Intelligent monitoring of material production

Abstract

The present disclosure relates to the intelligent monitoring of output materials using large language models. Disclosed are methods, apparatuses, systems for monitoring and / or controlling output materials produced or to be produced by a distributed chemical production network to enhance handling of the output materials for users.

Description

[0001] INTELLIGENT MONITORING OF MATERIAL PRODUCTION

[0002] TECHNICAL FIELD

[0003] The present disclosure relates to the intelligent monitoring of output materials using large language models. Disclosed are methods, apparatuses, systems for monitoring and / or controlling output materials produced or to be produced by a distributed chemical production network to enhance handling of output materials for users.

[0004] TECHNICAL BACKGROUND

[0005] Distributed chemical production environments are large scale networks producing multiple thousands of chemical products. Managing material or product flows from supply of input material for production of output material to providing of produced output material for further downstream production is challenging. Particularly the chemical environment is very diverse with respect to the output materials produced and provided and the users or applications for such output materials.

[0006] SUMMARY OF THE INVENTION

[0007] In one aspect disclosed is method for monitoring and / or controlling output materials produced or to be produced by a distributed chemical production network, the method comprising:

[0008] - receiving one or more request(s) to provide one or more output materials from one or more output material users, wherein the one or more request(s) include different request types in relation to the data structure of the one or more request(s),

[0009] - classifying the one or more request(s) based on the one or more request type(s) and routing the one or more request(s) based on the classification to one or more extraction agent(s) configured to extract a natural language representation from the one or more request(s),

[0010] - generating one or more instruction(s) for generating at least one common, machine-readable representation based on the natural language representation, wherein the at least one common, machine-readable representation is configured to monitor and / or control output materials produced or to be produced by the distributed chemical production network,

[0011] - generating the common, machine-readable representation by providing the generated one or more instruc- tion(s) to a general-purpose data-driven model, wherein the general-purpose data-driven model is trained on unstructured natural language data and is configured to process natural language;

[0012] - providing the common, machine-readable representation for monitoring and / or controlling output materials produced or to be produced by the distributed chemical production network.

[0013] In another aspect disclosed is an apparatus for monitoring and / or controlling output materials produced or to be produced by a distributed chemical production network, the apparatus comprising:

[0014] - an input interface configured to receive one or more request(s) to provide one or more output materials from one or more output material users, wherein the one or more request(s) include different request types in relation to the data structure of the one or more request(s), - a classifier configured to classify the one or more request(s) based on the one or more request type(s) and routing the one or more request(s) based on the classification to one or more extraction agent(s) configured to extract a natural language representation from the one or more request(s),

[0015] - an instruction generator configured to generate one or more instruction(s) for generating at least one common, machine-readable representation based on the natural language representation, wherein the at least one common, machine-readable representation is configured to monitor and / or control output materials produced or to be produced by the distributed chemical production network,

[0016] - a representation generator configured to generate the common, machine-readable representation by providing the generated one or more instruction (s) to a general-purpose data-driven model, wherein the general- purpose data-driven model is trained on unstructured natural language data and is configured to process natural language;

[0017] - an output interface configured to provide the common, machine-readable representation for monitoring and / or controlling output materials produced or to be produced by the distributed chemical production network.

[0018] In another aspect disclosed is a system for monitoring and / or controlling the providing of output materials the system comprising:

[0019] - a distributed chemical production network configured to provide multiple output materials,

[0020] - an apparatus for monitoring and / or controlling output materials produced or to be produced by the distributed chemical production network as disclosed herein.

[0021] In another aspect disclosed is the use of the common, machine-readable representation generated according to the methods disclosed herein or by the apparatuses disclosed herein for monitoring and / or controlling output materials produced or to be produced by a distributed chemical production network. In another aspect disclosed is a method for using the common, machine-readable representation generated according to the methods disclosed herein or by the apparatuses disclosed herein for monitoring and / or controlling output materials produced or to be produced by a distributed chemical production network. From the common, machine-readable representation monitoring and / or controlling instructions suitable for monitoring and / or controlling facilities of the distributed chemical production network may be generated.

[0022] EMBODIMENTS

[0023] Any disclosure and embodiments described herein relate to the methods, the apparatuses, the systems, the uses, and the computer elements lined out above and below and vice versa. Advantageously, the benefits provided by any of the embodiments and examples equally apply to all other embodiments and examples.

[0024] In the following, embodiments of the present disclosure will be outlined by ways of embodiments and / or example. It is to be understood that the present disclosure is not limited to said embodiments and / or examples. All terms and definitions used herein are understood broadly and have their general meaning. The methods, the apparatuses, the systems, the uses, and the computer elements disclosed herein provide an efficient, and reliable way for handling of output materials produced by a chemical production environment. The request may include one or more file / protocol types with (semi-) un-structured data that may be different for each chemical product order. On top of that the specifications and instructions in the order may differ e.g., in chemical product description, amounts and packaging. This makes order intake cumbersome and error prone. To provide the output materials more reliably and flexibly, large language model is used. By transforming the requests to text representations, the text representations may be provided to the pre-trained LLM together with instructions that allow for transforming the diverse file / protocol types of orders into a common format. By using the LLM with specific instruction representation the mapping to machine-readable representation suitable for triggering production and / or transport of the output material can be executed more reliably independent of the explicit specification provided in the request. In addition, standardization intermediaries translating requests to single formats may be eliminated.

[0025] The distributed chemical production network may include at least multiple warehouse, transport and / or production facilities configured to provide multiple output materials. The distributed chemical production network may include at least multiple warehouse facilities configured to store and provide produced output materials. The distributed chemical production network may include at least multiple transport facilities configured to transport and provide produced output materials. The distributed chemical production network may include at least multiple production facilities configured to produce multiple output materials. The warehouse, transport and / or production facilities may be distributed locally. The warehouse, transport and / or production facilities may be configured to store, transport and / or produce the same or substantially similar output materials or different output materials, respectively.

[0026] The distributed chemical production network may include multiple types of production processes for producing one or more output material(s) from one or more input material(s) or intermediate material(s). The chemical production network may derive one or more output material(s) from the input material(s) or intermediate material(s). The chemical production network may include a production network producing multiple output materials in or through multiple production chains e.g. composed of different production facilities. The chemical production network may include connected, interconnected and / or non-connected production chains. The chemical production network may include one or more production chains with multiple production facilities. The production facility may perform one or more production step(s) to transform input materials or intermediate materials provided to the facility to output materials or intermediate materials procured by the facility. The intermediate materials provided to the facility may be produced by another or preceding facility of the chemical production network e.g. an upstream facility. The output materials of the facility may be intermediate materials provided to another or subsequent facility of the chemical production network e.g. a downstream facility. The output materials may be provided to a warehouse for storage. The materials may be transported between warehouse and / or production facilities by transport facilities such as trains, trucks, air freight or other equipment suitable for movement of materials. The output materials may leave or exit the system boundary of the chemical production network. The output materials may be provided to the users of the output materials by transport facilities such as trains, trucks, air freight or other equipment suitable for movement of materials. The production facilities and / or production steps included in the chemical network may be defined by the physical system boundary of the chemical production network. The system boundary may be defined by location or control over production processes. The system boundary may be defined by the site of the chemical production network. The system boundary may be defined by production processes controlled by one entity or multiple entities jointly. The system boundary may be defined by the value chain with staggered production processes to an end product, which may be controlled by multiple entities jointly or separately. The chemical production network may include a waste collection and sorting step, a recycling step, a chemical reaction step, a separation step and further processing steps to convert such intermediates to output materials leaving the system boundary of the chemical production network. The entry points of the chemical production network may be marked by the entry of input materials to the chemical production network or the system boundary of the chemical network. The output materials may leave the physical system boundary of the chemical production network. The exit points of the chemical production network may be marked by the exit of output materials from the chemical production network or the system boundary of the chemical network.

[0027] The output material may include or be any material produced by the chemical production network using at least one input material. The output material may comprise or be any output material leaving the chemical production network at any exit point. The output material may comprise any chemical product obtained from chemical reactions as well as natural chemical products. Natural chemical products may encompass any chemical substance that is naturally occurring, i.e. any unprocessed chemical substance that is found in nature, such as chemicals from plants, microorganisms, animals, the earth and the sea or any chemical substance that is found in nature and extracted using a process that does not change its chemical composition Natural chemical products may include biologicals like enzymes as well naturally occurring inorganic or organic chemical products. Natural chemical products can be isolated and purified prior to their use or they can be used in non-isolated and / or unpurified form. Chemical products obtained from chemical reactions may be any inorganic or organic chemical product obtained by reacting inorganic and / or organic chemical reactants. The inorganic and organic chemical reactants may be naturally occurring chemical products or can be chemical products obtained from chemical reactions. Chemical reactions may include any chemical reaction commonly known in the state of the art in which the reactants are converted to one or more different chemical products. Chemical reactions may involve the use of catalysts, enzymes, bacteria, etc. to achieve the chemical reaction between the reactants.

[0028] Monitoring and / or controlling output materials produced or to be produced by the distributed chemical production network may include operating distributed facilities of the chemical production network. Monitoring and / or controlling output materials produced or to be produced by the distributed chemical production network may include operating at least production, warehouse and / or transport facilities of the chemical production network. Monitoring and / or controlling output materials produced or to be produced may include generating machine-readable monitoring and / or control data configured to initiate the providing of output materials. Machine-readable monitoring and / or control data may be generated based on or from the common, machine-readable representation. Machine-readable monitoring and / or control data may be provided to at least one monitoring and / or control system of at least one of the facilities of the chemical production network. The monitoring and / or control system may be configured to provide the monitoring and / or control data to an operator operating the facility and / or to operation equipment of the facility. The monitoring and / or control system of the warehouse facility may be configured to provide the monitoring and / or control data to warehouse equipment such as forklifts or robots configured to store packaged output materials to designated storage spaces. The monitoring and / or control system of the transport facility may be configured to provide the monitoring and / or control data to transport equipment such as trains or ships configured to move packaged output materials to users of the output materials. The monitoring and / or control system of the production facility may be configured to provide the monitoring and / or control data to an operator of a production equipment or to a production equipment such as valves or pumps controlling material flow of production processes.

