Product data driven monitoring and / or control of chemical production

Abstract

A computer-implemented method for generating chemical product production and / or processing data relating to production and / or processing of a chemical product, the method comprising: providing chemical product data relating to production and / or processing of the chemical product, providing a plurality of production and / or processing data templates indicative of structures associated with at least a portion of the chemical product data and / or associated element (s), selecting at least one production and / or processing data template based on the at least a portion of the chemical product data and / or associated element (s), generating contextualized chemical product data by combining at least a portion of the chemical product data with the at least one selected production and / or processing data template, providing the contextualized chemical product data to a data-driven model for generating the chemical product production and / or processing data, the data-driven model is parameterized and / or trained to provide the chemical product production and / or processing data in response to being provided the contextualized chemical product data, the chemical product production and / or processing data being provided.

Description

Technical Field

[0001] This disclosure relates to the technical field of controlling and / or monitoring the production and / or processing of chemical products. It also relates to methods and apparatus for generating chemical product production data associated with the production of chemical products, a method for generating chemical product processing data associated with the processing of the produced chemical products, uses of chemical product production and / or processing data, and / or computer elements. Background Technology

[0002] Chemical production is complex and consumes significant resources, such as energy, water, and raw materials. Because chemical production is typically measured in tons, even small relative changes can constitute major improvements. Therefore, improvements in chemical production are desirable. Summary of the Invention

[0003] In another aspect, this disclosure relates to a computer-implemented method for generating chemical product production and / or processing data related to the production and / or processing of a chemical product, the method comprising: providing chemical product data related to the production and / or processing of the chemical product; selecting at least one production and / or processing data template based on at least a portion and / or associated elements of the chemical product data; generating contextualized chemical product data by combining at least a portion of the chemical product data with the at least one selected production and / or processing data template; providing the contextualized chemical product data to a data-driven model for generating the chemical product production and / or processing data, wherein the data-driven model is parameterized and / or trained to provide the chemical product production and / or processing data in response to the provision of the contextualized chemical product data; and providing the chemical product production and / or processing data.

[0004] On the other hand, this disclosure relates to a computer-implemented method for generating chemical product production data related to the production of a chemical product or generating chemical product processing data related to the processing of a chemical product. The method includes: receiving the chemical product production data, or receiving the chemical product processing data; receiving a production data template indicating a structure associated with at least a portion of the chemical product production data, or receiving a processing data template indicating a structure associated with at least a portion of the chemical product processing data; generating contextualized chemical product data by combining at least a portion of the chemical product production data with the production data template, or generating contextualized chemical product data by combining at least a portion of the chemical product production data with the production data template. Processing data templates are combined to generate contextualized chemical product data; chemical product processing data is generated by providing the contextualized chemical product data to a data-driven model, or chemical product production data is generated by providing the contextualized chemical product data to the data-driven model, wherein the data-driven model is parameterized and / or trained based on a training dataset including historical data related to the chemical product production data and chemical product processing data to provide the chemical product production data in response to the provision of the contextualized chemical product data and / or to provide the chemical product processing data in response to the provision of the contextualized chemical product data; providing the chemical product processing data, or providing the chemical product production data.

[0005] On the other hand, this disclosure relates to the use of production and / or processing data related to the production and / or processing of a chemical product, obtained by any of the methods described herein, for the control and / or monitoring of the production and / or processing of that chemical product.

[0006] In another aspect, this disclosure relates to a computer-implemented method for generating chemical product processing data related to the processing of a chemical product, the method comprising: receiving chemical product processing data; receiving a processing data template indicating a structure associated with at least a portion of the chemical product processing data; generating contextualized chemical product data by combining at least a portion of the chemical product production data and the processing data template; generating chemical product production data by providing the contextualized chemical product data to a data-driven model, wherein the data-driven model is parameterized and / or trained based on a training dataset including historical data associated with the chemical product production data and the chemical product processing data to provide the chemical product processing data in response to the provision of the contextualized chemical product data; and providing the chemical product processing data.

[0007] In another aspect, this disclosure relates to a computer-implemented method for generating chemical product production data related to the production of a chemical product, the method comprising: receiving chemical product production data; receiving a production data template indicating a structure associated with at least a portion of the chemical product production data; generating contextualized chemical product data by combining at least a portion of the chemical product production data and the production data template; generating chemical product processing data by providing the contextualized chemical product data to a data-driven model, wherein the data-driven model is parameterized and / or trained based on a training dataset including historical data associated with the chemical product production data and the chemical product processing data to provide the chemical product production data in response to the provision of the contextualized chemical product data; and providing the chemical product production data.

[0008] On the other hand, this disclosure relates to the use of chemical product production data, such as that obtained by any of the methods presented herein, in connection with the production of chemical products.

[0009] On the other hand, this disclosure relates to an apparatus and / or system comprising: a processor; and a memory storing instructions that, when executed by the processor, configure the apparatus to perform steps of any of the methods presented herein.

[0010] On the other hand, this disclosure relates to a non-transitory computer-readable storage medium comprising instructions that, when executed by a computer, cause the computer to perform any of the methods presented herein. Example

[0011] Any disclosures, embodiments, and examples described herein relate to the methods, systems, apparatuses, chemical products, and computer elements listed above and below. Advantageously, the benefits provided by any embodiments and examples also apply to all other embodiments and examples.

[0012] In the following sections, the terminology and / or technical fields used herein and / or the scope of this disclosure will be outlined by way of definition and / or examples. Where examples are given, it should be understood that this disclosure is not limited to those examples.

[0013] Chemical products typically occupy the beginning of numerous supply chains. Performance requirements, quantities, and quality vary significantly depending on the consumers of these products. Therefore, it is desirable to provide customized chemicals for different intended uses. Furthermore, chemical production consumes substantial amounts of energy, water, and raw materials, while also emitting carbon dioxide and toxic waste. Therefore, improving the management of chemical production and processing while reducing carbon dioxide emissions and toxic waste is crucial for conserving resources such as energy, water, and raw materials. This can be achieved by customizing chemical production according to the needs of the chemicals being processed. Moreover, providing chemical processors and / or producers with accurate information about available chemicals helps match expectations with available chemical products. To this end, data-driven models can be deployed to generate chemical production and / or chemical processing data. Using such data-driven models enables rapid and reliable customization of chemical production to meet processing requirements. Because data-driven models are provided with inputs generated based on production and processing data templates, they help users achieve successful operations and save time in generating responses. Moreover, this makes the tools accessible to all users (not just technical experts). Therefore, this invention provides an unhindered possibility for generating chemical product processing data and chemical product production data. This enables more informed decision-making, thereby improving the suitability of measures related to chemical product processing and production while reducing the risk of errors. Ultimately, this contributes to the goal of improving the efficiency of chemical product-related processing and production, thereby reducing the amount of energy, water, and raw materials used, while generating less carbon dioxide and less toxic waste.

[0014] These and other objectives are addressed by the subject matter of the independent claims, and will become apparent upon reading the following description. The dependent claims relate to embodiments of the invention.

[0015] In the embodiments, a chemical product may refer to a product obtained through a chemical production process. A chemical production process may refer to a process comprising one or more chemical reactions. A chemical product may be characterized by at least one functional group. The functional group may be at least one of the following: alkyl, alkenyl, alkynyl, phenyl, carbonyl, ketone, aldehyde, hydroxyl, haloformyl, ester, carboxyl, halogen, carboxyl, peroxy, alkoxycarbonyl, hydroperoxy, ether, acetal, hemiacetal, hemiketal, ketal, carboxylic anhydride, amide, amidine, amino, ketimine, aldehyde imine, imide, cyanate, azo, nitrite, nitrate, nitro, nitrile, thioether, thiol, sulfinyl, sulfonyl, sulfonic acid, thiocyanate, thioester, thioester ester, phosphonic acid, phosphate, or any combination thereof.

