)). The request may further include authentication data and / or a decentral participant identifier associated with the decentral consumer environment neural network engine 410 is a part of (see FIG. 4, FIG. 9). Authentication data may be obtained by neural network engine 410 from decentral IAM network node 1 1502. Neural network engine 410 may request authentication data from decentral IAM network node 1 1502. The request for authentication data may include credentials associated with neural network engine 410. Credentials associated with neural network engine 410 may include credentials associated with the participant entity operating the decentral network environment where neural network engine 410 is a part of. Credentials may include certificate(s). Credentials may include verifiable credentials as previously mentioned. In response to the received request, decentral IAM network node 1 1502 may verify the data provided by neural network engine 410 and may issue authentication data, such as an access token, if the data provided by neural network engine 410 is verified.

[0213] In response to the request, relationship node 1504 may gather decentral identifier(s) associated with production input(s) and production input data, for example as described in the context of FIG. 15 and FIG. 16. The gathered production input data may be provided by relationship node 1504 to neural network engine 410.

[0214] With continued reference to FIG. 13, neural network engine 410 may be configured to provide a uniform data sets as output to consumer app 408. Neural network engine 410 may be configured to gather data based on received input data from a data storage used by consumer node 134 to store gathered data. The data may be gathered from the data storage by generating query data based on the received input data. The data may be gathered as described in the context of FIG. 8B. In particular, the data set(s) may be gathered by search result compositor 508 of neural network engine 410 by executing a data-driven model configured to generate gather data from a data storage based on received user input.

[0215] The data driven-model may be trained to gather data from data storage 522 based on input data received from consumer app 408. The data-driven model to be instantiated and executed on neural semantic converter 716 is described in more detail in the context of FIG. 7 and FIG. 8B. The data- driven model may be trained as described in the context of FIG. 8B. The data provided as output by neural semantic converter 716 to consumer app 408 may signify transformed data since the data provided as output may resemble a uniform data set generated from multiple data set(s) gathered by consumer node 134. With continued reference to FIG. 13, the transformed data may be provided by neural network engine 410 to consumer app 408. Consumer app 408 may be configured to display the received transformed data, for example within a graphical user interface 1402. The transformed data 1410 may be displayed below the input data 1406.

[0216] By using a data-driven model, in particular a trained data driven model, user input associated with a produced output product or a production input thereof may be readily transformed into query data allowing to query a decentral network storing such data. This allows to translate user input, such as unstructured data associated with or encoding a given context, into query data allowing to query the decentral network for data desired by the user. Hence, the user may not need to be familiar with highly defined and particular query data to be used to query the decentral network but may be able to formulate the query in natural language. In addition, the user may not require knowledge of the decentral identifier(s) or may have to be in the possession of physical entities of input materials and / or output products comprising identification elements encoding identifiers allowing discovery of the decentral identifiers within the decentral network. This may allow users to gather data on output product(s) and / or production input(s) in a reliable and efficient way without violating the data sovereignty of the data owners of such data irrespective of the semantics required by the decentral network and the decentral identifier(s) used within the decentral network to retrieve such data. Such efficient and reliable data retrieval allows to use such gathered data to determine suitable recycling operation(s) to be performed and / or to determine properties of output product(s) currently used by end product user(s). Such determined properties may be used to predict the amount and / or type of end-of-life product(s) to be recycled in the future.

[0217] By using a data-driven model, in particular a trained data-driven model, data set(s) gathered from various data providing nodes can be transformed into a uniform output data based on the user input. This allows to convert the data set(s) gathered from the decentral network based on the query data generated by neural network engine 410 in response to a received user input into a single data set containing only the data requested by the user hence avoiding provision of additional (e.g. not requested) data. The system allows to efficiently retrieve structured data associated with produced output products and production input(s) thereof from a decentral network based on unstructured data associated with or encoding a context, e.g. with associated with or encoding a request for specific data on produced output product(s) or production input(s) thereof.

[0218] FIG. 11 illustrates a flow chart of a further example method for retrieving data associated with product(s) and / or input material(s) used to produce the product(s) from a decentral network in accordance with an exemplary embodiment of the present invention. The method may be implemented by a consumer environment of the decentral network, such as data consumer environment 402 illustrated in FIG. 4 to FIG. 9. The method may be implemented by the neural network engine 410, such as described in the context of FIG. 8A and FIG. 8B.