[0029] One or more or multiple request(s) to provide one or more output materials received from one or more output material user(s) may be associated with the one or more output material (s) to be provided, produced or to be produced by the chemical production network. The multiple requests may relate to different request types. The different request types may relate to different data structure, such as file and / or protocol, types with structured, semi-structured and / or unstructured data representing at least one specification of the one or more output material(s) to be provided. The different request types may relate to pre-defined, semi-defined and / or undefined data structures representing at least one specification of the one or more output material(s) to be provided. Undefined, semi-defined or defined may relate to the content of the request in relation to the one or more output material(s) to be provided. For example, at least one request type may relate to undefined data structure representing at least one specification of the one or more output material(s) to be provided. The data points of the undefined data structure may include the at least one specification following an undefined data schema or format, such as natural language audio messages or natural language text messages. Further for example, at least one request type may relate to the semi-defined data structure representing at least one specification of the one or more output material(s) to be provided. The data points of the semi-defined data structure may include the at least one specification following a semi-defined data schema, such as Java Script Object Notation (JSON) or Extensible Markup Language (XML). The semi-defined data schema may be suitable for annotating and validating structure, constraints, and / or data types of the data structure. Further for example, at least one request type may relate to the defined data structure representing at least one specification of the one or more output material(s) to be provided. The data points of the defined data structure may include the at least one specification following a defined data format for data messages, and protocols governing exchange of such data messages, such as Electronic Data Interchange (EDI), decentralized network protocols such as bitcoin or ethereum or other peer-to-peer network protocols. The multiple requests may be provided by different requesting entities associated with the one or more output material user(s) configured to provide one or more request type(s).

[0030] The request may include one or more indication(s) or specification (s) relating to the providing of output material(s) produced or to be produced. The request may include one or more indication(s) or specification (s) relating at least to output material(s), quantity / ies of output material(s), user(s) of the output material(s), location (s) for providing and / or using the output material(s), time for providing and / or using the output material(s), transport or shipping property / ies for providing and / or using the output material (s), packaging property / ies for providing and / or using the output material (s), technical property / ies of the output material or combinations thereof.

[0031] The common, machine-readable representation may relate to structured, semi-structured and / or unstructured data representing at least one specification for monitoring and / or controlling output materials produced or to be produced by the distributed chemical production network. The common, machine-readable representation may include one or more standardized data representation (s) for monitoring and / or controlling output materials produced or to be produced by the distributed chemical production network. The common, machine-readable representation may be configured for monitoring and / or controlling output materials produced or to be produced by the distributed chemical production network. The common, machine-readable representation may relate to one or more standardized data representation(s) depending on the providing of the output material to be produced or produced. The common, machine-readable representation may be pre-defined e.g. depending on the providing of the output material to be produced or produced. The common, machine-readable representation may relate to one or more standardized or predefined data representation(s) configured for controlling and / or monitoring the warehouse, production and / or transport facilities of the distributed chemical production network.

[0032] In one embodiment the one or more or multiple request(s) include at least one request type per request. Multiple requests of different request type may be received e.g. by or from different requesting entities associated with the one or more output material user(s). The request types may differ in relation to the data structure of the one or more request(s).

[0033] The classification of the one or more request(s) may be based on the one or more request type(s). The classification may include classifying the requests received e.g. by or from different requesting entities associated with the one or more output material user(s) according to their respective request type. The classification may include classifying the requests according to their respective data structure, such as file and / or protocol types, with structured, semistructured and / or unstructured data representing at least one specification of the one or more output material(s) to be provided. The classified request(s) may be routed to one or more extraction agent(s) configured to extract the natural language representation from the respectively classified request(s). The extraction agent(s) may be configured to extract the natural language representation per request type(s). The extraction agent(s) may be configured to extract the natural language representation depending on request type(s).

[0034] In another embodiment the extraction of the natural language representation from the one or more request(s) includes a transformation of the data structure of the request to a natural language or text-based data structure. The extraction of the natural language representation from the one or more request(s) may include extraction of data associated with the one or more output material(s) to be provided. The extraction of the natural language representation from the one or more request(s) may include extraction of data associated with the one or more output materials) produced or to be produced by the chemical production network. The extraction of the natural language representation from the one or more request(s) may include language detection based on rule-based language classifica- tion and / or statistical model-driven language classification. The extraction of the natural language representation from the one or more request(s) may include language translation of at least part of the extracted natural language representation to one or more pre-defined target language(s). The natural language representation may include natural language or text-based data associated with the one or more output material(s) to be provided. The natural language representation may include natural language or text-based data associated with the one or more output material (s) produced or to be produced by the chemical production network.

[0035] In another embodiment the one or more instruction (s) relate to generating a common, machine-readable data structure. Generating the one or more instruction(s) may include retrieving one or more historical pair(s) including request and common, machine-readable representation. The common, machine-readable data structure may be configured to be processed by one or more systems for controlling and / or monitoring the distributed chemical production network, in particular in relation to facilities of the chemical production network, further in particular in relation to controlling and / or monitoring the providing of output material by facilities of the chemical production network.

[0036] In another embodiment the one or more instruction (s) are generated based on at least one template instruction. The template instruction may include at least one task instruction for the general-purpose data-driven model, at least one context instruction relating to the natural language representation of the request, and / or context data retrieved from at least one context database based on the one or more request(s). The at least one template instruction may depend on the request type and / or the data points provided by the request. For example, some request types may be more specific and structured with respect to the data points provided than other request types. The template instruction may include different sets of context data depending on request type or the specificity of the request type. The one or more instruction (s) may be generated by adding the natural language representation of the request and context data to the instruction template. Based on the natural language representation extracted from the request, context data stored in knowledge or context database(s) may be retrieved. The context data may relate to one or more historical pair(s) of requests and corresponding common, machine-readable representation. The context data may relate to the providing of output material(s) by the output material producer to the output material user(s). The context data may include request-specific context data associated with the request for providing of output material(s) by the output material producer to the output material user(s), for example the context data may be specific to the output material user requesting the providing of the output material. The context data may relate to historical request(s) by the output material user. The context data may relate to the output material requested by the output material user. The at least one template instruction may include output instruction(s) specifying at least in part the common data structure of the common, machine-readable representation. The output instruction(s) may depend on the providing of the output material to be produced or produced, such as the monitoring and / or controlling systems associated with the facilities of the chemical production network providing the output material.

[0037] In another embodiment the common, machine-readable representation is generated by a data-driven model based on the request and a pre-defined data structure of the common, machine-readable representation. The instruction(s) for generating the common, machine-readable representation of the request may be provided to at least one general- purpose data-driven model suitable or adaptable for generating the common, machine-readable representation of the request. The instruction(s) for generating the common, machine-readable representation of the request may be provided to a pre-trained, transformer-based model for mapping the text-based, natural language-based and / or unstructured instruction to the common, machine-readable representation. The data-driven model may be a pre-trained model. The pre-trained model may be a general-purpose model trained or parametrized based on general data sets including data pairs not specific to a task, in particular not specific to generating common, machine-readable representation, further in particular not specific to the request(s) received. The pre-defined data structure of the common, machine-readable representation may be configured to provide output materials produced or to be produced by the distributed chemical production network. The pre-defined data structure of the common, machine-readable representation may be configured to monitor and / or control output materials produced or to be produced by the distributed chemical production network. The (e.g. general-purpose) data-driven model may be a generative data-driven model, it may be a model, e.g. implemented in a computer system, that, based on historical data it has been trained on, may generate new instances of said data e.g. by sampling from a probability distribution, which may have been learned during training, and generate an according output data set after receiving a task instruction. A generative data-driven model may have been trained on general purpose training data sets (and via that training may be configured to) to generate an output data set in response to obtaining (e.g. receiving) the task instruction. A generative data-driven model may be or comprise a transformer-based data driven model, preferably a decoder-only transformer-based model such as a generative pre-trained transformer, e.g. Large Language Model Meta Al (Llama), Llama 2, Llama 3, Mistral 7B, GPT 3.5, GPT 3.5 turbo, GPT 4, GPT 4o. A generative data-driven model may be or comprise a mixture of experts architecture based model e.g. Mixtral 8x7B, Mixtral 8x22B. In a mixture of experts model several decoder blocks may be operated in parallel representing different experts or a feed-forward layer in a block may be split into separate parallel feed-forward layer, wherein each of the parallel feed-forward layers may be regarded as an expert and may learn to focus on different tasks during training. A gating network may be used to switch a particular input to the respective expert, e.g. by training the gateway network alongside the experts for instance using an expectationmaximization algorithm or a gradient descent algorithm. A generative data-driven model may be or comprise a selective state space sequence architecture based model e.g. Mamba. A generative data-driven model may be or comprise a combined architecture such as Mamba LLM or Mamba Mixture of Experts (which may comprise alternating Mamba and mixture of experts layers). A selective or structured state space sequence architecture (e.g. SSMs, S4, or 86 models) may allow for using more context in generation and allow for generating larger output data sets.

[0038] In another embodiment the method additionally includes the step of enriching at least one common, machine- readable representation by providing reference data retrieved from at least one reference database based on the at least one common, machine-readable representation. The reference data points may be used to adapt data points of the common, machine-readable representation and / or to add data points to the common, machine-readable representation. The at least one reference database may store reference data points that may be specified by the common data structure of the common, machine-readable representation. The reference data point(s) may relate to specification (s) as used by the provider of the output material produced or to be produced. The reference data points may be retrieved from the at least one reference database based on at least part of the generated common, ma- chine-readable representation. The part of the generated common, machine-readable representation may relate at least to the specification of the output material, the specification of the facility the output material may be provided by, the specification relating to the user of the output material or any combinations thereof. The reference data point(s) provided by the common, machine-readable representation may be compared to the reference data point(s) e.g. stored or available in the at least one reference database. The reference data point(s) may be related to matching tables including synonymous or equivalent data point(s). If the comparison leads to synonymous or equivalent data point(s) used by the respective part of the common, machine-readable representation, the respective part of the common, machine-readable representation may be adapted. The retrieved reference data point may replace the respective part of the common, machine-readable representation. If the comparison leads to missing data point(s) of the respective part of the common, machine-readable representation, the respective part of the common, machine- readable representation may be added.

[0039] In another embodiment method additionally including the step of validating the at least one common, machine- readable representation by one or more validation instruction(s) to the general-purpose data-driven model.

[0040] In another embodiment the step of validating includes providing at least one validation template. The validation template may include task instructions for the general-purpose data-driven model to validate the common representation. The at least one validation instruction may be generated based on the validation template.

[0041] In another embodiment the validation template includes at least one placeholder for the common, machine-readable representation to be validated. The validation template may relate to rules to be checked on validation. The rules to be checked on validation may be pre-defined e.g in natural language or text The validation template may include correction instructions to be used for error handling on validation The correction instructions may be pre-defined e.g. in natural language or text. The validation template may include one or more process trigger(s) in case at least one error is found upon validation and cannot be resolved. The one or more process trigger(s) may be pre-defined e.g. in natural language or text. The validation template may include at least one placeholder for one or more historical example(s) relating to validated common, machine-readable representations. The one or more historical example(s) may include the historical request and the corresponding common, machine-readable representation as described above and below for the context data with one or more historical pair(s) of requests and corresponding common, machine-readable representation.

[0042] In another embodiment the request relates to at least to one or more output material(s), one or more user(s) of the output material, one or more appl ication (s) of the output material, one or more producer(s) of the output material, one or more quantity / ies per output material, one or more time specification (s) related to the providing of the output material, one or delivery specification(s) related to the providing of the output material and / or one or more technical prop- erty / ies per output material. In another embodiment additionally including the step of determining a total quantity of output material to be provided based on the at least one common, machine-readable representation. The total quantity may be determined per output material produced or to be produced. Based on the total quantity one or more facility / ies of the distributed chemical production network to provide the output material may be determined.