[0016] In this embodiment, chemical product data may be related to, and / or indicate, the production and / or processing of chemical products. Chemical product data may include and / or may be chemical product production data and / or chemical product processing data.

[0017] In embodiments, chemical product data may indicate the production and / or processing of a chemical product. Chemical product data may include chemical product manufacturer data, chemical product production instructions, potential chemical product processing data and / or potential chemical product production data, one or more material properties associated with a chemical product, chemical product processor data, chemical product processing instructions, potential chemical product production data, potential chemical product processing data, one or more material properties associated with a product based on a chemical product, one or more input products based on the chemical product used to produce the product based on the chemical product, one or more chemical compounds on which the chemical product is based, or combinations thereof. A product based on a chemical product may refer to a product obtained by processing the chemical product.

[0018] Chemical product manufacturer data can indicate manufacturers associated with a chemical product and / or production facilities associated with the production of the chemical product. For example, chemical product manufacturer data can indicate one or more manufacturers associated with a chemical product, quantitative production data, qualitative production data, or a combination thereof. Qualitative production data can indicate the chemical structure associated with the chemical product to be produced. For example, qualitative production data can indicate the name of the chemical product, the mass of the chemical product, one or more chemical compounds associated with the chemical product, one or more material properties associated with the chemical product, one or more chemical compounds on which the chemical product is based, or a combination thereof. Quantitative production data can indicate the quantity of the chemical product.

[0019] Potential chemical product production data can indicate the anticipated production of a chemical product based on historical production and / or production associated with substitutes for the chemical product. For example, potential chemical product production data can indicate historical production associated with the chemical product, substitutes for the chemical product, the product to be produced based on the chemical product, the application areas associated with the chemical product, or a combination thereof. In embodiments, chemical product production data can be received in audio and / or text formats.

[0020] Chemical product processor data can indicate chemical product processors associated with a chemical product and / or processing facilities associated with the processing of the chemical product. For example, chemical product processor data can indicate one or more chemical product processors associated with a chemical product, quantitative processing data, qualitative processing data, application areas associated with the chemical product, the product to be produced based on the chemical product, the application areas of the product to be produced based on the chemical product, or a combination thereof. Qualitative processing data may include the name of the chemical product, the quality of the chemical product, one or more target material properties associated with the chemical product, one or more chemical compounds included in the chemical product, or a combination thereof. Quantitative processing data can indicate quantities associated with one or more components associated with the chemical product. Chemical product processing instructions can indicate processing instructions associated with the processing of the chemical product. For example, chemical product processing instructions may be related to processing conditions such as temperature, pressure, reaction rate, and catalyst.

[0021] Potential chemical product processing data can indicate the anticipated processing of a chemical product based on historical processing and / or processing associated with its substitutes. For example, potential chemical product processing data can indicate historical processing associated with a chemical product, its substitutes, or a combination thereof.

[0022] In embodiments, chemical product data can be received and / or provided in audio and / or numerical and / or text and / or digital formats. In embodiments, the chemical product data may include one or more elements, preferably production and / or processing parameter values. One or more elements (preferably production and / or processing parameter values) may be adapted to be combined with a production and / or processing data template. Preferably, one or more elements can be inserted into the production and / or processing data template, preferably at the location in the production and / or processing data template that specifies the production and / or processing parameters. Production and / or processing parameters may be placeholders indicating where at least a portion of the chemical product data (preferably production and / or processing parameter values) is inserted.

[0023] In this embodiment, chemical product processing data can indicate processing related to the chemical product. Chemical product processing data may include chemical product processor data, chemical product processing instructions, potential chemical product production data, and / or potential chemical product processing data.

[0024] Chemical product processor data can indicate chemical product processors associated with a chemical product and / or processing facilities associated with the processing of the chemical product. For example, chemical product processor data can indicate one or more chemical product processors associated with a chemical product, quantitative processing data, qualitative processing data, application areas associated with the chemical product, the product to be produced based on the chemical product, the application areas of the product to be produced based on the chemical product, or a combination thereof. Qualitative processing data may include the name of the chemical product, the quality of the chemical product, one or more target material properties associated with the chemical product, one or more chemical compounds included in the chemical product, or a combination thereof. Quantitative processing data can indicate quantities associated with one or more components associated with the chemical product. Chemical product processing instructions can indicate processing instructions associated with the processing of the chemical product. For example, chemical product processing instructions may be associated with processing conditions such as temperature and pressure.

[0025] Potential chemical product processing data can indicate the anticipated processing of a chemical product based on historical processing and / or processing associated with its substitutes. For example, potential chemical product processing data can indicate historical processing associated with a chemical product, its substitutes, or a combination thereof.

[0026] In embodiments, chemical product processing data can be received in audio and / or text formats. In embodiments, the chemical product processing data may include one or more elements, preferably processing parameter values. One or more elements (preferably processing parameter values) may be suitable for combination with a processing data template. Preferably, one or more elements may be inserted into the processing data template at the location specified by the processing data template for the processing parameters. The processing parameters may be placeholders indicating where at least a portion (preferably processing parameter values) of the chemical product processing data should be inserted.

[0027] In embodiments, chemical product production data can indicate production associated with a chemical product. Chemical product production data may include chemical product manufacturer data, chemical product production instructions, potential chemical product processing data, and / or potential chemical product production data. Chemical product manufacturer data can indicate manufacturers associated with a chemical product and / or production facilities associated with the production of the chemical product. For example, chemical product manufacturer data can indicate one or more manufacturers associated with a chemical product, quantitative production data, qualitative production data, or a combination thereof. Qualitative production data can indicate the chemical structure associated with the chemical product to be produced. For example, qualitative production data can indicate the name of the chemical product, the quality of the chemical product, one or more chemical compounds associated with the chemical product, one or more material properties associated with the chemical product, one or more chemical compounds on which the chemical product is based, or a combination thereof. Quantitative production data can indicate the quantity of the chemical product. The name of the chemical product can refer to a name that can be used to sell the chemical product and / or specify the name of one or more chemical compounds associated with the chemical product.

[0028] Potential chemical product production data can indicate the anticipated production of a chemical product based on historical production and / or production associated with substitutes for the chemical product. For example, potential chemical product production data can indicate historical production associated with the chemical product, substitutes for the chemical product, the product to be produced based on the chemical product, the application areas associated with the chemical product, or a combination thereof. In embodiments, chemical product production data can be received in audio and / or text formats.

[0029] In an embodiment, chemical product production data may include one or more elements, preferably production parameter values. One or more elements (preferably production parameter values) may be suitable for combination with a production data template. Preferably, one or more elements may be inserted into the production data template at the location specified by the production parameter in the production data template. Production parameters may be placeholders indicating where at least a portion (preferably production parameter values) of the chemical product production data should be inserted.

[0030] In this embodiment, contextualized chemical product data may refer to a combination of input data and at least one production and / or processing data template. Contextualized chemical product data may include contextualized chemical product data and / or contextualized chemical product data. Contextualized chemical product data may be a second set of contextualized chemical product data.

[0031] In an embodiment, the data-driven model may include one or more machine learning architectures and model parameters. The one or more machine learning architectures may be at least one of the following: one or more convolutional layers, one or more fully connected layers, one or more pooling layers, one or more transformer encoders, one or more classification layers, one or more transformer decoders, one or more feedforward layers, one or more linear layers, one or more transformer encoder-decoders, or combinations thereof.