[0219] With reference to FIG. 6, FIG. 7 and FIG. 13, search data from a user indicating a search for data associated with produced output products or production input(s) thereof may be received. The input data may be received via consumer App 408. The user may input the search data in a graphical user interface, such as graphical user interface 1402 displayed in FIG. 14A. The graphical user interface 1402 may be displayed by consumer App 408. The graphical user interface 1402 may be displayed by the frontend of consumer App 408. The search data 1406 may be typed by the user in a text field displayed within graphical user interface 1402. Upon detecting a user input on the send button 1404, the search data 1406 entered by the user may be send by consumer App 408 to neural network engine 410 and vector database 506. The search data may be sent via a backend server of consumer App 408 to neural network engine 410 and vector database 506.

[0220] The search data may include unstructured data. The unstructured data may be associated with a context. The context may be associated with the produced output product or the production input(s) thereof. The context may be associated with a property of the produced output product or the production input(s) thereof. The context may encode the data associated with the produced output product and / or production input(s) to be gathered or retrieved from the decentral network. The context may encode the property / properties of the produced output product(s) or production input(s) to be retrieved or gathered from the decentral network.

[0221] With reference to FIG. 6, embeddings of the search data may be generated by providing the search data to an embedding model 606. The embedding model 606 may be part of neural network engine 410. The embedding model may be part of neural semantic parser 608. The embedding model 606 may be trained as described in the context of FIG. 17.

[0222] With continued reference to FIG. 5 and FIG. 6, a vector database 506 may be provided. The vector database 506 may store embeddings of data associated with access elements. The embeddings may be generated as described in the context of FIG. 6.

[0223] With continued reference to FIG. 5 and FIG. 6, embeddings from the vector database 506 may be gathered based on the generated embeddings. The embeddings may be gathered from the vector database 506 using a nearest neighbor search. The embeddings gathered from the vector database 506 may correspond to vectors that most closely match the input data (e.g. the context included in the input data).

[0224] Query data may be generated by converting the received input data and the gathered embeddings by neural network engine 410, for example as described in the context of FIG. 10 and FIG. 13. Afterwards, the method may proceed to block 1006 of FIG. 10 (e.g. the method may include further blocks described in FIG. 10).

[0225] By adding semantic context to the search data received from the user, the accuracy of the classification generated by query data generator 504 of neural network engine 410 may be improved, hence allowing to provide query data more accurately matching the intent of the user encoded in the search data. More accurate query data may allow to retrieve or gather output product data and / or input production data more accurately matching the intent of the user without requiring the user to know identifiers allowing discovery of the decentral identifier(s) within the decentral network or the decentral identifiers, hence allowing more accurate monitoring and / or controlling of the production of further product(s), the recycling and / or re-use of the output product(s) and / or the monitoring of the environmental impact of the input materials, the output product and / or the production based on the generated response.

[0226] FIG. 15 illustrates an example system for gathering production input data based on a decentral identifier(s) associated with output product(s) produced from such production input(s). The system of FIG. 15 may be used to gather access elements of such production input(s) as described in the context of FIG. 16. The output product may be produced using a production chain. The production chain may include one or more production step(s). Production of the output product may hence include production of intermediate products. Production input(s) may include any input material(s) used within one or more of the production step(s).

[0227] With reference to FIG. 10, neural network engine 410 may be configured to provide a request for production input data to relationship node 1504. The request may include decentral identifier(s) associated with the output product(s) (denoted as parent identifier(s) hereinafter). The request may further include authentication data, such as an access token. The request may further include a decentral participant identifier associated with the entity operating the decentral environment including neural network engine 410.

[0228] In response to receiving the request from neural network engine 410, relationship node 1504 may be configured to verify the authentication data included in the received request with decentral IAM network node 1 1502. Relationship node 1504 may be configured to determine the decentral data providing network node (e.g. provider node) associated with the provided decentral identifier (e.g. associated with the decentral registry storing the access element including the decentral identifier). With reference to FIG. 4, FIG. 9 and FIG. 13, relationship node 1504 may gather the decentral participant identifier(s) associated with said decentral identifier(s) from participant ID resolver 418. Relationship node 1504 may gather the endpoint(s) of provider node(s) associated with said decentral participant identifier(s) from provider node resolver 420.