[0043] The disclosed method may for instance be performed, carried-out, executed and / or controlled by an / the computing apparatus, for instance a server, a server cloud, a computer-system, or part thereof. The disclosed method or any step of the method may be computer-implemented. For instance, the method or any step of the method may be performed and / or controlled by using at least one processor e.g. of an / the apparatus. Alternatively, the method according to any aspect may be performed, carried-out, executed and / or controlled by more than one apparatus, for instance a server cloud comprising at least two servers or a system of apparatus, e.g. a system comprising at least one server providing at least one data base comprising production data represented in a graph structure, at least one server providing a generative data-driven model, and an apparatus comprising means for carrying-out the respective steps of the method according to the first aspect. For instance, an apparatus is disclosed, the apparatus comprising respective means for carrying out or performing the steps of the method according to the first aspect and / or any embodiment or example and combinations thereof of the method. Additionally or alternatively the apparatus may comprise at least one processor and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to carry out the steps of the disclosed method and / or any embodiment or example and combinations thereof of the disclosed method. Additionally or alternatively the apparatus may comprise circuitry (e.g. hardware-only circuitry, digital circuitry and / or a combination of hardware circuits and software) designed or configured the implement the functions for carrying out the steps of the method according to the first aspect (and / or any embodiment or example and combinations thereof of the method). Circuitry may be implemented in a chipset or a chip or an integrated circuit.

[0044] In another embodiment additionally including the step of determining, based on the at least one common, machine- readable representation, at least one facility of the distributed chemical production network for providing the output material as requested. A trigger signal may be generated and provided to the at least one determined facility of the distributed chemical production network.

[0045] BRIEF DESCRIPTION OF DRAWINGS

[0046] In the following, the present disclosure is further described with reference to the enclosed figures:

[0047] Fig. 1 illustrates an example of an intelligent platform for monitoring and / or controlling material production by a distributed chemical production environment.

[0048] Fig. 2 illustrates an example method for monitoring and / or controlling output materials produced or to be produced by a distributed chemical production network. Fig. 3 illustrates an example apparatus for monitoring and / or controlling output materials produced or to be produced by a distributed chemical production network.

[0049] Fig. 4 illustrates an example of data transmissions in relation to the method steps illustrated in Fig. 3.

[0050] Fig. 5 illustrates example request type and corresponding text extraction.

[0051] Fig. 6 illustrates examples of common representations prior and after enrichment and / or validation.

[0052] Fig. 7 illustrates example production facilities controlled based on received request.

[0053] Figs. 8-11 illustrate an example transformer architecture of a general-purpose model. Specifically, FIG. 8 illustrates an embodiment of training an embedding layer. FIG. 9A illustrates an embodiment of a transformer encoder architecture. FIG. 9B illustrates an embodiment of a transformer decoder architecture. FIG. 90 illustrates an embodiment of a transformer encoder-decoder architecture. FIG. 10 illustrates an embodiment of training and / or deploying the transformer encoder, the transformer decoder and / or the transformer encoder-decoder. FIG. 11 illustrates an embodiment of input embedding.

[0054] DESCRIPTION OF EMBODIMENTS

[0055] Fig. 1 illustrates an example of an intelligent platform for monitoring and / or controlling material production by a distributed chemical production environment.

[0056] The monitoring and / or controlling platform encompasses multiple computational layers including request providing systems, intake processing layers, operational instruction generation layers in connection with or communicatively coupled to monitoring and / or controlling systems of the distributed chemical production environment.

[0057] Request providing systems may include document providing systems configured to provide the request for providing one or more output materials via text-based files (order), such as document providing systems like e-mail servers, paper scanners, electronic documents, pdf files, doc(x) files, txt files, shared documents on shared storage or the like. Request providing systems may include electronic message providing systems configured to communicate the request for providing one or more output materials (order) via electronic message protocols, such as peer-to-peer communication channels, wireless communication channels, internet-based communication channels or the like. Request providing systems may include text or speech message providing systems configured to provide the request for providing one or more output materials (order) in spoken natural language, such as phones, voicemails, speechbased assistance or the like.

[0058] The request for providing one or more output materials may include multiple data points. The data points may be embedded in different electronic formats or data structures. Multiple requests may have different format per request. The request may be associated with one or more output material(s) produced or to be produced by the chemical production network. The request may be associated with the production and / or transport of output material produced or to be produced. The request may relate to the one or more output material(s), one or more user(s) of the one or more output material(s), one or more applications of the output material, one or more producer(s) of the one or more output material(s), one or more quantity / ies per output material, one or more time specification(s) related to the providing of one or more output material(s), one or more delivery specification(s 9related to the providing of one or more output material(s) and / or one or more technical properties of the one or more output material(s). The one or more output material(s) may relate to any chemical product produced by the chemical production network including at least one chemical reaction between educts to form chemical products. The chemical product may include a base chemical such as propylene or ethylene. The chemical product may include an intermediate chemical such as polyolefins, polyol or isocyanate. The chemical product may include a specialty product such polyurethane, polyamid or surfactants. The chemical product may include a mixture of chemical components such as herbicides, fungicides or insecticides. The chemical product may be a biological, organic and / or inorganic material. The chemical product may include biological, organic or inorganic components. One or more applications of the output material may relate to technical applications or uses of the output material such production of further products based on the output material. One or more technical properties of the output material may relate to technical properties of the output material itself such as quality properties or properties that may be provided by a certificate of analysis or safety relevant properties such as may be provided by a safety data sheet. Quality properties may relate to the application of the output material such as purity, shelf life, isomer ratio, acidity, content of specific components such as chloride, viscosity, boiling point, molecular weight, flash point, melting, point, vapor pressure, vapor density or the like. The safety data may be associated with hazardous properties of the output material. The certificate of analysis may be associated with laboratory measurement data as measured for the output material, the output material class or the output material type. One or more technical properties of the output material may relate to technical application properties of the output material in relation to one or more specific applications or uses of the output material such as efficacy of a herbicides, fungicides or insecticides with respect to an application target or species, foamability of a formulation or cleaning properties of a formulation.

[0059] The request may include identifiers or representations that signify or specify the one or more output materials such as CAS number, smile string or IUPAC name, one or more users of the output material such as customer ID or customer name, one or more applications of the output material such as material for building insulation, material for footwear, material for packaging, material for toys or material for food packaging, one or more producers of the output material, such as any operator of the chemical production network, one or more quantities per output material such as amount, volume or pieces, one or more time specifications related to the providing of the output material, such as delivery time, one or more delivery specifications related to the providing of the output material, such as packaging type, location of user of output material or the like, and / or one or more technical properties of the output material. The identifiers or representations may be provided in a structured, unstructured, or semi-structured format. The request may include unstructured data such as text or speech data that may be provided by the user of the output material e.g. in text format such as doc(x), txt, emails or pdf or audio formats such as mp3 or wave. The request may include semi structured data such as machine-readable text data including semi-structured key-value pairs with numeric- and / or text-based data points as may be provided by electronic devices e.g. in JSON, avro, ocr format or electronic communication protocols such as html according to internet standards, 5G, Bluetooth or other communication standards. The request may include structured data including structured key-value pairs with numeric and / or text-based data points such as provided by electronic storages e.g. in parquet, csv format.

[0060] The intake layer may be configured to receive the request for providing one or more output material(s) from the different types or forms of request providing systems. The intake layer may hence act or be configured as central instance for ingesting any request or multiple requests related to the providing of one or more output materials. The intake layer may be configured to receive multiple request(s) for providing one or more output material(s) with or in different electronic formats. The intake system may further be configured to process the different forms or types of request(s) to a common machine-readable format suitable for further processing.

[0061] The operation instruction layer may be configured to receive the processed request in common machine-readable format(s) and to generate operation instructions for one or more system(s) monitoring and / or controlling the distributed chemical production network. The system(s) for monitoring and / or controlling the distributed chemical production network may include distributed system(s) for monitoring and / or controlling components of the chemical production network. The system(s) for monitoring and / or controlling the distributed chemical production network may include system(s) configured to control and / or monitor production, system(s) configured to control and / or monitor warehouses, system(s) configured to control and / or monitor transport or the like or combinations thereof

[0062] The monitoring and / or controlling platform may be implemented on a cloud infrastructure configured for distributed computing or cloud computing. “Cloud computing'' may refer a model for enabling on-demand network access to a shared pool of configurable computing resources or nodes (e.g., networks, servers, storage, applications, and services). When distributed, cloud computing environments may be distributed internationally within an organization and / or across multiple organizations. In this example, the distributed cloud computing environment may contain the following computing resources: mobile device(s), operation equipment, applications, databases, data storages and server(s). The cloud computing environment may be deployed as public cloud, private cloud or hybrid cloud. A private cloud may be owned by an organization and only the members of the organization with proper access can use the private cloud through protected communication channels, rendering the access to data and usage of data in the private cloud similar to on-premises server usage. In contrast, data stored in a public cloud may be accessible over the internet. The hybrid cloud may be a combination of both private and public clouds and may allow to keep some of the data private while other data may be accessible over the internet. The term “computing resource or node" is defined herein broadly and may refer to 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. Computing resources or nodes are now increasingly taking a wide variety of forms. Computing resources or nodes may, for example, be handheld devices, production facilities, sensors, monitoring systems, control systems, appliances, laptop computers, desktop computers, mainframes, data centers, or even devices that have not conventionally been considered a computing node, such as wearables (e.g. , glasses, watches or the like). The memory may take any form and depends on the nature and form of the computing node or resource. The term “processor” may refer to an arbitrary logic circuitry configured to perform basic operations of a computer or system, and / or, generally, to a device which is configured for performing calculations or logic operations. In particular, the processor, or computer processor may be configured for processing basic instructions that drive the computer or system. It may be a semi-conductor-based processor, a quantum processor, or any other type of processor configures for processing instructions. As an example, the processor may comprise at least one arithmetic logic unit ("ALU"), at least one floating-point unit ("FPU)", such as a math coprocessor or a numeric coprocessor, a plurality of registers, specifically registers configured for supplying operands to the ALU and storing results of operations, and a memory, such as an L1 and L2 cache memory. In particular, the processor may be a multicore processor. Specifically, the processor may be or may comprise a Central Processing Unit ("CPU"). The processor may be a (“GPU”) graphics processing unit, (“TPU”) tensor processing unit, ("CISC") Complex Instruction Set Computing microprocessor, Reduced Instruction Set Computing ("RISC") microprocessor, Very Long Instruction Word ("VLIW") microprocessor, or a processor implementing other instruction sets or processors implementing a combination of instruction sets. The processing means may also be one or more special-purpose processing devices such as an Application- Specific Integrated Circuit ("ASIC"), a Field Programmable Gate Array ("FPGA"), a Complex Programmable Logic Device ("CPLD"), a Digital Signal Processor ("DSP"), a network processor, or the like. The methods, systems and devices described herein may be implemented as software in a DSP, in a micro-controller, or in any other sideprocessor or as hardware circuit within an ASIC, CPLD, or FPGA. It is to be understood that the term processor may also refer to one or more processing devices, such as a distributed system of processing devices located across multiple computer systems (e.g., cloud computing), and is not limited to a single device unless otherwise specified. The memory may refer to a physical system memory, which may be volatile, non-volatile, or a combination thereof. The memory may include non-volatile mass storage such as physical storage media. The memory may be a computer-readable storage media such as RAM, ROM, EEPROM, CD-ROM, or other optical disk storage, magnetic disk storage, or other magnetic storage devices, or any other physical and tangible storage medium which can be used to store desired program code means in the form of computer-executable instructions or data structures and which can be accessed by the computing system. Moreover, the memory may be a computer-readable media that carries computer- executable instructions (also called transmission media). Further, upon reaching various computing system components, program code means in the form of computer-executable instructions or data structures can be transferred automatically from transmission media to storage media (or vice versa). For example, computer-executable instructions or data structures received over a network or data link can be buffered in RAM within a network interface module (e.g., a “NIC"), and then eventually transferred to computing system RAM and / or to less volatile storage media at a computing system. Thus, it should be understood that storage media can be included in computing components that also (or even primarily) utilize transmission media.