[0032] Data-driven models can be trained and / or parameterized based on training datasets. Training datasets may include historical data related to chemical product data and chemical product production and / or processing data. Further, training datasets may include at least one dataset containing chemical product data and corresponding chemical product production and / or processing data. Training data may include one or more element sequences. Preferably, at least one dataset in the training dataset may include element sequences specifying chemical product data and corresponding chemical product production and / or processing data. Data-driven models can be trained and / or parameterized based on training datasets to provide chemical product production and / or processing data in response to the provision of contextualized chemical product data. Data-driven models can be generative models. Generative models can be trained and / or parameterized based on training datasets to preferably generate chemical product production and / or processing data based on provided contextualized chemical product data. Preferably, training datasets may include numerical historical data and / or sentence-based historical data related to chemical product production and / or processing data and chemical product data. Numerical historical data may refer to historical data including one or more numbers. Sentence-based historical data may refer to historical data including at least a portion of a sentence. Data-driven models can be trained and / or parameterized to sequentially generate chemical product production and / or processing data, particularly based on contextualized chemical product data.

[0033] Data-driven models can be trained and / or parameterized to sequentially generate chemical product production and / or processing data, particularly based on provided contextualized chemical product data. This can mean that the data-driven model can be trained and / or parameterized to generate chemical product production and / or processing data comprising a sequence of elements by generating one element (preferably the first element of the sequence) based on the contextualized chemical product data, and generating subsequent elements (preferably elements after the first element in the sequence) based on the contextualized chemical product data and previously generated elements in the sequence. By taking previously generated elements into account, the data-driven model can generate chemical product production and / or processing data more accurately, and this data is better linked to other elements within the sequence. Ultimately, this enables customized production for the quantity and / or quality associated with the chemical product, improves the efficiency of processes associated with processing and producing the chemical product, and helps to use resources as efficiently as possible.

[0034] In an embodiment, the data-driven model may be a pre-trained data-driven model. The pre-trained data-driven model may be trained on general data to provide output data based on provided input data. In another embodiment, the data-driven model may be a fine-tuned data-driven model. The fine-tuned data-driven model may preferably be trained on general data during a first training process to provide output data based on provided input data. The fine-tuned data-driven model may preferably be further trained in a second training process on historical data and chemical product data to provide chemical product production and / or processing data based on provided contextualized chemical product data. General data may include historical input data and / or historical output dates. Input data and / or output data may refer to data that can be processed by the data-driven model (preferably a parameterized and / or initialized data-driven model). Input data and / or output data may include and / or may represent one or more element sequences, wherein the element sequence comprises two or more elements. In particular, the element sequence may indicate a sequence of two or more elements in a sequence.

[0035] Training the data-driven model can refer to and / or the training process can be the process of building a data-driven model, specifically the process of determining and / or updating the parameters of the data-driven model. During the training process, the data-driven model can be adjusted to achieve a best fit with the training dataset, for example, by best-fitting at least one input value with at least one target output value. For example, if the neural network is a feedforward neural network (e.g., a convolutional neural network (CNN)), the backpropagation algorithm can be applied to train the neural network. In the case of a recurrent neural network (RNN), the gradient descent algorithm can be used to achieve the training objective. The gradient descent algorithm uses gradients to update the parameters. The gradient can indicate the degree of change of the parameters of the data-driven model. The gradient can be obtained through backpropagation. Therefore, the gradient descent algorithm can be based on backpropagation. The training process can terminate when the deviation of the output generated by the data-driven model from the target output specified by the training dataset falls within a predetermined range. When the training process can terminate, the determination and / or updating of the parameters of the data-driven model can be terminated. The output generated by the data-driven model can be chemical product production and / or processing data, historical data related to chemical product production and / or processing data, or a combination thereof. The target output specified by the training dataset can be chemical product production and / or processing data, historical data related to chemical product production and / or processing data, or a combination thereof. During and / or while training the data-driven model, the training dataset can include one or more element sequences, and these sequences can be provided sequentially to the data-driven model, and / or the model can generate outputs sequentially. Therefore, the data-driven model can generate a first element from one or more element sequences based on contextualized chemical product data, and can generate additional elements, such as a second element from one or more sequences, based on one or more previously generated elements and contextualized chemical product data. A second training process following the first training process can be advantageous for customizing the data-driven model for use cases related to the production and processing of chemical products. This helps to further improve the accuracy of the data-driven model.

[0036] In an embodiment, a transformer encoder may include an encoder input, one or more encoder blocks, and / or an encoder output. The encoder input may generate embedded contextualized chemical product data based on received contextualized chemical product data. One or more encoder blocks may generate a production and / or processing data context tensor based on the received embedded contextualized chemical product data. The encoder output may generate chemical product production and / or processing data based on the received production and / or processing data context tensor. A transformer decoder may include a decoder input, one or more decoder blocks, and / or a decoder output. The decoder input may generate embedded contextualized chemical product data based on received contextualized chemical product data. One or more decoder blocks may generate a production and / or processing data context tensor based on the received embedded contextualized chemical product data. The decoder output may generate chemical product production and / or processing data based on the received production and / or processing data context tensor. A transformer encoder-decoder may include a transformer encoder comprising an encoder input and one or more encoder blocks, and a transformer decoder comprising a decoder input, one or more decoder blocks, and a decoder output.

[0037] In an embodiment, generating chemical product production and / or processing data by providing contextualized chemical product data to a data-driven model may include: receiving contextualized chemical product data at encoder inputs and / or decoder inputs; applying input embeddings to the contextualized chemical product data, for example, to generate embedded contextualized chemical product data; applying position encoding to the contextualized chemical product data and / or the embedded contextualized chemical product data; applying self-attention to the contextualized chemical product data and / or the embedded contextualized chemical product data, for example, to generate a production and / or processing data context tensor; and applying a softmax function to the contextualized chemical product data.

[0038] In embodiments, the method may further include, preferably, embedding contextual chemical product data by a data-driven model in response to the provision of contextual chemical product data. Embedding contextual chemical product data may refer to applying input embedding and / or positional encoding to the contextual chemical product data. Embedding contextual chemical product data may include transforming the contextual chemical product data into machine-processable contextual chemical product data, such as tensors. The data-driven model can process the machine-processable contextual chemical product data into chemical product production data. By doing so, contextual chemical product data can be received in user language, and the data-driven model can handle accurate, robust, and efficient processing of user input. This enables barrier-free use of the data-driven model to generate chemical product production data. Furthermore, providing access to all users provides all users with the opportunity to deploy the methods and systems presented herein. This will increase the number of requests processed by these methods and systems, resulting in more customized chemical product processing and production, while reducing the amount of error in chemical product processing and production. Therefore, resources used for chemical product processing and production are used more efficiently and with fewer errors.

[0039] Additionally or alternatively, embedding contextualized chemical product data may include passing the contextualized chemical product data through an embedding layer. The embedding layer may be adapted to transform the contextualized chemical product data into a machine-processable format. A machine-processable format may refer to a numeric (particularly tensor-based) representation of the contextualized chemical product data. Embedding contextualized chemical product data can produce embedded contextualized chemical product data. Embedded contextualized chemical product data may include tensors representing the contextualized chemical product data. A tensor-based representation of the contextualized chemical product data may be referred to as embedded contextualized chemical product data. Embedded contextualized chemical product data may be referred to as a machine-processable format of the contextualized chemical product data.