[0229] With reference to FIG. 16, relationship node 1504 may be configured to gather access element associated with the decentral identifier(s) from the provider node(s), for example as described in the context of FIG. 10 and FIG. 16. Relationship node 1504 may parse the received access element data to determine aspect(s) contained within the received access element data. Relationship node 1504 may determine access data associated with relationship data set(s) (e.g. an aspect defining a relationship between the output product and production input(s) used to produce the output product via respective decentral identifiers). The access data may be determined by matching key value pair(s) signifying such aspect(s) with access element data, e.g. with key value pairs included in the retrieved access element(s). Matching may include performing a similarity search between key value pairs signifying such aspect(s) and key value pairs included in the retrieved access element data. The access data may include decentral identifier(s) (e.g. decentral relationship identifier(s)) associated with the relationship data set(s) and digital representation(s) pointing to the respective relationship data set(s). The digital representation pointing to the respective relationship data set may point to the provider node associated with the respective relationship data set. The respective relationship data set may be part of the digital twin of the output product the decentral identifier(s) provided by neural network engine 410 is / are associated with. Relationship node 1504 or a decentral data consuming network node associated with said relationship node 1504 (not shown) may be configured to request relationship data set(s) associated with the decentral relationship identifier(s) from respective provider nodes, for example as described in the context of FIG. 10.

[0230] With reference to FIG. 10, FIG. 13 and FIG. 16, aspect(s) (e.g. relationship data set(s)) may be gathered based on the determined access data, for example as described in the context of FIG. 10 and FIG. The relationship data set may be included in a digital twin of the end product. The relationship data sets may define a relationship between the decentral identifier associated with the output product (e.g. the decentral digital twin identifier) and the decentral identifier(s) associated with production input(s) used to produce the output product, for example via a parent child relationship. The relationship data set may define decentral identifier associated with the output product as parent identifier and the decentral identifier(s) associated with the production input(s) as child identifier(s). The relationship data set may hence allow to determine decentral identifier(s) of production input(s) used to produce the end product. Via the decentral identifier(s) associated with the production input(s), production input data associated with the respective production input (e.g. input material) may be gathered via the decentral network.

[0231] With continued reference to FIG. 16, relationship node 1504 may be configured to parse the gathered relationship data set(s) to determine the decentral child identifier(s) included in said relationship data set(s). If production input data is to be gathered, relationship node 1504 may be configured to gather access element(s) associated with said decentral child identifier(s), for example as previously described. Relationship node 1504 may be configured to parse the access element data to determine access data associated with production input data set(s) (e.g. aspect(s) of a digital twin of the respective production input) as previously described. Using said decentral child identifier(s) and the access data, relationship node 1504 or a decentral data consuming network node associated with said relationship node 1504 (not shown) may be configured to gather the respective production input data set(s) from determined provider node(s) as previously described.

[0232] With continued reference to FIG. 16, relationship node 1504 may be configured to determine whether the access elements associated with the child identifier(s) include access data associated with relationship data set(s) signifying relationships between a child identifier and decentral identifier(s) associated with input material(s) used to produce the respective production input associated with such production input, for example as previously described. Hence, the child identifier(s) associated with the production input(s) may be considered as parent identifier(s). If such access data is included, relationship node 1504 may gather such relationship data set(s) as previously described. Relationship node 1504 may be configured to parse the gathered relationship data set(s) to determine the decentral child identifier(s) included in said relationship data set(s). If production input data is to be gathered, relationship node 1504 may be configured to gather access element(s) associated with said decentral child identifier(s), for example as previously described. Relationship node 1504 may be configured to parse the access element data to determine access data associated with production input data set(s) (e.g. aspect(s) of a digital twin of the respective production input) as previously described. Using said decentral child identifier(s) and the access data, relationship node 1504 or a decentral data consuming network node associated with said relationship node 1504 (not shown) may be configured to gather the respective production input data set(s) from determined provider node(s) as previously described.