[0063] To enable monitoring and / or controlling output materials produced by the distributed chemical production network in a simple, reliable, and robust manner, the methods, apparatuses, systems and uses disclosed herein are based on a three-prong approach: 1) Simplification of intake retrieval by consolidating data in a way that enables interaction with the request in natural language.

[0064] 2) Robust and reliable processing for triggering actions related to production, storage and / or transport of output materials.

[0065] 3) Robust and reliable action execution based on validation and autocompletion mechanisms.

[0066] Overall, the approaches described herein allow for easy, fast, and reliable reaction to requests for providing output materials (order processing). This way output materials can be made available in a fast and reliable manner to output material users.

[0067] Fig. 2 illustrates an example method for monitoring and / or controlling output materials produced or to be produced by a distributed chemical production network.

[0068] One or more request(s) to provide one or more output material (s) may be received from one or more output material user(s). The request may include one or more file / protocol type(s) with structured, semi-structured and / or unstructured data representing at least one or more output material(s) and quantity to be provided, produced and / or transported per output material. The request may further include one or more indication(s) of user(s) of output material, location as to where the output material to be used and / or time indication as to when output material is to be used.

[0069] The request may be received from different sending entities in different formats such as different file formats or via different communication protocols with different formats. One example may include decentral communication requests provided by one or more decentral network node(s) associated with one or more network(s) with network specific protocol. Requests may be provided via peer-to-peer communication in semi- or unstructured manner (not predefined data formats). Another example may include electronic requests provided as electronic document e.g. via an electronic communication channel such as e-mail or pdf documents. Another example may include audio-based requests provided by an audio system or a digital audio transcription service. The request may relate to one or more output material(s), one or more user(s) of the output material, one or more application (s) of the output material, one or more producer(s) of the output material, one or more quantity / les, such as pieces, amounts or volumes, per output material, one or more time specification(s) related to the providing of the output material, one or more delivery specification^) related to the providing of the output material, and / or one or more technical property / ies of the output material.

[0070] Classification and text extraction:

[0071] The request may be classified and routed to one or more extraction agent(s) configured to generate or extract a natural language or text representation such as txt as target format. The extraction agent(s) type 1, 2, ... n may transform the, per or each request from the provided data structure or format to a natural language or text representation. The extraction agent(s) type 1-n may extract a natural language or text representation from the one or more request(s) provided in one or more or multiple different format(s). The request may be classified by request type e.g. by data format, file type and / or protocol type. The request may be transformed based on the request type to a natural language or text representation. For example, requests of type e-mail in msg format may be provided to an extraction engine implemented as e-mail transformation engine. The e-mail plain text or the e-mail attachment(s) such as pdf documents may be converted to text or natural language format. Further for example, requests of type files documents such as pdf, doc or text may be transformed by optical character recognition (OCR) to machine encoded text and converted to text or natural language representation. Further for example, requests of type EDI (Electronic Data Interchange) with different format may be provided to an EDI transformation engine. EDI messages may relate to different standards such as ISO Electronic data interchange for administration, commerce, and transport (EDIFACT), ANSI X12, GS1 EDI, or the like. EDI messages may include documents related to identifiers that signify the type of transaction such as order, invoice or the like. The documents may include pre-defined header and details related to the identified type of transaction in standardized formats. The EDI messages may be transformed to intermediate representations such as xml or others and converted to text or natural language representation.

[0072] Processing of extracted text e.g. by using Large language Model (LLM, see Figs. 8-11):

[0073] Further for example, one or more language type(s) contained in the request such as Chinese or English may be recognized or detected. For language detection a classification model such as rule-based engines and / or probabilistic n-gram models trained on character distributions of one or more languages, e.g as implemented in opennlp package from Apache Foundation Version 2.3.3, may be used. One or more target language(s) may be pre-defined and text passages deviating from one or more target language(s) may be provided to a translation model such as transformer-based translation models e.g. as implemented by transformer-based models such as large language or foundation models like GPT4.0 or Gemini.

[0074] One or more instruction(s) for generating a common, semi-structured or structured and / or machine-readable representation including the natural language representation may be generated. The common, semi-structured or structured and / or machine-readable representation may be pre-defined. An instruction template relating to at least one task instruction configured to specify the task for the data-driven model to be executed may be provided. The instruction template may further include output instructions configured to specify format and / or structure of the common, machine-readable representation to be generated. The instruction template may further include one or more placeholders for inserting instruction context. The instruction may include at least one context placeholder for inserting the natural language representation as context. The instruction may include a context placeholder for inserting the request type as context. The context instruction may include the request type.

[0075] The instruction may include a context placeholder for inserting the historical pair(s) as context. The instruction context may include one or more historical pair(s) of requests and corresponding common, machine-readable representation. The one or more historical pair(s) of request and corresponding common, machine-readable representation may be retrieved from one or more context data bases storing historical pair(s) of request and corresponding common, machine-readable representation. The one or more historical pair(s) of request and common, machine-readable representation may be retrieved from one or more vector data bases storing historical pair(s) of request and corresponding common, machine-readable representation as embedded numerical vectors. The received request may be embedded in numerical vector. The one or more historical pair(s) of request and corresponding common, machine- readable representation may be retrieved based on a distance measure between the stored vectors of historical requests and the vector of the received request. The one or more historical pair(s) of request and common, machine- readable representation with closest distance between the stored vectors of historical requests and the received request may be selected and provided as context data.

[0076] The instruction may include a context placeholder for inserting additional data points related to the specific request as context. The instruction context may be enriched by one or more knowledge or context database(s). The knowledge or context database may store plain data points or embedded data points such numeric vector representations of text related to the request. For example, the knowledge or context database(s) may store data related to output materials, users of the output material, applications of the output material, producers of the output material, users of the output material, quantities, such as pieces, amounts or volumes per output material, time specifications related to the providing output material, delivery specifications, such as delivery date, packaging, address or type, related to the providing output material and / or technical properties of output material. For enrichment, one or more output materials, one or more users of the output material, delivery specifications, such as delivery date, packaging, address or type, related to the providing output material, one or more applications of the output material, one or more producers of the output material, one or more quantities, such as pieces, amounts or volumes, per output material, one or more time specification (s) related to the providing of the output material and / or one or more technical properties of the output material may be extracted from the request. The respective extraction may be embedded as e.g. described in the context of Figs. 8-11. Based on the extraction from the request, data stored in knowledge or reference database(s) may be retrieved. For example, if specific user of the output material or sender of order is extracted, additional data related to the specific user of the output material as stored in the knowledge or reference database may be retrieved. Further for example, stored requests processed in relation the specific user of the output material may be retrieved. The stored previous requests may be retrieved by finding the nearest neighbor embeddings of the current user's request and previous requests. Both current and previous requests may be embedded using the same embedding model. The embedding model may be pre-trained or fine-tuned to specific request type domain. The retrieved data from knowledge or context database(s) may be inserted into the instruction template as request specific context.

[0077] The instruction (s) for generating a common, semi-structured or structured and / or machine-readable representation of the request may be provided to at least one data-driven model configured, adapted, adaptable or suitable to at least one data-driven model configured to generate the semi-structured or structured and machine-readable representation of the request. The semi-structured or structured and machine-readable representation of the request may be based on a common or standardized data structure or format. Such data structure or format may for example be provided at least in part by the instruction (s) to the data-driven model. The machine-readable representation of the request may be based on a pre-defined common or standardized data structure or format. The machine-readable representation may be configured to monitor and / or control output materials produced or to be produced by the distributed chemical production network. The machine-readable representation may be configured to trigger production and / or transportation of the output material.

[0078] The instruction (s) for generating the common and machine-readable representation of the request may be provided to at least one data-driven model configured to generate the common and machine-readable representation of the request. The instruction(s) for generating the common and machine-readable representation of the request may be provided to a pre-trained, transformer-based model for mapping the text-based, natural language-based and / or unstructured instruction to machine-readable representation.

[0079] The data-driven model may include a general-purpose model trained at least on unstructured data. The data-driven model may be trained on unstructured natural language data and may be configured to process natural language. The data driven model may include a transformer-based model. The transformer-based model may include at least one processing layer configured to transform the instruction representation to an instruction vector representation, at least one processing layer configured to map the instruction vector representation to output vectors, and / or at least one processing layer configured to map the output vectors to the machine-readable representation. The data-driven model may be a general purpose model trained or parametrized on text data that is not specific to the task of generating common and machine-readable representation of the request.

[0080] The one or more instruction(s) may be provided to a data driven model together with configuration parameters for the data-driven model such as maximal token specification or authentication / authorization parameters The data-driven model may be selected from one or more data-driven models based on a sensitivity and / or confidentiality level of the request. Further details of possible data-driven models and their selection are described for example in the context of Figs. 8-11. The data-driven model may generate and / or provide semi-structured or structured and machine-readable representation of the request according to the instruction. For example, a semi-structured JSON representation including human readable or text-based key-value pairs may be generated. This way the request of different format(s) and / or type(s) may be brought into a common representation that may be further processed based on the machine- readable, semi structured, or structured data structure provided by the common representation.