[0040] In embodiments, the method may further include processing embedded contextualized chemical product data into a production and / or processing data context tensor using a data-driven model. Processing embedded contextualized chemical product data into a production and / or processing data context tensor can refer to transforming the embedded contextualized chemical product data into a production and / or processing data context tensor. Preferably, the embedded contextualized chemical product data can be transformed into a production and / or processing data context tensor by forming one or more tensor products associated with the embedded contextualized chemical product data and / or one or more representations associated with the embedded contextualized chemical product data. The one or more representations associated with the embedded contextualized chemical product data can be obtained by applying one or more mathematical operations to the embedded contextualized chemical product data. The one or more mathematical operations may include, for example, summation, subtraction, division, integration, differentiation, multiplication, normalization, or combinations thereof.

[0041] In embodiments, embedded contextualized chemical product data can be sequence-specific. Sequence-specific embedded contextualized chemical product data can refer to a representation of embedded contextualized chemical product data that takes into account a sequence of data points associated with the embedded contextualized chemical product data. Sequence-specific embedded contextualized chemical product data can be obtained by applying input embedding and / or positional encoding to the contextualized chemical product data. For example, applying positional encoding can refer to adding and / or multiplying one or more portions of the embedded contextualized chemical product data with positional factors indicating the position of one or more portions of the embedded contextualized chemical product data within the embedded contextualized chemical product data. Sequence-specific embedded contextualized chemical product data can be processed (especially transformed) in a manner similar to embedded contextualized chemical product data.

[0042] In this embodiment, the received chemical product production data can be sentence-based. Therefore, production and / or processing data templates indicating structures associated with at least a portion of the chemical product data and / or contextualized chemical product data can be sentence-based. Thus, the data-driven model can be a natural language processing model. The natural language processing model can be a model suitable for processing sentence-based inputs (e.g., sentence-based chemical product data and / or chemical product data including one or more elements). Contextualized chemical product data based on sentences can be embedded to produce embedded contextualized chemical product data. Sentence-based contextualized chemical product data can be embedded via input embedding (preferably word embedding). Word embedding can refer to transforming sentence-based contextualized chemical product data into a machine-processable format.

[0043] In embodiments, the determined chemical product production and / or processing data may include two or more portions, specifically a first portion and a second portion of the chemical product production and / or processing data. Where the chemical product production and / or processing data is sentence-based, the two or more portions may refer to words, numbers, and / or parts of words. The first portion of the chemical product production and / or processing data may refer to a first part of a sequence of chemical product production and / or processing data, and the second portion may refer to a second part of a sequence of chemical product production and / or processing data. The first portion may be determined at a first time step, and the second portion may be determined at a second time step.

[0044] In the first time step, production and / or processing data for a first chemical product can be determined based on embedded contextualized chemical product data. In the second time step, production and / or processing data for a second chemical product can be determined based on the embedded contextualized chemical product data and the embedded first chemical product production and / or processing data. The embedded first chemical product production and / or processing data can refer to first chemical product production and / or processing data embedded similarly to the contextualized chemical product data. Therefore, production and / or processing data for chemical products can be generated based on the contextualized chemical product data and at least in part based on at least a portion of the generated production and / or processing data, preferably by further providing at least a portion of the generated production and / or processing data to a data-driven model, particularly by providing at least a first portion of the generated production and / or processing data to the data-driven model.

[0045] In the embodiments, material properties can refer to physical properties, chemical properties, biological properties, or a combination thereof. Chemical properties can be properties that can only be established by altering one or more chemical structures associated with at least one chemical product. Examples of chemical properties can be acidity, oxidation state, or reactivity. Physical properties can be one of the following: mechanical properties, electrical properties, optical properties, thermal properties, etc. For example, physical properties can include one or more of the following: density, scratch resistance, electrical conductivity, color, absorption, heat capacity, etc. Biological properties can refer to toxicity, bioactivity, biodegradability, bioaccumulation, etc.

[0046] In an embodiment, the processing data template indicates a structure associated with at least a portion of the chemical product processing data. The processing data template may indicate a sequence of one or more elements associated with at least a portion of the chemical product processing data. Preferably, the processing data template may indicate one or more elements associated with at least a portion of the chemical product processing data, as well as a sequence of one or more elements included in the processing data template.

[0047] A machining data template can indicate one or more machining parameters and one or more elements associated with the relationship between the machining parameters (specifically, the machining parameter values). Machining parameters can specify where machining parameter values ​​can be inserted. One or more machining parameters and one or more elements associated with the relationship between the machining parameters (specifically, the machining parameter values) can constitute a sequence of two or more elements.

[0048] Therefore, an element associated with a relationship between one or more processing parameters can specify a portion of the element sequence associated with contextualized chemical product data. Additionally or alternatively, an element associated with a relationship between one or more processing parameters can specify one or more portions of the sequence adjacent to (preferably before, after, and / or between) the position where a processing parameter value can be inserted.

[0049] In this embodiment, the processing data template may include at least a portion of chemical product processing data. The processing data template may include one or more of the following: chemical product processor data, chemical product processing instructions, potential chemical product processing data, or combinations thereof.

[0050] In an embodiment, the production and / or processing data template may refer to a production data template and / or a processing data template. The production and / or processing data template may indicate a structure associated with at least a portion of the input data, chemical product production data, and / or chemical product processing data.

[0051] In embodiments, a production data template indicates a structure associated with at least a portion of chemical product production data. The production data template may indicate a sequence of one or more elements associated with at least a portion of the chemical product production data. Preferably, the production data template may indicate one or more elements associated with at least a portion of the chemical product production data, as well as a sequence of one or more elements included in the production data template. The production data template may indicate one or more production parameters and one or more elements associated with the relationship between one or more production parameters (particularly one or more production parameter values). Production parameters may specify where production parameter values ​​can be inserted. One or more production parameters and one or more elements associated with the relationship between one or more production parameters (particularly one or more production parameter values) may constitute a sequence of two or more elements. Thus, the element associated with the relationship between one or more production parameters may specify a portion of the element sequence associated with the contextualized chemical product data. Additionally or alternatively, the element associated with the relationship between one or more production parameters may specify one or more portions of the sequence adjacent to (preferably before, after, and / or between) the position where a production parameter value can be inserted.

[0052] In this embodiment, the production data template may include at least a portion of chemical product production data. The production data template may include one or more of the following: chemical product manufacturer data, chemical product production instructions, and / or potential chemical product production data, or combinations thereof.

[0053] In an embodiment, chemical product data may include one or more elements associated with chemical product data, and wherein production and / or multiple production and / or processing data templates may include one or more elements.

[0054] In embodiments, an element may refer to at least a portion of a word, at least a portion of a number, a portion of a table, or a combination thereof. Additionally or alternatively, a sequence of two or more elements may refer to at least a portion of a sentence, at least a portion of a number sequence, at least a portion of a table, or a combination thereof. A portion of a table may refer to row numbers, column numbers, and / or table entries. The contextualized chemical product data may include two or more elements that are equal to one or more elements associated with the chemical product data and / or one or more elements of at least one production and / or processing data template. Further, the contextualized chemical product data may include a sequence of two or more elements that are equal to one or more elements associated with the chemical product data and / or one or more elements of at least one production and / or processing data template. Additionally or alternatively, the chemical product data may include one or more elements associated with the chemical product data. Further, at least one of a plurality of production and / or processing data templates may include one or more elements.