[0233] With reference to FIG. 16, relationship node 1504 may be configured to repeat blocks 1612, 1614 and 1608 until no further child identifier(s) may be existing. Hence, relationship node 1504 may be configured to resolve at least a part of the bill of material tree associated with a given output product. The bill of material tree may reflect all production input(s) used to produce the output product. The bill of material tree may reflect all production input(s) used per production step of the production chain associated with the end product. The bill of material tree may include production input(s) being discrete materials and production input(s) being chemical materials. The bill of material tree may include raw materials, chemical intermediate products, chemical products, parts or components and / or component assemblies.

[0234] The production input data gathered by relationship node 1504 may be provided to neural network engine 410. The gathered production input data may be inputted to into the data-driven model to generate transformed data, as described in the context of FIG. 10.

[0235] FIG. 17 illustrates an embodiment of obtaining a trained embedding model. The trained embedding model may be a trained embedding layer. The trained embedding layer may be obtained by training for example a continuous bag of words model (CBOW) or a skip-gram model. The embedding layer may be suitable for generating embedded input data based on input data. Generating embedded input data may refer to embedding input data. Embedding input data may result in a representation associated with the input data. Thus, the embedded input 1714 may be the representation associated with the input data. The input data may comprise one or more elements. The one or more elements may be represented by the input vector 1706. In particular, the embedded input 1714 and / or the input vector 1706 may be machine- readable and / or processable by a processor. For this purpose, the embedded input 1714 and / or the input vector 1706 may be a tensor, in particular a first-rank tensor. Specifically, the input vector 106 may be a one-hot vector or a summation of a plurality of one- hot vectors. A one-hot vector may be a vector with one entry unequal to zero. Examples for one-hot vectors may be 1608, 1710 and 1712. The entries unequal to zero in the one-hot vector and / or in the input vector 1706 may indicate the element. For example, a lookup table may define the relation between the position of the entries unequal to zero and the element indicated by the one-hot vector. The lookup table may specify a plurality of different elements. The number of different elements may be equal to the number of entries in the one-hot vector. The number of different elements may be referred to as vocabulary size. In an example, the elements may be represented by tokens and a sequence of elements may refer to at least a part of a sentence. The at least a part of the sentence may be represented by a plurality of tokens. A token may represent at least a part of the element and / or word. For example, where one element would be associated with only one word, words such as “embeddings", “embedding” or “embed” would constitute different elements. A first token may represent the stem “embed” and the endings, typically appearing in a plurality of word, may be represented by a second token, a third token and a fourth token. The second token, the third token and the fourth token may be used for representing other words such as “look”, “looking” or the like, preferably together with a fifth token representing the stem “look”. Ultimately, this tokenization of elements associated with a plurality of stems and a plurality of endings results in less tokens to be used for representing a plurality of elements and thus, uses less computational resources.

[0236] A lookup table specifying a subset of the vocabulary size e.g. of the English language may comprise 10,000 words or more. The embedded input 1714 may be a lower-dimensional representation than the input vector 1706. For example, typical embedded input 1714 may comprise some hundreds of different entries. Followingly, the embedded input 1714 constitute a densified representation of one or more elements using less computational resources. More than that, the embedded input 1714 may represent a relation between two or more elements. For example, the words “Italy” and “Germany” may be similar or may be more closely related since they both define European countries, whereas the word “embodiment” may be very different from the two respective words. The smaller the dot product between two embedded input 1714 may be the more similarthe two elements associated with the embedded input 1714 may be. Hence, the embedded input 1714 may represent one or more elements accurately and lead to accurate results based on processing the embedded input 1714.