[0081] Enrichment of common machine-readable representation:

[0082] The common representation may be enriched via one or more knowledge or reference database(s). The knowledge or reference database may store plain data points or embedded data points such numeric vector representations of text. For example, the knowledge or reference database(s) may store reference data related to output materials, users of the output material, applications of the output material, producers of the output material, quantities, such as pieces, amounts or volumes, per output material, time specifications related to the providing output material, delivery specifications, such as delivery date, packaging, address or type, related to the providing output material, and / or technical properties of output material. For enrichment, one or more output materials, one or more users of the output material, one or more applications of the output material, one or more producers of the output material, one or more quantities, such as pieces, amounts or volumes, per output material, one or more time specifications related to the providing of the output material, delivery specifications, such as delivery date, packaging, address or type, related to the providing output material, and / or one or more technical properties of the output material may be extracted from the common representation. The respective extraction may be embedded as e.g. described in the context of Figs. 8- 11. Based on the extraction from the common representation, data stored in knowledge or reference database(s) may be retrieved. For example, if specific output material is extracted, data related to the specific output material as stored in the knowledge or reference database(s) may be retrieved. Specifically in the field of chemistry, specific chemical names and associated material keys may differ for users of the output material and producers of the output material. In such instances, the knowledge database retrieval may serve as a reference database for providing the extracted part of the machine-readable representation in conformity with the output material producer’s systems configured to generate operational instructions. The retrieved data from knowledge or reference database(s) may replace the extracted part of the machine-readable representation. In other embodiment, syntactic and semantic analysis of the mismatched materials names may be used to identify the material names of the producer. This could be followed by identifying similar materials from the vector representation of the chemical inventory from the chemical product producer and to provide the match with closest distance. By resolving the material name(s) of the request to the material name(s) of the producer the material identifier(s) used by the production monitoring and / or control system of the producer may be provided for triggering providing of the output material The material identifier(s) used by the production monitoring and / or control system(s) of the producer may relate to the material and the production plant the material is to be provided from or produced by.

[0083] Beyond the material information, output material user and / or sender information existing in the knowledge or reference database may be used to enrich the machine-readable representation in conformity with the output material producer's systems, e.g., shipping and delivery information like packaging, transport, addresses or the like. In addition to harmonize the incoming messages and mapping to output material producer's data points, enrichment could be carried out using an agent-based process that based on the observed data types fetches the normalized, harmonized or standardized information from one or more knowledge or context data bases of the output material producer.

[0084] Validation of common machine-readable representation:

[0085] The enriched common representation may be validated based on one or more representation rules associated with the data contained in the enriched common representation. For validation a validation templates may be provided. The validation template may include in natural language task instructions for the data-driven model to validate the enriched common representation. The validation template may include at least one placeholder for the enriched common representation to be validated. The validation template may relate to rules to be checked on validation provided in natural language or text. The validation template may relate to correction instructions in case at least one error is found upon validation. The validation template may relate to one or more process trigger(s) in case at least one error is found upon validation and cannot be resolved. The one or more process triggers) may depend on a validation failure type, e.g. the part of the common representation which caused the failure of validation. The validation instruction may include system instructions relating to the output structure or conditions to be fulfilled by the output or the instruction to restrict execution to fully defined enriched common representation. The validation template may be used to generate one or more validation instruction(s) to be provided to the data-driven model. Upon receipt of the response from the data-driven model a validated enriched common representation may be provided.

[0086] The common representation may not be validated or rejected e.g. based on one or more pre-defined rule(s). The predefined rules may relate to the data specified in the request, missing data fields, syntax errors contained in the common representation or the like. For example, after enrichment the conformed or common representation may be defined to include mandatory fields or data necessary for processing the common representation. If validation of mandatory data fails, a failure of validation may be provided. If some of the mandatory fields or data are missing, the validation may provide a process trigger for sending automatic messages to the sender to clarify and / or correct the original request. Based on the failure of validation one or more process trigger(s) in accordance with the validation failure type may be provided. For example, if the common representation relating to the output material user or order sender caused the failure, a message to the output material user may be triggered to provide the missing data. Further for example, if the common representation relates to an output material specification that cannot be matched to an output material produced by the producer, a message to the output material user may be triggered to provide the missing data. For such message the format or protocol types used for the request may be used. Upon receipt of the missing data, the missing data may be inserted into the common representation. The thus amended common representation may be validated as described above. Further for example, if the common representation relating to the output material user or order sender caused the failure, the process trigger may relate to retrieval of the missing data from one or more knowledge databases storing such information.

[0087] Further for example, the validation may require the common representation to fulfill predefined rules such as derivable from local or global regulations. If validation of predefined rules fails, a failure of validation may be provided as process trigger. The common representation and the request may be rejected. If validation of predefined rules fails, a failure of validation may be provided as process trigger. The process trigger may relate to an alternative output material and / or output material packaging by triggering a message to the output material user to confirm the alternative output material and / or output material packaging. Upon receipt of confirmation, the alternative output material and / or output material packaging may be inserted into the common representation. The thus amended common representation may be validated as described above.

[0088] Action trigger for production and / or transportation:

[0089] A monitoring and / or controlling action to provide output materials produced to be produced by the distributed chemical production network to output material user may be triggered based on the common representation. For example, the common representation may be configured to comply with a data format or schema that is suitable for direct or indirect ingestion by monitoring and / or controlling systems of the distributed chemical production environment. The common representation may for example include a data structure format or schema compatible with a data structure format or schema of monitoring and / or controlling systems of the distributed chemical production environment. This way the distributed chemical production environment may be monitored and / or controlled in a reliable and effective manner based on the request received by users of the output materials. Through the direct actionable system communication, the chemical production network may be controlled and / or monitored in accordance with demand received by output material users. Thus, supply and demand challenges around inventory levels, use of production capacities and / or market demand can be solved and the chemical production network can be operated in a highly flexible while still reliable manner.

[0090] Fig. 3 illustrates an example apparatus for monitoring and / or controlling output materials produced or to be produced by a distributed chemical production network.

[0091] The computing apparatus may be configured to monitoring and / or controlling output materials produced or to be produced by a distributed chemical production network. In particular, the computing apparatus may be configured to perform the steps as described in the context of Fig. 2. The computing apparatus may include a monitor- ing / controlling apparatus communicatively coupled to one or more data storage(s) providing knowledge data bases for context or reference functionalities, one or more template store(s), one or more validation processing systems and / or one or more monitoring and / or controlling systems S1 ..n associated with the components of the distributed chemical production network. The computing apparatus may include and / or be communicatively coupled to one or data driven model interfaces configured to access a data driven model and to provide the result of the processing by the data-driven model.

[0092] The computing apparatus may include one or more request ingestion interface(s) configured to obtain one or more request(s) to provide one or more output materials from one or more output material users. The monitoring and / or controlling apparatus may include a routing engine configured to classify the request and configured to routing the request to one or more extraction engine(s). The monitoring and / or controlling apparatus may include one or more extraction engine(s) configured to extract a natural language or text representation of the request. The monitoring and / or controlling apparatus may include an instruction agent configured to generate one or more instruction(s) for generating a common, semi-structured or structured and / or machine-readable representation including the extracted text or natural language representation. Example instructions are illustrated in the insets on the bottom (dotted arrows). The monitoring and / or controlling apparatus may include or be communicatively coupled to one or more model execution engines for generating model results based on provided instructions. The monitoring and / or controlling apparatus may include an enrichment agent that may access one or more database(s) for enriching the machine- readable representation for monitoring and / or controlling output materials produced or to be produced by the distributed chemical production network. The monitoring and / or controlling apparatus may include a validation engine that may be communicatively coupled to one or more validation processing system(s) for validating the machine-readable representation for monitoring and / or controlling output materials produced or to be produced by the distributed chemical production network. The monitoring and / or controlling apparatus may include a monitoring and / or controlling engine communicatively coupled to one or more monitoring and / or controlling system(s) S1...n configured to monitor and / or control components of the distributed chemical production network.

[0093] Fig. 4 illustrates an example of data transmissions in relation to the method steps illustrated in Fig. 3.

[0094] Multiple requests 1 , 2, .., n may be received for classification with respect to the type of the request. The classification may depend on the format and / or schema type. Fig. 5 illustrates an example of an EDI 850 message as request type and the extracted text from such EDI 850 message. Depending on the type of the requests or depending on the type per request or the classification based on the on the type per request, the requests may be routed to text extraction suitable for the respective request type. For example, the EDI 850 message may be routed to a different text extractor than the plain email. Further for example, the pdf format purchase order may be routed to a different text extractor than the plain email or EDI 850 message.

[0095] After text extraction, instructions to be provided to a large language model (LLM) may be generated. The generation may include accessing the knowledge or context data base to retrieve further instruction context. The generated instructions may be provided to the Large Language Model (LLM) such as GPT-4 or Gemini. The LLM may provide the common representation as for example illustrated in Fig. 6, top based on the generated instructions. The common representation may be enriched by accessing knowledge or reference database(s) as for example illustrated in Fig. 6, middle, where for example the material ID field is matched to the producer's material ID and replaced. The enriched common representation may be validated by using an LLM The validation may optionally include triggering validation actions through validation processing system(s) or accessing knowledge or reference database(s) such as the addition of address details. The validated common representation as for example illustrated in Fig. 6, bottom may be provided to systems configured to monitor and / or control components of the distributed chemical production network.

[0096] Fig. 5 illustrates example request type and corresponding text extraction.

[0097] The request may relate to an EDI 850 message or document according to EDI X12 standard. The message may be received through any communication protocol such as VAN (Value-Added-Network; private, hosted service), AS2 (Applicability Statement 2; secure messaging over the Internet using digital certificates and encryption), (s)FTP ((se- cure)File Transfer Protocol; standard network protocol built on a client-server architecture), HTTPS (Hypertext Transfer Protocol Secure; an extension of the HTTP and a secure way for communication over the Internet) or the like. Different EDI X12 requests transferred over different communication channels may be received.

[0098] An example request relating to receiver ID, sender ID, product ID, quantity is schematically shown in Fig. 5. The request adheres to standard format including Header, Functional Groups, Document and Summary. Based on the standard format the different elements of the EDI 850 message may be provided. The text extraction of the different elements may include converting the EDI 850 message to a text format in this example a XML format as illustrated in Fig. 5, bottom.

[0099] Fig. 6 illustrates examples of common representations prior and after enrichment and / or validation.

[0100] The common representation may relate to a JavaScript Object Notation (JSON) format as for example described in internet standards RFC 8259 or ECMA 404. JSON is a lightweight, text-based, language-independent data interchange format that defines a set of formatting rules for the portable representation of structured data. The JSON shown in Fig. 6 shows an object including multiple name / value pairs. Name is a string. Value is a string, number, boolean, null, object, or array. Strings represented in a JSON text are composed of Unicode characters.

[0101] The top illustration of Fig. 6 shows a JSON format including selected values of the order object prior to enrichment and / or validation. The middle illustration shows the same JSON with replaced materiallD value after enrichment and prior to validation. The bottom illustration shows the same JSON with address value received by the sender of the request. The bottom illustration shows the JSON after validation triggered the process for requesting the address from the sender of the request. The communication channel the request was received via may be used for such process.

[0102] Fig. 7 illustrates example production facilities controlled and / or monitored based on received requests.

[0103] The distributed chemical production network may include multiple production facilities. The production facilities may be connected by material streams. For example, one production facility may produce an intermediate product as output material that may be used as input material by one or more subsequent production facility / ies. Further for example, one production facility may produce a byproduct as output material that may be used as input material by one or more preceding production facility / ies.