[0055] In an embodiment, the method may further include: receiving second chemical product production data (particularly based on chemical product processing data), or receiving second chemical product processing data (particularly based on chemical product production data); and optionally receiving a second production data template, or receiving a second processing data template; and generating second contextualized chemical product data based on the second chemical product production data and the production data template (optionally the second production data template), or generating the second contextualized chemical product data based on the second chemical product processing data and the processing data template (optionally the second processing data template); and generating second chemical product processing data based on the second contextualized chemical product data, or generating second chemical product production data based on the second contextualized chemical product data; and providing the second chemical product processing data or the second chemical product production data. Additionally or alternatively, second contextualized chemical product data may be received based on at least a portion of the chemical product processing data, or second contextualized chemical product data may be received based on at least a portion of the chemical product production data; and second chemical product processing data may be generated based on the second contextualized chemical product data, or second chemical product production data may be generated based on the second contextualized chemical product data; and the second chemical product processing data or the second chemical product production data may be provided. Second chemical product production data, second chemical product processing data, second contextualized chemical product data, second contextualized chemical product data, or combinations thereof may be received via a user interface. This allows for customized solutions tailored to the user's specific circumstances. Furthermore, second chemical product production data, for example, from the user, may be obtained to further improve the generated data based on previously generated data or to respond to changes in conditions involving the production and / or processing of the chemical product. This is resource-efficient and allows coverage of use cases with constantly changing conditions.

[0056] In embodiments, the method may further include providing an instruction for selecting a data-driven model from a plurality of data-driven models, and selecting a data-driven model from the plurality of data-driven models based on the instruction. Preferably, the instruction for selecting a data-driven model can be provided via a user interface, preferably from a user. Two or more data-driven models may be available. The two or more data-driven models can be distinguished by their intended use. For example, a first data-driven model among the two or more data-driven models may be more accurate than a second data-driven model among the two or more data-driven models. On the other hand, a second data-driven model may generate chemical product processing data and / or chemical product production data faster than a first data-driven model. The corresponding first data-driven model can be selected when the accuracy of the generated data is more valuable than the speed at which the data is generated. A second data-driven model can be selected when the speed at which the data is generated is more valuable than the high accuracy of the generated data. Additionally or alternatively, one or more of the two or more data-driven models may be more resource-efficient than the other two or more data-driven models. Therefore, the most suitable model can be selected for the corresponding use case. This makes it possible to customize chemical product processing data and chemical product production data while minimizing resources. This instruction can be provided by the user. Additionally or alternatively, the instruction can be based on one or more elements associated with the chemical product data, and / or the instruction can be provided by providing the chemical product data. Therefore, the instruction may include one or more elements associated with the chemical product data. A data-driven model can be selected based on the chemical product data, preferably based on one or more elements associated with the chemical product data.

[0057] In embodiments, multiple production and / or processing data templates may include various types of production and / or processing data templates. Selecting at least one production and / or processing data template based on at least a portion and / or associated elements of chemical product data may refer to selecting at least one production and / or processing data template based on its type. Further, the type of production and / or processing data template may be selected based on determining that its type corresponds to chemical product data (particularly one or more elements associated with the chemical product data). Chemical product data may include one or more elements associated with the chemical product data. The type of production and / or processing data template may be associated with one or more elements associated with the chemical product data. Therefore, selecting at least one production and / or processing data template may refer to selecting at least one production and / or processing data template whose type corresponds to one or more elements associated with the chemical product data. The type of this production and / or processing data template specifies that the chemical production and / or processing data includes at least one of the following: chemical product manufacturer data, chemical product production instructions, chemical product processor data, chemical product processing instructions, one or more material properties associated with the chemical product, and one or more material properties associated with products based on the chemical product. By doing so, the interaction with the data-driven model can be customized for the target use case, resulting in more accurate chemical product production and / or processing data. This enables greater resource efficiency in the production and processing of chemical products.

[0058] In embodiments, the method may further include selecting a data-driven model from multiple data-driven models based on chemical product data. Preferably, the data-driven model can be selected from multiple data-driven models based on at least one production and / or processing data template, and particularly the type of at least one production and / or processing data template. By doing so, the interaction with the data-driven model can be customized for the target use case, thereby generating more accurate chemical product production and / or processing data. This enables higher resource efficiency in the production and processing of chemical products.

[0059] In an embodiment, the plurality of production and / or processing data templates may be provided via a database, and wherein selecting at least one production and / or processing data template based on at least a portion and / or associated elements of the chemical product data may include generating embedded chemical product data and generating one or more embedded production and / or processing data templates, and selecting the production and / or processing data template based on determining that the tensor product between the tensor associated with the embedded chemical product data and the tensor associated with at least one of the one or more production and / or processing data templates can be within a predefined numerical range.

[0060] Generating embedded chemical product data can refer to providing production and / or processing data to an encoder and receiving embedded chemical product data from the encoder. Generating one or more embedded production and / or processing data templates can refer to providing production and / or processing data templates to an encoder and receiving one or more embedded production and / or processing data templates from the encoder. The encoder can be configured to transform the chemical product data and production and / or processing data templates into a machine-readable format, such as tensors. Therefore, the encoder can reduce the dimensionality of the chemical product data and production and / or processing data templates. An example of such an encoder could be as shown in... Figure 8 The embedding layer of the Continuous Bag-of-Words (CBOW) model is described in the context of this example. Various encoders are available for embedding data, such as word2vec, GloVe, and FastText. The tensor product between embedded chemical product data and embedded production and / or processing data templates can be used to determine the similarity between the embedded chemical product data and one or more embedded production and / or processing data templates. Selecting at least one production and / or processing data template based on the tensor product can be used to select at least one production and / or processing data template that is most similar to the chemical product data. Searching for processing production and / or processing data templates based on embedded chemical product data allows for searches beyond the scope of classification, but enables content comparison. In the example, chemical compounds are often associated with multiple names, such as the UIPAC nomenclature and historically used colloquial names. Therefore, when ethylene and ethene are mentioned, they refer to the same chemical compound, just with different names. Searching via embedded data allows linking these two different words to the same chemical compound. Therefore, the quality of the search is significantly improved by using an embedded database.

[0061] In embodiments, chemical product production data may include one or more elements associated with chemical product production data, and chemical product processing data may include one or more elements associated with chemical product processing data. These one or more elements may be part of an element sequence (particularly a sequence specified by contextualized chemical product data and / or contextualized chemical product data).

[0062] In embodiments, contextualized chemical product data may include sequences of two or more elements associated with contextualized chemical product data, and / or wherein contextualized chemical product data may include sequences of two or more elements associated with contextualized chemical product data.

[0063] In this embodiment, an element may refer to a word and / or a portion of a word. In this embodiment, a sequence of two or more elements may be a sentence and / or a portion of a sentence.

[0064] By doing so, contextualized chemical product data, chemical product production data, chemical product processing data, and contextualized chemical product data can be received in the user's language. The data-driven model can handle the user and the accurate, robust, and efficient processing of user input. This enables the seamless use of the data-driven model to generate chemical product production and processing data. Furthermore, providing access to all users provides them with the opportunity to deploy the methods and systems presented herein. This will increase the number of requests processed by these methods and systems, resulting in more customized chemical product processing and production, while reducing the amount of error in chemical product processing and production. Consequently, resources used for chemical product processing and production are used more efficiently and with fewer errors.

[0065] In an embodiment, the system may further include one or more databases configured to provide processing data templates and / or production data templates. Storing the processing data templates and production data templates in the databases enables robust retrieval of the processing data templates and production data templates, and makes them readily available when needed.

[0066] In an embodiment, the system may further include a user interface configured to receive chemical product processing data and / or chemical product production data, particularly from a user, and / or configured to provide chemical product processing data and / or chemical product production data, particularly to the processor. The user interface may allow the user to interact with a data-driven model. Thus, customized chemical product production data and chemical product processing data are generated under the user's control.