[0237] For transforming the input vector 1706 into the embedded input 1714, the embedding layer may comprise a number of neurons equal to the number of entries in the embedded input 1714. Based on the embedded input 1714, the output layer may generate the output vector 1716. The output vector may be a vector and / or may indicate one or more elements. The output vector 1716 may indicate one or more elements different from the input vector 1706 and / or the one-hot vectors associated with the input vector 1706. For this purpose, the output layer may comprise a number of neurons equal to the number of entries of the input vector 1706 and / or the output vector 1716. The output layer may apply a softmax function to the embedded input 1714. By doing so, the output vector may comprise the probabilities associated with the elements associated with the entries of the output vector 1716 unequal to zero. Hence, from the output vector 1716 one or more elements may be obtained with a corresponding probability. Where the input vector 1706 may specify one or more sequence(s) of elements, the output vector 1716 may specify one or more elements corresponding to the sequence(s) of elements specified by the input vector 1706. In the example of FIG. 17, the element associated with vector 1718 may correspond to the input vectorwith a probability of 71 %. Additional or alternative elements may correspond to the input vector as indicated by the output vector with lower probability. By defining a threshold to which the probability may be compared, the selection of the corresponding elements may be tailored to the needs of the user. The elements generated by the model comprising the embedding layer 1702 and the output layer 1704 may refer to the most probable elements indicated by the output vector 1716. Hence, the model depicted in FIG. 17 may generate the element associated with the vector 1718 with a confidence score of 71 %. The model of FIG. 17 may be continuous bag of words (CBOW) model. The CBOW model may be trained based on a training data set comprising a plurality of input vectors and corresponding output vectors. As the training data set may not be labelled, the training of the CBOW model may be referred to as self-supervised. Before training of the CBOW model, the CBOW model may be initialized with random values assigned to the weights of the neurons. During the training of the CBOW model, the input vectors may be passed through the initialized embedding layer and the output layer and a loss may be determined by comparing the output vector obtained by passing the input vector 1706 through the model to the output vector corresponding to the input vector 1706 as specified by the training data set. Based on the determined loss, backpropagation may be applied to determine the gradients associated with the neurons of the embedding layer 1702 and the output layer 1704 to lower the loss. According to the determined gradients, the weights of the neurons may be updated by using a gradient descent algorithm. If a predetermined loss may be achieved by the CBOW model, the training may be terminated and a trained CBOW model may be obtained. From the trained CBOW model, the embedding layer 1702 may be suitable for embedding input data comprising one or more elements. This embedding layer 1702 may be used in other machine-learning architectures requiring an embedding layer 1702 such as a transformer encoder, transformer decoder or transformer encoder decoder architecture.

[0238] The present disclosure has been described in conjunction with preferred embodiments and examples as well. However, other variations can be understood and effected by those persons skilled in the art and practicing the claimed invention, from the studies of the drawings, this disclosure and the claims.

[0239] Any steps presented herein can be performed in any order. The methods disclosed herein are not limited to a specific order of these steps. It is also not required that the different steps are performed at a certain place or in a certain computing node of a distributed system, i.e. each of the steps may be performed at different computing nodes using different equipment / data processing.

[0240] As used herein ..determining" also includes ..initiating or causing to determine", “generating" also includes ..initiating and / or causing to generate" and “providing” also includes “initiating or causing to determine, generate, select, send and / or receive”. “Initiating or causing to perform an action” includes any processing signal that triggers a computing node or device to perform the respective action.

[0241] In the claims as well as in the description the word “comprising” or “including” or similar wording does not exclude other elements or steps and shall not be construed limiting to the elements or steps lined out. The indefinite article “a” or “an” does not exclude a plurality. A single element or other unit may fulfill the functions of several entities or items recited in the claims. The mere fact that certain measures are recited in the mutual different dependent claims does not indicate that a combination of these measures cannot be used in an advantageous implementation or further elements may be included.

[0242] Providing in the scope of this disclosure may include any interface configured to provide data. This may include an application programming interface, a human-machine interface such as a display and / or a software module interface. Providing may include communication of data or submission of data to the interface, in particular display to a user or use of the data by the receiving entity.

[0243] Any disclosure and embodiments described herein relate to methods, systems, apparatuses, devices, chemicals, materials, services, uses, computer program elements lined out above and vice versa. Advantageously, the benefits provided by any of the embodiments and examples equally apply to all other embodiments and examples and vice versa.

Claims

CLAIMS1. A method for retrieving data associated with produced output product(s) and / or input material(s) used to produce the output product(s) from a decentral network, the method comprising the steps of:(a) providing search data related to the data associated with the output product(s) and / or the input material(s) to be retrieved from the decentral network;(b) converting the provided search data into query data by a query data generator model trained to generate query data for retrieving output product data associated with the output product(s) and / or input material data associated with the input material(s) from the decentral network based on the provided search data;(c) gathering decentral identifier(s) associated with the output product(s) from the decentral network based on the query data generated by the query data generator model;(d) gathering output product data from the decentral network based on the gathered decentral identifier(s) associated with the output product(s) and / or gathering input material data from the decentral network based on the gathered decentral identifier(s) associated with the produced output product(s);(e) generating the data associated with the output product(s) and / or the input material(s) by transforming the gathered output product data and / or the input material data by a data transformer model trained to transform the gathered output product data and / or input material data based on the provided search data;(f) providing the generated data associated with the output product(s) and / or the input material(s).