[0104] The production facilities may be associated with production systems configured to monitor and / or control the production of the production facility. The production systems may be communicatively coupled to the monitoring and / or controlling apparatus as e.g. described in the context of Fig. 3 or an intermediate connector that connects the production systems with the monitoring and / or controlling apparatus as e.g. described in the context of Fig. 3. Through the example architecture illustrated in Fig. 7, the requests to provide output material may be received by the monitoring and / or controlling apparatus. The requests may be processed by the monitoring and / or controlling apparatus to provide machine-readable representation for triggering a monitoring and / or controlling action to provide output materials produced by the distributed chemical production network to output material user. The connector may receive the machine-readable representations and aggregate the quantity per output material to be provided to a total quantity to be provided per output material. The connector may determine the facility to be triggered based on the output mate- rial to be produced. The connector may provide the trigger for production to the determined facility. The trigger may include for example the output material and the total quantity to be produced.

[0105] The distributed chemical production network shown on Fig. 7 is merely one example and shall not be considered limiting. Distributed chemical production networks may include further facilities. The facilities may be segregated or non-segregated with respect to material flows. The facilities may include production facilities, logistics facilities, warehouses or the like.

[0106] Figs. 8-11 illustrate an example transformer architecture of a general-purpose model. Fig. 8 illustrates an embodiment of training an embedding layer.

[0107] The embedding layer may be obtained by training for example a continuous bag of words model (CBOW) or a skipgram 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 814 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 806. In particular, the embedded input 814 and / or the input vector 806 may be machine-readable and / or processable by a processor. For this purpose, the embedded input 814 and / or the input vector 806 may be a tensor, in particular a first-rank tensor. Specifically, the input vector 806 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 808, 810 and 812. The entries unequal to zero in the one hot vector and / or in the input vector 806 may indicate the element. For example, a look up table may define the relation between the position of the entries unequal to zero and the element indicated by the one hot vector. The look up 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.

[0108] A look up table specifying a subset of the vocabulary size eg of the English language may comprise 10,000 words or more. The embedded input 814 may be a lower-dimensional representation than the input vector 806. For example, typical embedded inputs 814 may comprise some hundreds of different entries. Followingly, the embedded inputs 814 constitute a densified representation of one or more elements using less computational resources. More than that, the embedded input 814 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 inputs 814 may be the more similar the two elements associated with the embedded inputs 814 may be. Hence, the embedded inputs 814 may represent one or more elements accurately and lead to accurate results based on processing the embedded inputs 814.

[0109] For transforming the input vector 806 into the embedded input 814, the embedding layer may comprise a number of neurons equal to the number of entries in the embedded input 814. Based on the embedded inputs 814, the output layer may generate the output vector 816. The output vector may be a vector and / or may indicate one or more elements. The output vector 816 may indicate one or more elements different from the input vector 806 and / or the one hot vectors associated with the input vector 806. For this purpose, the output layer may comprise a number of neurons equal to the number of entries of the input vector 806 and / or the output vector 816. The output layer may apply a softmax function to the embedded inputs 814. By doing so, the output vector may comprise the probabilities associated with the elements associated with the entries of the output vector 816 unequal to zero. Hence, from the output vector 816 one or more elements may be obtained with a corresponding probability. Where the input vector 806 may specify one or more sequence(s) of elements, the output vector 816 may specify one or more elements corresponding to the sequence(s) of elements specified by the input vector 806. In the example of FIG. 8, the element associated with vector 818 may correspond to the input vector with 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 802 and the output layer 804 may refer to the most probable elements indicated by the output vector 816. Hence, the model depicted in FIG. 8 may generate the element associated with the vector 818 with a probability determined by the data driven model or in case of a calibrated model a confidence score of 71 %.

[0110] The model of FIG. 8 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 labeled, 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 806 through the model to the output vector corresponding to the input vector 806 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 802 and the output layer 804 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 802 may be suitable for embedding input data comprising one or more elements. This embedding layer 802 may be used in other machine-learning architectures requiring an embedding layer 802 such as a transformer encoder, transformer decoder or transformer encoder decoder architecture as described within the context of FIG. 9A, FIG. 9B and FIG. 9C. For training these architectures, a trained embedding layer 802 may be required. Hence, a model such as a CBOW model may be trained prior to training the transformer encoder, transformer decoder or transformer encoder decoder architecture.

[0111] Fig. 9A illustrates an embodiment of a transformer encoder architecture.

[0112] The transformer encoder comprises an encoder input 978, one or more encoder blocks 974, 914 and an encoder output. The transformer encoder architecture may be derived from the transformer encoder-decoder architecture as known in the art and shown in FIG. 9C In particular, the transformer encoder may be referred to as X-former. The transformer encoder architecture may correspond to the encoder architecture associated with the transformer encoder-decoder architecture with an additional encoder output instead of connecting the encoder block directly to the decoder of the transformer encoder-decoder architecture. A plurality of transformer encoder architectures is available in the art such as the bi-directional encoder representations from transformers (BERT).

[0113] The input data may be received at the encoder input 978. The input data may be contextualized chemical product data. The encoder input 978 may apply an input embedding 902. Applying the input embedding 902 may refer to passing the input data through an embedding layer eg as described within the context of FIG. 8. Applying the input embedding 902 to the contextualized chemical product data may result in embedded contextualized chemical product data.

[0114] The encoder input 978 may apply positional encoding 904. Applying positional encoding 904 may refer to adding a positional factor to the embedded input obtained via input embedding. Preferably, the input data may specify a sequence of elements. The positional factor Ppos may be indicative of the position of the elements within the sequence. For example, the positional factor Ppo» may be obtained based on the following equation: where pos may refer to the position of the element within the sequence, I may refer to the dimension associated with the input embedding and d may refer to the dimension of the model, eg transformer decoder, transformer encoder or transformer encoder-decoder. This may be referred to as absolute positional embeddings. Alternatively, the positional encoding may be based on rotary positional embeddings (RoPE). Positional encoding is beneficial since it enables the processing of sequential data without requiring further dimensions indicating the position of each element. Followingly, the positional encoding 904 reduces the computational resources needed for embedding the input data. By passing the input data through the encoder input, the input data may be transformed into a second-rank tensor representing the sequence of elements. This second-rank tensor may be referred to as embedded input data. The embedded input data may be processed by the encoder block. The embedded input data may be provided to the layer normalization 908 by a residual connection. Multi-head self-attention 906 may be applied to the embedded input data. Multi-head self-attention 906 may comprise the two components multi-head and self-attention. Self-attention may be understood as being a filter applied to the embedded input data. By applying the filter to the embedded input data, the elements associated with the embedded input data contributing to the to be generated output data may be identified for generating the output data. Hence, the filter may represent the degree of contributing to the to be generated output data by the elements associated with the embedded input data. Applying the filter may be referred to as weighting the elements associated with the embedded input data. This is advantageous specifically regarding long sequences of elements. The filter may be learned and improved during the training by learning to identify the contribution of elements associated with the embedded input data. For example, in the partial sentence “I went to the bakery to buy a” the last word may be generated by the data-driven model such as the transformer encoder. The self-attention may focus the transformer encoder to attend to the word "bakery" and “buy” mostly to generate the word “bread”. Self-attention may refer to attention generated based on the input data. Hence, the filter may be determined based on the input data, preferably the embedded input data. The embedded input data may serve as query Q, key K and value V with respect to the self-attention operation. The self attention may refer to attention based on the received input data. Hence, the filter may be calculated based on the following formula by inserting the respective tensors based on the embedded input data: where dft corresponds to the dimension of the key.

[0115] For improving the efficiency of the transformer encoder further, the multiple heads are used to apply the filter resulting in the multi-head self-attention 906. Multi-head self-attention 906 may comprise applying the filter to two or more parts of the embedded input data. Hence, the tensor may be split into two or more parts and the filter may be applied to the two or more parts separately by two or more heads according to the following equation:head i =Attention (QWtQ, KW,K, VWTy) with parameter matrices i may refer to the number of heads, dy, d^ and may refer to the dimensions of the value, key and query.

[0116] The result of the two or more head may be concatenated according to the following equa-tion :MultiHead (Q, K, V) = Concat (head 1, . . . , headh) W° e jjWs,-xd and h may refer to the number of heads. The embedded input data may be transformed via the multi-head self-attention 906 into a context tensor. The context tensor may represent the sequence of elements and the relation between two or more elements of the input data. Transforming the embedded contextualized chemical product data may result in chemical product and environmental property data context tensor 622. Transforming the embedded contextualized chemical product data may result in chemical product data context tensor 620. Hence, the context tensor may be chemical product data context tensor 620 and / or chemical product and environmental property data context tensor 622. The context tensor may be a second rank tensor and / or may comprise one or more first rank tensor(s). After the multi-head self-attention 906 layer normalization 908 may be applied based on the context tensor and / or the embedded input data from the residual connection. Applying layer normalization 908 may refer to normalizing the context tensor. Normalizing the context tensor may lower the values of the entries of the context tensor. This reduces the computational cost associated with processing the context tensor. Layer normalization 908 may be followed by passing the context tensor to a feed forward layer 910 again followed by layer normalization 912 based on the residual connection to the context tensor and / or the output of the feed forward layer 910. The feed forward layer 910 may be a feed-forward neural network. The feed-forward neural network may comprise of a plurality of fully connected neurons. Passing the context tensor through the feed-forward neural network may result in transforming the context tensor linearly. Additionally or alternatively, the neural network may comprise one or more activation functions such as a rectified linear unit (ReLU). Hence, the neural network may be configured for performing one or more non-linear operations to the context tensor and / or transforming the context tensor non-linearly. After the context tensor has been transformed and / or normalized by the feed forward layer 910 and the layer normalization 912, the context tensor may be provided to one or more further encoder blocks 914 Having passed the context tensor through the feed forward layer 910 may adapt the context tensor for the processing by a further attention layer of the one or more further encoder blocks 914 for applying a self-attention filter, preferably multi-head self-attention 906. The context vector after being transformed by the layer normalization 912 and the feed forward layer 910 may be referred to as hidden state.

[0117] The encoder output 976 comprises of a linear layer 916 and a softmax layer 918. The linear layer 916 may transform the context vector into a logits vector The linear layer may be fully-connected. The logits vector obtained by passing the context tensor through the linear layer 916 may be passed through the softmax layer 918. Passing the logits vector through the softmax layer 918 may refer to applying the softmax function to the logits vector. Applying the softmax function to the logits vector may result in a probability distribution of one or more elements corresponding to the sequence of elements in the input data. From the probability distribution based on predefined selection criteria, one or more elements may be chosen. The one or more chosen elements may be referred to as the one or more elements generated by the transformer encoder. The one or more generated elements may be provided to the encoder input for generating further one or more elements corresponding to the sequence of the input data and the one or more elements generated by the transformer encoder as described within the context of FIG. 10. The output data from the encoder output 976 may be chemical product production and / or processing data. Hence, the result of transforming the chemical product data context tensor 620 may be chemical product production and / or processing data.