[0067] In embodiments, a processor can refer to any logical circuit system configured to perform basic operations of a computer or system, and / or generally refers to a device configured to perform computations or logical operations. Specifically, a processor or computer processor can be configured to process the basic instructions that drive a computer or system. A processor can be a semiconductor-based processor, a quantum processor, or any other type of processor configured to process instructions. As an example, a processor can be or can include a Central Processing Unit (“CPU”). A processor can be a Graphics Processing Unit (“GPU”), a Tensor Processing Unit (“TPU”), a Complex Instruction Set Computing Microprocessor (“CISC”), a Reduced Instruction Set Computing (“RISC”) microprocessor, a Very Long Instruction Word (“VLIW”) microprocessor, or a processor implementing other instruction sets or multiple processors implementing combinations of instruction sets. A processing device can 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, etc. The methods, systems, and devices described herein can be implemented as software in a DSP, microcontroller, or any other auxiliary processor, or as hardware circuitry within an ASIC, CPLD, or FPGA. It should be understood that the term "processor" can also refer to one or more processing devices, such as a distributed processing device system located across multiple computer systems (e.g., cloud computing), and is not limited to a single device unless otherwise stated. A processor can also be an interface to a remote computer system, such as a cloud service. A processor can include or be a Secure Isolation Zone (SEP). An SEP can be secure circuitry configured to process the spectrum. "Secure circuitry" is circuitry that protects isolated internal resources from direct access by external circuitry. A processor can be an image signal processor (ISP) and can include circuitry systems suitable for processing images, particularly images containing personal and / or confidential information.

[0068] In embodiments, the memory can be physical system memory, which can be volatile, non-volatile, or a combination thereof. The memory can include non-volatile mass storage devices, such as physical storage media. The memory can be a computer-readable storage medium (e.g., RAM, ROM, EEPROM, CD-ROM) or other optical disc storage, disk storage, other magnetic storage devices, non-disk storage (e.g., solid-state drives), or any other physical tangible storage medium that can be used to store object program code means in the form of computer-executable instructions or data structures and is accessible by the computing system. Furthermore, the memory can be a computer-readable medium carrying computer-executable instructions (also referred to as a transmission medium). Further, upon arrival at various computing system components, program code means in the form of computer-executable instructions or data structures can be automatically transferred from the transmission medium to the storage medium (and vice versa). For example, computer-executable instructions or data structures received via a network or data link can be buffered in RAM within a network interface module (e.g., a "NIC") and then ultimately transferred to the computing system RAM and / or a less volatile storage medium at the computing system. Therefore, it should be understood that storage media may be included in computing components that also (or even primarily) utilize transmission media.

[0069] In this embodiment, the plurality of production and / or processing data templates may be provided via a database, and selecting at least one production and / or processing data template based on at least a portion and / or associated elements of the chemical product data includes providing a query to the database based on the production and / or processing data and / or receiving the at least one production and / or processing data template based on the query. The query can define categories associated with the search for the production and / or processing data template. This allows for predictable, robust, and human-relevant searches. Furthermore, it allows users to define and refine their searches as needed.

[0070] In one embodiment, providing the chemical product data includes retrieving the chemical product data from the data source by providing a request to the data source for receiving the chemical product data and receiving the chemical product data from the data source in response to providing the request. The data source can be configured to provide the requested data upon receiving a request to receive the data. By doing so, real-time data can be retrieved. Thus, up-to-date data related to the production and / or processing of chemical products can be retrieved. This improves the reliability of chemical production and / or processing data generated based on the chemical product data. Ultimately, this improves the production and / or processing of chemical products.

[0071] In embodiments, any of these methods may further include retrieving additional chemical product data from a data source by providing a request to the data source for receiving additional chemical product data and receiving the chemical product data from the data source in response to providing the request. The data source may be configured to provide the requested data upon receiving a request to receive the data. The additional chemical product data may be associated with the chemical product data. For example, the chemical product data may include a first portion of associated data points, and / or the additional chemical product data may include a second portion of the chemical product data, particularly multiple data points corresponding to the chemical product data. Contextualized chemical product data may be generated by combining at least a portion of the chemical product data, at least a portion of the additional chemical product data, and at least one selected production and / or processing data template.

[0072] In embodiments, numerical identifiers associated with one or more data points linked to chemical product data and / or other chemical product data may be provided, for example, via a user interface and / or along with contextualized chemical production and / or processing data. The numerical identifiers may point to one or more data points associated with the chemical product data and / or other chemical product data. For example, the numerical identifiers may include links to one or more datasets associated with a data source, and / or may point to one or more datasets associated with a data source. Chemical product data and / or other chemical product data can be accessed based on the numerical identifiers. By doing so, raw data stored in a database or website can be accessed to examine chemical production and / or processing data generated by the data-driven model. This allows for the evaluation of the quality of responses from the data-driven model. Therefore, the user has complete control over where the data is obtained and how the data-driven model processes the data. Thus, the user can interact with the data-driven model when errors are detected. This improves the accuracy of the chemical production and / or processing data generated by the data-driven model. Ultimately, this improves the production and / or processing of chemical products.

[0073] In this embodiment, contextualized chemical product data may be suitable for triggering data-driven models to generate chemical product production and / or processing data.

[0074] In the embodiments, chemical product production and / or processing data can be provided for monitoring the production of chemical products and / or for the production of chemical products. Specifically, the method for generating chemical product production and / or processing data related to the production and / or processing of chemical products can be a method for producing and / or processing chemical products. Attached Figure Description

[0075] The disclosure will be further described below with reference to the accompanying drawings. In the drawings and the disclosure, the same reference numerals are intended to refer to the same or similar elements, components and / or parts.

[0076] Figure 1 It demonstrates the production and processing of chemical products.

[0077] Figure 2 An example of a chemical production system is shown.

[0078] Figure 3 A system for generating chemical product production and / or processing data related to the production and / or processing of chemical products is demonstrated.

[0079] Figure 4A An example for selecting at least one production and / or processing data template is shown.

[0080] Figure 4B Another embodiment for selecting at least one production and / or processing data template is shown.

[0081] Figure 5 Examples of methods for generating chemical product production and / or processing data related to the production and / or processing of chemical products are shown.

[0082] Figure 6 Another embodiment of a method for generating chemical product production and / or processing data related to the production and / or processing of chemical products is shown.

[0083] Figure 7A Examples of contextualized chemical product data, as well as production and / or processing data, are presented.

[0084] Figure 7B Another embodiment of contextualized chemical product data, as well as production and / or processing data, is presented.

[0085] Figure 8 An example of training the embedding layer is shown.

[0086] Figure 9A An example of a transformer encoder architecture is shown.

[0087] Figure 9B An example of a transformer decoder architecture is shown.

[0088] Figure 9C An example of a transformer encoder-decoder architecture is shown.

[0089] Figure 10Examples of training and / or deploying transformer encoders, transformer decoders, and / or transformer encoder-decoders are shown.

[0090] Figure 11 Examples of input embedding for embedding numbers and / or tables are shown.

[0091] Figure 12 An example of the Mamba architecture is shown. Detailed Implementation

[0092] The following examples are merely instances for implementing the methods, systems, or application devices disclosed herein and should not be considered limiting.

[0093] Figure 1 The production and processing of chemical products are illustrated. Chemical products can be produced by chemical product manufacturer 104. Typically, generating a chemical product from raw materials supplied by input material supplier 102 or from recycled materials based on chemical products supplied by recycling system operator 114 may require several steps and / or synthesis. Therefore, chemical products may be produced within a chemical production network comprising multiple plants. The chemical products can then be supplied to chemical product user 106. Chemical product user 106 can process the chemical products into intermediate products, which are chemical product-based products. Intermediate products can be further supplied and / or processed until final product manufacturer 108 can process one or more intermediate products into a final product. This final product may be based on the chemical product. The final product can be provided to final product user 110. Once the final product becomes waste, it can be provided to waste collectors and / or sorters 112. Waste can be separated based on waste fractionation for recycling. These fractions can be provided to recycling system operator 114. However, waste and its fractions can be based on chemical products. Therefore, the processing of chemical product 118 may include all actions performed on chemical products, chemical product-based products, and chemical product-based waste and waste fractions.