2. The method of claim 1 , wherein the search data includes unstructured data associated with a context, in particular wherein the context is associated with the data associated with the output product(s) and / or input material(s).

3. The method of claim 1 or 3, wherein the query data includes structured data, in particular structured data included in access element(s) associated with the output product(s) and / or the input material(s), wherein the access element(s) provide access to the output product data and / or the input material data.

4. The method of any one of claims 1 to 3, wherein the query data generator includes at least one data-driven model parametrized according to training data set(s) including unstructured data and associated query data.

5. The method of any one of claims 1 to 4, wherein the at least one data-driven model trained to generate the query data generates at least one classifier discriminating between one or more output product types, one or more manufacturer types, one or more input material types, one or more output product type properties and / or one or more input material type properties.

6. The method of any one of claims 1 to 5, wherein the query data relates to the data associated with the produced output product(s) and / or the input material(s).

7. The method of any one of claims 1 to 6, wherein the data associated with the output product(s) is related to access element(s) configured to provide access to the output product data, wherein the query data includes key-value pair(s) contained within such access element(s) and being related to the data associated with the output product(s).

8. The method of any one of claims 1 to 7, wherein the data associated with the input material(s) is related to access element(s) configured to provide access to the input material data, wherein the query data includes key value pair(s) contained within such access element(s) and being related to the data associated with the input material(s).

9. The method of any one of claims 1 to 8, wherein the input material data is gathered based on relationship data signifying a relationship between the produced output product and the input material(s) used to produce the output product.

10. The method of any one of claims 1 to 9, wherein gathering the output product data includes: gathering access element(s) associated with the produced output product based on the gathered decentral identifier(s) associated with the output product(s), gathering the output product data based on access data included in the access element and the query data.11 . The method of any one of claims 1 to 10, further including a step of storing the gathered output product data and / or input material data in a database, in particular wherein the gathered output product data and / or input material data is stored in the database prior to transforming the gathered output product data and / or input material data.

12. The method of any one of claims 1 to 11 , wherein the data transformer model includes at least one data-driven model parametrized according to training data set(s) including unstructured data and associated data transformation scheme(s)13. The method of any one of claims 1 to 12, wherein the data transformer model trained to transform the gathered output product data and / or input material data determines database query data to retrieve output product data and / or the input material data from a database storing gathered output product data and / or input material data based on the provided user input data.

14. An apparatus for retrieving data associated with produced output product(s) and / or input material(s) used to produce the output product(s) from a decentral network, the apparatus comprising:(a) a data providing interface configured to provide search data related to the data associated with the output product(s) and / or the input material(s) to be retrieved from the decentral network;(b) a query data generator model configured to convert the provided input data into query data, wherein the query data generator model is trained to generate query data for retrieving output product data associated with the output product(s) and / or input material dataassociated with the input material(s) from the decentral network based on the provided search data;(c) a decentral network interface configured to• gather decentral identifier(s) associated with the output product(s) from the decentral network based on the query data generated by the query data generator model and• gather output product data from the decentral network based on the gathered decentral identifier(s) associated with the produced output product(s) and / or to gather input material data from the decentral network based on the gathered decentral identifier(s) associated with the produced output product(s);(d) a data transformer model configured to generate the data associated with the output product(s) and / or the input material(s), wherein the data transformer model is trained to transform the gathered output product data and / or input material data based on the provided search data;(e) a data providing interface configured to provide the generated data associated with the output product(s) and / or the input material(s).

15. Use of the data associated with produced output product(s) and / or input material(s) used to produce the output product(s) retrieved from the decentral network according to the method of any one of claims 1 to 13 or the apparatus of claim 14 for controlling and / or managing the production of further output product(s) and / or for controlling and / or managing the re-use and / or recycling of end-of- life product(s).

Images