[0118] FIG. 9B illustrates an embodiment of a transformer decoder architecture. Input data, embedded input data, context tensor and / or output data may be as defined within the context of FIG. 9A.

[0119] The transformer decoder comprises a decoder input 984, one or more decoder blocks 980, 932 and a decoder output 992. The transformer decoder architecture may be derived from the transformer encoder-decoder architecture as known in the art and shown in FIG. 9C. The transformer decoder may be referred to as X-former. The transformer decoder architecture may correspond to the decoder architecture associated with the transformer encoder-decoder architecture independent of receiving one or more hidden states from the encoder of the transformer encoder-decoder. A plurality of transformer decoder architectures is available in the art such as the generalized pretrained transformers (GPT).

[0120] The decoder input 984 may apply input embedding 920 and positional encoding 922 analogous to analogous to the input embedding 902 and the positional encoding 904 as described within the context of FIG. 9A.

[0121] The decoder block 980 may comprise the layer normalizations 926, the masked multi-head self-attention 924, the feed forward layers 928 and / or the layer normalization 930. The embedded input data resulting from passing the input data through the decoder input 984 may be provided to the layer normalization 926 via a residual connection. Further, masked multi-head self-attention 924 may be applied to the embedded input data.

[0122] Masked multi-head self-attention 924 corresponds to the multi-head self-attention 906 as described within the context of FIG. 9A with additionally masking a part of the embedded input data associated with elements later in the sequence than the element to be generated. Additionally or alternatively, the part of the input data associated with elements later in the sequence than the element to be generated may not be received and / or transformed into the embedded input data. Thus, the transformer decoder may be suitable for generating a subsequent element to a sequence, whereas the transformer encoder may be suitable for generating a missing element in within one sequence and / or between two or more sequences. Therefore, the transformer encoder may be configured for classification tasks. The transformer decoder may be configured for text generation.

[0123] Similar to the transformer encoder as described within the context of FIG. 9A, a context tensor may be generated by applying the masked multi-head self-attention 924 and the layer normalization 926. The context tensor may be provided to the layer normalization 930 via a residual connection. Further, the feed forward layer 928 and the layer normalization 930 may be analogous to the feed forward layer 910 and the layer normalization 912 as described within the context of FIG. 9A. The context tensor may be provided to one or more further decoder blocks 932.

[0124] The decoder output 992 may comprise of a linear layer 934 and a softmax layer 936. The linear layer 934 and the softmax layer 936 may be analogous to the linear layer 916 and the softmax layer 918 as described within the context of FIG. 9A.

[0125] FIG. 9C illustrates an embodiment of a transformer encoder-decoder architecture. Input data, embedded input data, context tensor and / or output data may be as defined within the context of FIG 9A.

[0126] The transformer encoder-decoder may comprise the encoder input 988, the one or more encoder blocks 986, 964, the decoder input 994, the decoder block 990 and the decoder output 992. The encoder input 988 may correspond to the encoder input 978 of FIG. 9A. The one or more encoder block 986, 964 may correspond to the one or more encoder blocks 974, 914 of FIG. 9A. The decoder input 994 may correspond to the decoder input 984 of FIG. 9B.

[0127] The decoder block 990 may comprise a masked multi-head self-attention 970, a layer normalization 972, a feed forward layer 938 and a layer normalization 940 analogous to the masked multi-head self-attention 924, the layer normalization 926, the feed forward layer 928 and the layer normalization 930 as described within the context of FIG. 9B. The decoder block 990 may further comprise a multi-head self-attention 950 and a layer normalization 948. Analogous to the description of FIG. 9B, the context tensor may be obtained from the masked multi-head self-attention 970 and the layer normalization 972. Multi-head self-attention 950 analogous to the multi-head self-attention 906 of FIG. 9A may be applied to the context vector obtained from the layer normalization 972 and the hidden states of the one or more encoder blocks 986, 964. Layer normalization 948 may be applied to the context vector obtained from the multi-head self- attention 950 and the context vector obtained from the layer normalization 972 provided via a residual connection. The context vector resulting from the layer normalization 948 may be processed via the feed forward layer 938 and the layer normalization 940 analogous to the description of FIG. 9B. The context vector resulting from the layer normalization 940 may be provided to further decoder blocks 942 analogous to the decoder block 990. The context vector obtained from the one or more decoder blocks 990, 942 may be provided to the decoder output 992. The decoder output 992 may correspond to the decoder output 982 of FIG. 9B.

[0128] With the above-described architecture, the transformer encoder-decoder may receive and process input data at the encoder input 988 and the one or more encoder blocks 986, 964 and the decoder block 990 and the decoder output 992. Based on the input data, the transformer encoder-decoder may generate output data part by part or sequentially. The sequentially generated output data may be provided to and / or may be processed by the decoder input 994, the one or more decoder blocks 990, 942 and the decoder output 992. Preferably, a sequence may be provided to the encoder input 988 and after having generated at least a part of the output data, the decoder input 994 may be provided with at least the part of the elements of the output data already generated. By doing so, the next elements of the output data may be generated with a higher accuracy by taking the input data and the generated output data into account since more data is received by the transformer encoder-decoder may be received over time.

[0129] Because of the transformer encoder-decoder architecture, the transformer encoder-decoder may be configured for transforming a sequence into another representation of the sequence. An example for transforming one sequence into another representation may be translation of one sentence into another language. A plurality of transformer encoder-decoders are available in the art such as BART, T5 or the like.

[0130] In an embodiment, the layer normalization 908, 912 may be applied prior to the masked multi-head selfattention 924, multi-head self-attention 906 and / or the feed forward layer 910 in the transformer decoder, the transformer encoder and / or the transformer encoder-decoder. By doing so, the computational resources for applying the multi-head self-attention 906 and / or the feed forward layer 910 to the embedded input data and / or the context tensor may be decreased as the entries of the respective tensors may be lower after normalization.

[0131] In an embodiment, the decoder output 992 may comprise of a classification neural network, further feedforward layers, convolutional layers, fully connected layers or the like. For example, the transformer encoderdecoder may be configured for choosing between a plurality of options. For this purpose, the transformer encoder-decoder may be provided with three different input data sets and may classify the context vectors obtained from the one or more decoder blocks 990 via one or more linear layers. Followingly, the architecture may be extended depending on the use case to be solved.

[0132] FIG. 10 illustrates an embodiment of training and / or deploying the transformer encoder, the transformer decoder and / or the transformer encoder-decoder.

[0133] The encoder / decoder / encoder-decoder architecture 1002 may correspond to the transformer decoder, the transformer encoder and / or the transformer encoder-decoder as describe within the context of FIG. 9A- FIG. 9C. Input data, embedded input data, context tensor and / or output data may be as defined within the context of FIG. 9A. The output data generated by the encoder / decoder / encoder-decoder architecture 1002 may comprise of one or more elements, in particular a sequence of elements. The previously generated elements of the output data may be provided as input for generating the next element in the sequence of the output data.

[0134] In the example of FIG. 10, the input data may comprise of N elements, in particular input tokens. An input token may be a token dedicated to be inputted into a data-driven model such as the transformer decoder, the transformer encoder or the transformer encoder-decoder. The output data to be generated may comprise of M elements. The encoder / decoder / encoder-decoder architecture 1002 may generate one element of the output data based on receiving the input data and optionally previously generated elements of the output data at a timestep. Hence, for generating M elements M time steps are required. A time step comprises of providing input 1010, 1012, 1014 to the encoder / decoder / encoder-decoder architecture 1002 and receiving output data 1004, 1008, 1006 from the encoder / decoder / encoder-decoder architecture 1002. In a first timestep, the input 1010 may comprise of N input tokens. The N input tokens may be associated e.g. with N words, stems or endings. Preferably, the N input tokens may specify a question. One or more input tokens may specify the beginning of the sequence of tokens and / or the end of the sequence of tokens. The input 1010 may be processed by the encoder / decoder / encoder-decoder architecture 1002. Based on the input 1010 at least a part of the output data 1004 may be generated. The at least a part of the output data may comprise a first output token. In the next timestep, the generated first output token may be provided together with the input 1012. Specifically, where the input 1012 may be received by a transformer encoder-decoder the input tokens may be received at the encoder input 988 and the first output token may be received at the decoder input 994. Where the input 1012 may be received by the transformer encoder, the input 1012 may be received by the encoder input 978 and analogously regarding the transformer decoder and the decoder input 984. Based on the input 1012, the output data 1008 comprising the first output token and a second output token may be generated. Generating the output data 1008 based on the input 1012 may refer to generating the second token based on the first token and the N input tokens, wherein the first token may have been generated based on the N input tokens This process may be repeated until the last token in the sequence of the output data 1006 may be generated. Preferably, the last token may be an end token. The end token may terminate the generation of a further output token.

[0135] Similarly, to the data processing during deployment of the encoder / decoder / encoder-decoder architecture 1002, the encoder / decoder / encoder-decoder architecture 1002 may be trained. The training data set may comprise a plurality of sequences comprising a plurality of elements. The sequences may be associated with the input data and / or the output data. Additionally or alternatively, the sequences may be independent of the input data and / or the output data. For example, where the input data and the output data may refer to chemical compositions represented via text, the training data set may comprise sequential text data independent of chemical compositions. In this example, the training data set may comprise sequences of words originating from a conversation. In an embodiment, the training data set may comprise at least partially input data sets and / or output data sets.

[0136] The training may be initialized by initializing the encoder / decoder / encoder-decoder architecture 1002. In an embodiment, the parameters associated with the encoder / decoder / encoder-decoder architecture 1002 may be initialized randomly. Additionally or alternatively, the input embedding of the encoder / decoder / encoder- decoder architecture 1002 may be obtained by training a CBOW model or a skip gram model as described within the context of FIG. 8. The trained embedding layer may be used during training. The parameters associated with the embedding layer may be kept constant and / or may be updated after a predefined number of training epochs. By doing so, the number of parameters to be updated is lower enabling a faster and less computational resources-consuming training. Further, the accuracy associated with the embedding layer may be constant and / or may be increased by avoiding error compensation in relation to the just initialized encod- er / decoder / encoder-decoder architecture 1002.