[0094] Figure 2 It demonstrates a distributed production environment, such as one or more chemical plants.

[0095] The distributed production environment may include devices and sensors 204 that generate one or more sensor-related data streams. The distributed production environment may produce one or more products as defined above, wherein characteristics of the one or more products can be measured, extracted, or calculated, thereby generating one or more product-related data streams. Plant data 228 (plant-based data) may include data obtained from each of the one or more data streams.

[0096] The equipment can be any equipment in a distributed production environment, such as pumps, heat exchangers, valves, reaction vessels, separation chambers and / or the like.

[0097] The sensor can be any kind of sensor in a distributed production environment, such as temperature sensors, flow sensors, pressure sensors and / or the like.

[0098] One or more products produced in a distributed production environment can be any type of product as described above. The properties of the products can be measured, for example, by gas chromatography.

[0099] Plant data 228 can be stored in a database, for example, as historical data. Plant data 228 can be provided to a preprocessing engine, which preprocesses the data and provides plant-based input data to an analysis engine 230, which analyzes the plant-based input data and generates machine-readable instructions 236 for controlling and / or monitoring engine 232. Additionally or alternatively, analysis engine 230 can analyze and / or receive production and / or processing data 234 and generate machine-readable instructions 236 for controlling and / or monitoring engine 232. The generation of machine-readable instructions 236 can be performed automatically by analysis engine 230 based on the analysis of input plant data (i.e., without operator intervention). For example, analysis engine 230 can continuously receive input plant data and analyze the data in a continuous mode. When an anomaly occurs, analysis engine can generate machine-readable instructions 236 for controlling and / or monitoring engine to identify and / or remove the anomaly. The analytics engine can identify solutions for improving production efficiency by analyzing factory-based input data within the context of the production process, and can send machine-readable instructions 236 to the control and / or monitoring engine for improving production. The control and / or monitoring engine 232 can display push notifications to operators to review the machine-readable instructions 236, and based on this review, can generate machine-readable instructions 236. The control and / or monitoring engine 232 can change the operating parameters of one or more devices and sensors 204 based on the machine-readable instructions 236. Reviewing these machine-readable instructions 236 can allow the safe integration of trained transformer-based models into distributed production environments (such as chemical production).

[0100] Alternatively, or in addition to the machine-readable instructions 236 automatically generated by the analysis engine 230, the operator may prompt the analysis engine 230 to provide machine-readable instructions 236 based on prompts and / or queries. Prompts and / or queries may indicate context-dependent operations.

[0101] Operators can use machine-readable instructions 236 to control and / or monitor one or more production operations in a distributed production environment.

[0102] Machine-readable instructions 236 may include operational instructions for production, such as machine-readable instructions for controlling equipment and sensors 204 and / or machine-readable instructions for monitoring equipment and sensors 204. Machine-readable instructions 236 may also include operational instructions for operator control and / or monitoring of a distributed production environment, wherein the operator may be a human-based operator, a computer-based operating system, or a hybrid system including human operators and computer-assisted systems.

[0103] The operator can review the machine-readable instructions 236, and based on that review, can generate machine-readable instructions 236 for controlling the devices and sensors 204. The operator can prompt the analysis engine 230 to provide operating instructions (machine-readable instructions 236) based on prompts, queries, and / or context.

[0104] The operator may be a human operator reviewing the machine-readable instructions 236. The operator may be a human operator with a computer-assisted system for reviewing the machine-readable instructions 236. The operator reviewing the machine-readable instructions 236 may be a computer-based system based on a computer program that includes a set of instructions for reviewing the machine-readable instructions 236.

[0105] A control system for a distributed production environment may include one or more computing units that can manipulate one or more process parameters related to the production process by controlling one or more actuator or switch and / or end effector units (e.g., via manipulating one or more of the equipment operating conditions). Control is typically performed in response to one or more signals retrieved from the equipment.

[0106] The control and / or monitoring engine 232 may include one or more computer processors for modifying machine-readable instructions 236 and generating machine-readable instructions 236 for a control system used in distributed production. The control system for distributed production can adjust equipment operating conditions based on the machine-readable instructions 236, such that the adjusted process parameters and / or equipment operating conditions produce a controlled product (e.g., a chemical product) with one or more desired or predetermined characteristics or performance parameters. Therefore, production can be controlled in real time while ensuring that equipment operating conditions adapt to undesirable changes in process parameters.

[0107] Figure 3 A system for generating chemical product production and / or processing data related to the production and / or processing of chemical products is demonstrated.

[0108] The system may include a user interface 306 for receiving chemical product data 308 and / or for providing chemical product production and / or processing data 310. One or more production and / or processing data templates may be received based on the chemical product data 308. Therefore, one or more production and / or processing data templates can be selected from a plurality of production and / or processing data templates. Production and / or processing data templates may refer to production data templates and / or processing data templates. Database 302 may include a plurality of production and / or processing data templates. One or more production and / or processing data templates may be received from database 302, as in... Figure 4A and Figure 4B As described in the context, contextualized chemical product data can be generated based on received chemical product data 308 and one or more production data templates or one or more processing data templates. Contextualized chemical product data can be provided to the data-driven model 304 via an application programming interface (API) 312. The data-driven model 304 can provide chemical product production and / or processing data based on the received contextualized chemical product data. Chemical product production and / or processing data that can be generated by the data-driven model 304 can be provided via API 314. API 312 can be the same as or different from 314. Received chemical product production and / or processing data can be provided via a user interface 306 in response to being received and / or provided via API 314.

[0109] Figure 4A An example for selecting at least one production and / or processing data template is shown.

[0110] At least one production and / or processing data template 406 can be received from a database including production and / or processing data template 418. Figure 4AIn this context, chemical product data 402 may point to and / or link to one or more production and / or processing data templates 418. Specifically, chemical product data 402 may point to and / or link to at least one query associated with chemical product data 404. Preferably, the query associated with chemical product data 404 may point to and / or link to one or more production and / or processing data templates 418. The query associated with chemical product data 404 may be adapted to retrieve one or more production data templates and / or processing data templates from a database comprising multiple production and / or processing data templates 418. Based on the query associated with chemical product data 404, one or more production data templates or processing data templates 406 may be received. Therefore, receiving at least one production and / or processing data template 406 may include providing a request to a database comprising multiple production and / or processing data templates 418 for receiving one or more production and / or processing data templates 406, wherein the request may include the query associated with chemical product data 404. Furthermore, receiving at least one production and / or processing data template 406 may include receiving at least one production and / or processing data template 406 from a database based on a request already provided for receiving at least one production and / or processing data template 406. This is advantageous because receiving at least one production and / or processing data template 406 requires only lower computational resources.

[0111] Figure 4B Another embodiment for selecting at least one production and / or processing data template is shown.