[0137] During the training of the encoder / decoder / encoder-decoder architecture 1002, at least a part of the sequences of the training data set may be provided to the encoder / decoder / encoder-decoder architecture 1002 one by another and one or more elements may be generated based on the sequences of the training data set one by another. The elements generated based on the sequences may follow the elements of the parts of sequences the encoder / decoder / encoder-decoder architecture 1002 may have been provided with. The generated one or more elements may be compared to the one or more elements following the at least a part of the sequences provided to the encoder / decoder / encoder-decoder architecture 1002 as specified by the training data set. Hence, during the training the encoder / decoder / encoder-decoder architecture 1002 may generate a guess on the next element and the guess on the next element in a sequence may be compared to the ground truth specifying the actual next element according to the training data set. Based on the guess on the next element and the ground truth a loss may be determined. The loss may define the similarity between the guess on the next element and the ground truth. The loss may be determined by forming a vector dot product between the token associated with the one or more elements and the token associated with the ground truth. A loss unequal to zero may result in updating the parameters associated with encoder / decoder / encoder-decoder architecture 1002. Preferably the parameters associated with the encoder / decoder / encoder-decoder architecture 1002 may be independent of the embedding layer. For example, the parameters associated with the encod- er / decoder / encoder-decoder architecture 1002 may be weights of the neurons of the encod- er / decoder / encoder-decoder architecture 1002.

[0138] Based on the determined loss, backpropagation may be applied to determine the gradients associated with the parameters of the parameters associated with encoder / decoder / encoder-decoder architecture 1002 to lower the loss. According to the determined gradients, the parameters associated with the encod- er / decoder / encoder-decoder architecture 1002, preferably the weights of the neurons associated with the encoder / decoder / encoder-decoder architecture 1002, may be updated by using a gradient descent algorithm.

[0139] The training data set may be unlabeled. The sequences of elements within the training data set may inherently comprise the ground truth for determining the loss with respect to the one or more elements generated during the training of the encoder / decoder / encoder-decoder architecture 1002. Hence, the encoder / decoder / encoder- decoder architecture 1002 may be trained self-supervised. This is advantageous since time and resources for creating a labeled training data set may be saved. Furthermore, this enables the usage of large training data sets associated with a size of several tera bytes. Consequently, the data-driven model may be accurate in generating elements of a sequence. In addition, the large training data set enables few shot predictions or even zero shot predictions. Hence, the data-driven models trained as described above are versatile contributing to saving resources needed for training and / or hosting a plurality of purpose-driven models such as CNNs. The training described above may be referred to as pretraining. The data-driven model may be configured for performing few shots or even zero shot predictions with respect to a plurality of use cases after pretraining. The performance of the data-driven model may be increased further by additional training referred to as finetuning.

[0140] FIG. 11 illustrates an embodiment of input embedding. Input data, embedded input data, context tensor and / or output data may be as defined within the context of FIG. 9A.

[0141] Where the sequence of elements associated with the input data, preferably comprised in the input data, may be of one type, the input embedding 902, 920, 952, 966 as described within the context of FIG. 9A - 2C may be used. For example, a type of input data may be text where the elements may be associated with at least a part of a word, a punctuation character, a start token specifying the beginning of one or more sequences associated with the input data and / or the end token. In another example, the input data may be at least partially numerical. Hence, the input data may comprise a plurality of numbers. Numerical input data may be for example tabular data. Tabular data may specify one or more rows and / or one or more columns. Hence, the tabular data may comprise one or more cells, wherein the cells may be associated with one or more numerical values.

[0142] Numerical input data may require a different embedding than text input data. Input embeddings for numerical input data may comprise a token embedding, a positional embedding, a column embedding, a row embedding or a combination thereof.

[0143] Applying a token embedding to one or more elements, in particular tokens associated with the input data may result in a machine-processable representation associated with the one or more elements, in particular tokens. Applying the token embedding to one or more elements may refer to passing the one or more elements through the embedding layer, eg as described within the context of FIG. 8. Hence, token embeddings may specify the one or more elements, in particular tokens in a machine-processable representation. For example, the token embedding may transform a numerical value into a vector. This is advantageous since this representation can be enriched by further information such as the position of the token within the sequence and / or within a table associated with the sequence of tokens. The positional embedding may be analogous to the positional embedding as described within the context of FIG. 8, FIG. 9A-2C. Where the input data may be tabular data, column embedding may be applied. Applying a column embedding to one or more elements, in particular tokens associated with the input data may result in a machine-processable representation specifying the location of the one or more elements within a table 1102, preferably within the columns of the table 1102. Applying the column embedding may refer to adding a column factor to the input data embedded via token embeddings, in particular the embedded input data. The column factor may be the same for elements associated with the same column and / or may differ between two or more elements associated with different columns. Analogous, row embeddings may be applied where the input data may be tabular data. Applying a row embedding to one or more elements, in particular tokens associated with the input data may result in a machine-processable representation specifying the location of the one or more elements within a table 1102, preferably within the rows of the table 1102. Applying the row embedding may refer to adding a column factor to the input data embedded via token embeddings, in particular the embedded input data. The row factor may be the same for elements associated with the same row and / or may differ between two or more elements associated with different rows.

[0144] In an embodiment, input data may be at least partially numerical and at least partially text. Hence, the input data may comprise two or more types of data. A type of data may refer to a modality. Followingly, different embeddings may be applied to the input data. To parts of the input data comprising text the input embedding referred to in FIG. 8, FIG. 9A-2C may be applied. To parts of the input data being numerical token embeddings, positional embeddings, column embeddings and row embeddings may be applied. Further, segment embeddings may be applied to the input data independent of the type of input data. The segment embedding may specify the type of input data one or more elements may be associated to. For example, if the input data comprises of text and numbers, the input data may comprise of two types of input data. Applying the segment embedding to the input data may refer to adding a segment factor to the input data, preferably the embedded input data and / or the input data after having applied the token embedding. The segment factor may specify the type of data associated with the one or more elements. The segment factor may be the same for one or more elements associated with the same type of input data and / or may differ between two or more elements associated with different types of input data.

[0145] Applying the token embedding, the positional embedding, the segment embedding, the column embedding, the row embedding, or a combination thereof may result in embedded input data and / or may be the output of any one of the encoder input 978, 984, 988 or decoder input 984, 994. The data obtained by applying the token embedding, the positional embedding, the segment embedding, the column embedding, the row embedding, or a combination thereof may be processed by the encoder block 974, 986, decoder block 980, 990, encoder output 976, decoder output 992, 982.

[0146] 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.

[0147] 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.

[0148] 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, obtain, retrieve, ingest 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.

[0149] 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.

[0150] 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.

[0151] 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.

[0152] All terms and definitions used herein are understood broadly and have their general meaning if not indicated otherwise.

Claims

CLAIMS1. A method for monitoring and / or controlling output materials produced or to be produced by a distributed chemical production network, the method comprising:- receiving one or more request(s) to provide one or more output materials from one or more output material users, wherein the one or more request(s) include different request types in relation to the data structure of the one or more request(s),- classifying the one or more request(s) based on the one or more request type(s) and routing the one or more request(s) based on the classification to one or more extraction agent(s) configured to extract a natural language representation from the one or more request(s),- generating one or more instruction(s) for generating at least one common, machine-readable representation based on the natural language representation, wherein the at least one common, machine-readable representation is configured to monitor and / or control output materials produced or to be produced by the distributed chemical production network,- generating the common, machine-readable representation by providing the generated one or more instruc- tion(s) to a general-purpose data-driven model, wherein the general-purpose data-driven model is trained on unstructured natural language data and is configured to process natural language;- providing the common, machine-readable representation for monitoring and / or controlling output materials produced or to be produced by the distributed chemical production network.

2. The method of claim 1 , wherein the one or more request(s) include at least one request type per request, wherein multiple requests of different request type are received.

3. The method of any of the preceding claims, wherein the extraction of the natural language representation from the one or more request(s) includes a transformation of the data structure of the request to a text-based data structure.

4. The method of any of the preceding claims, wherein the one or more instruction(s) are generated based on at least one template instruction, wherein the template instruction includes at least one task instruction for the general-purpose data-driven model, at least one context instruction relating to the natural language representation of the request, and / or context data retrieved from at least one context database based on the one or more request(s).

5. The method of any of the preceding claims, additionally including the step of enriching at least one common, machine-readable representation by providing reference data retrieved from at least one reference database based on the at least one common, machine-readable representation.

6. The method of any of the preceding claims, additionally including the step of validating the at least one common, machine-readable representation by one or more validation instruction(s) to the general-purpose data- driven model.

7. The method of claim 6, wherein the step of validating includes providing at least one validation template, wherein the validation template includes task instructions for the general-purpose data-driven model to validate the common representation, wherein at least one validation instruction is generated based on the validation template.

8. The method of claim 6 or 7, wherein the validation template includes at least one placeholder for the common, machine readable representation to be validated, wherein the validation template relates to rules to be checked on validation, wherein the validation template includes correction instructions to be used for error handling on validation, wherein the validation template includes one or more process trigger(s) in case at least one error is found upon validation and cannot be resolved, wherein the validation template includes at least one placeholder for one or more historical example(s) relating to validated common, machine-readable representations.

9. The method of any of the preceding claims, wherein the request relates to at least to one or more output materials), one or more user(s) of the output material, one or more application(s) of the output material, one or more producer(s) of the output material, one or more quantity / ies per output material, one or more time specifications) related to the providing of the output material, one or delivery specification (s) related to the providing of the output material, one or more technical property / les per output material or combinations thereof.

10. The method of any of the preceding claims, wherein the common, machine-readable representation is generated based on the request and at least one pre-defined data structure.

11. The method of any of the preceding claims, wherein the one or more instruction(s) relate to generating a common data structure for the common, machine-readable representation.

12. The method of any of the preceding claims, wherein generating the one or more instruction(s) includes retrieving one or more historical pair(s) including the request and the respective common, machine-readable representation.

13. The method of any of the preceding claims, additionally including the step of additionally including the step of determining a total quantity of output material to be provided based on the at least one common, machine- readable representation, and / or additionally including the step of determining, based on the at least one common, machine-readable representation, at least one facility of the distributed chemical production network forproviding the output material as requested, wherein a trigger signal is generated and provided to the at least one determined facility of the distributed chemical production network.

14. An apparatus for monitoring and / or controlling a chemical production environment including distributed systems for tracking material flows of the chemical production environment, the apparatus comprising:- an input interface configured to receive one or more request(s) to provide one or more output materials from one or more output material users, wherein the one or more request(s) include different request types in relation to the data structure of the one or more request(s),- a classifier configured to classify the one or more request(s) based on the one or more request type(s) and routing the one or more request® based on the classification to one or more extraction agent(s) configured to extract a natural language representation from the one or more request®,- an instruction generator configured to generate one or more instruction(s) for generating at least one common, machine-readable representation based on the natural language representation, wherein the at least one common, machine-readable representation is configured to monitor and / or control output materials produced or to be produced by the distributed chemical production network,- a representation generator configured to generate the common, machine-readable representation by providing the generated one or more instruction(s) to a general-purpose data-driven model, wherein the general-purpose data-driven model is trained on unstructured natural language data and is configured to process natural language;- an output interface configured to provide the common, machine-readable representation for monitoring and / or controlling output materials produced or to be produced by the distributed chemical production network.

15. Use of the common, machine-readable representation generated according to the methods of any of claims 1 to 13 or by the apparatus of claim 14 for monitoring and / or controlling output materials produced or to be produced by a distributed chemical production network.

Images