[0112] At least one production and / or processing data template 412 can be received from a database comprising multiple production and / or processing data templates 414. To receive at least one production and / or processing data template 412, it can be done via, for example... Figure 8 The chemical product data 408 is embedded using the embedding layer described in the context. Embedding the chemical product data 408 can produce embedded chemical product data 410. The embedded chemical product data 410 can be a lower-dimensional representation than the chemical product data 408 and can be machine-processable. This can be achieved via, for example... Figure 8The embedding layer described in the context is used to embed production and / or processing data template 414. Embedding production and / or processing data template 414 can produce embedded production and / or processing data template 416. The similarity between embedded chemical product data 410 and embedded production and / or processing data template 416 can be compared. The more similar embedded chemical product data 410 and embedded production and / or processing data template 416 are, the closer the tensors associated with embedded chemical product data 410 and embedded production and / or processing data template 416 are likely to be. Therefore, in order to select one or more of the production and / or processing data templates 414 corresponding to chemical product data 408, embedded chemical product data 410 and embedded production and / or processing data template 416 are compared, for example, by calculating the dot product between two or more tensors associated with embedded production and / or processing data template 416 and embedded chemical product data 410. At least one production and / or processing data template 412 can be selected from production and / or processing data template 414 via the similarity between embedded chemical product data 410 and embedded production and / or processing data template 416. This selection can be independent of, for example, categories specified by a query associated with chemical product data 404, and can be based solely on content associated with chemical product data 408. Therefore, using embedding to retrieve data from the database improves the accuracy of retrieving at least one production and / or processing data template 412 that matches chemical product data 408.

[0113] Figure 5 Examples of methods for generating chemical product production and / or processing data related to the production and / or processing of chemical products are shown.

[0114] For example, via... Figure 3 The user interface described in the context is used to receive 502 chemical product data. Through the user interface, the user can select chemical product data and / or input chemical product data. For example, the user can select chemical product data from data obtained during the production and / or processing of the chemical product (particularly from historical data obtained during the production and / or processing of the chemical product).

[0115] Furthermore, you can select the 506 production and / or processing data template, such as in Figure 4A and Figure 4B As described in the context. For this purpose, more than 504 production and / or processing data templates can be provided before selection, and at least one production and / or processing data template can be selected from among the multiple production and / or processing data templates.

[0116] Two or more data-driven models may be available for providing contextualized chemical product data. These two or more models may include different model parameters that result in different outputs and / or different timing associated with providing chemical product production and / or processing data. For example, a first data-driven model among the two or more models may include fewer model parameters than a second data-driven model. Therefore, the first data-driven model may provide chemical product production and / or processing data faster than the second data-driven model. Conversely, the second data-driven model may provide more accurate chemical product production and / or processing data than the first data-driven model. Depending on the target output quality, a data-driven model may preferably be selected from multiple data-driven models. For this purpose, a user may receive 508 selections of one or more data-driven models via a user interface.

[0117] Furthermore, contextualized chemical product data can be generated by combining at least a portion of the production and / or processing data template with the chemical product data 510, as in... Figure 6 As described in the context.

[0118] Contextualized chemical product data can be fed into a data-driven model to generate chemical product data512. The data-driven model can be used as follows: Figure 6 and Figure 10 Training and / or parameterization can be performed as described in the context. Data-driven models can process contextualized chemical product data into chemical product production and / or processing data, as in... Figures 6 to 11 As described in the context.

[0119] For example, via Figure 3 The user interface described in the context of providing 514 chemical product production and / or processing data is preferably in response to receiving chemical product production and / or processing data from a data-driven model.

[0120] Figure 6 Another embodiment of a method for generating chemical product production and / or processing data related to the production and / or processing of chemical products is shown.

[0121] The received chemical product data 602 is used to generate contextualized chemical product data 606 based on a selected production and / or processing data template. The production and / or processing data template may indicate a structure associated with at least a portion of the chemical product data 602. For example, the production and / or processing data template may specify one or more portions of one or more sequences and one or more production and / or processing parameters located adjacent to one or more portions of one or more sequences. Production and / or processing parameters may be placeholders for the production and / or processing parameter values ​​to be inserted. To generate contextualized chemical product data, at least a portion of the chemical product data (preferably including the chemical product data and / or production and / or processing parameter values ​​associated with the chemical product data) may be inserted into the production and / or processing data template. The generated contextualized chemical product data may be input data to a data-driven model 624. The data-driven model may be and / or may include, as in... Figures 9A to 9C and / or Figure 11 One or more architectures described in the context of [the data-driven model]. A data-driven model can be like... Figure 10 Train as described in the context.

[0122] The generated contextualized chemical product data 606 is provided to a data-driven model 624, and / or the generated contextualized chemical product data 606 is received by the data-driven model 624. The data-driven model can be trained to embed the received contextualized chemical product data 606. For example, the contextualized chemical product data 606 may include one or more sequences, such as at least a portion of a sentence, one or more numbers, or combinations thereof. At least a portion of the sentence may be received via a user interface, such as in text form, audio form, etc.

[0123] To process the contextualized chemical product data 606 via the data-driven model 624, the data-driven model 624 can transform the contextualized chemical product data 606 into embedded contextualized chemical product data 618. Embedded contextualized chemical product data 618 can be a second-order tensor. Embedded contextualized chemical product data 618 can be represented in a machine-processable format and / or can be in a machine-processable format. Where the contextualized chemical product data 606 may include at least a portion of a sentence, words, numbers, and / or portions of words associated with the contextualized chemical product data 606 can be embedded. Words, numbers, and / or portions of words associated with the contextualized chemical product data 606 can be examples of elements associated with input data. Further examples of elements associated with input data can be numbers and / or at least a portion of a table including, for example, row and / or column numbers. Embedded contextualized chemical product data 618 can include one or more data points. Embedding contextualized chemical product data 606 enables computational processing of the contextualized chemical product data 606. Furthermore, embeddings (especially word embeddings) are an efficient data representation that uses less storage and requires less computational resources to process the corresponding data.

[0124] In particular, where contextualized chemical product data 606 may include at least a portion of a sequence, it can be as follows: Figures 9A to 9C The context describes the use of position encoding to consider the position of one or more elements associated with contextualized chemical product data 606. By applying position encoding to contextualized chemical product data 606, sequence-specific embedded contextualized chemical product data 618 can be obtained.

[0125] In an embodiment, the contextualized chemical product data 606 can be embedded in portions. For example, a first portion of the contextualized chemical product data 606 can be embedded to generate a first embedded contextualized chemical product data 618, and a second portion of the contextualized chemical product data 606 can be embedded to generate a second embedded contextualized chemical product data 618. The first embedded contextualized chemical product data 618 and the second embedded contextualized chemical product data 618 can be combined to generate embedded contextualized chemical product data 618. The first embedded contextualized chemical product data 618 and the second embedded contextualized chemical product data 618 can be combined by cascading a first vector associated with the first embedded contextualized chemical product data 618 and a second vector associated with the second embedded contextualized chemical product data 618. Cascading the first embedded contextualized chemical product data 618 and the second embedded contextualized chemical product data 618 can generate embedded contextualized chemical product data 618. Additionally or alternatively, the first contextualized chemical product data 606 and the second contextualized chemical product data 606 can be embedded together in a second-order tensor.

[0126] Embedded contextualized chemical product data 618 can be a second-order tensor representing contextualized chemical product data 606. The second-order tensor can be combined with itself and / or its transformations to form a tensor product, thereby generating a production and / or processing data context tensor 620, as shown in... Figures 9A to 9C and Figure 10 The production and / or processing data context tensor 620 may include a second-order tensor and / or may represent relationships between one or more elements associated with the embedded contextualized chemical product data 606, as described in the context.

[0127] In the contextualized chemical product data 606, which is sentence-based, the production and / or processing data context tensor 620 can represent at least a portion of words and / or numbers, as well as relationships between words associated with the production and / or processing data context tensor 620. This also applies to... Figure 11 The numerical inputs are described in the context of the chemical product production and / or processing data 612. By doing so, the relationship between two or more components of the contextualized chemical product data 606 is considered when generating the chemical product production and / or processing data 612. This makes it possible to determine the chemical product production and / or processing data 612 more accurately.

[0128] The production and / or processing data context tensor 620 can be transformed into chemical product production and / or processing data 612 via one or more machine learning architectures (e.g., linear layers, classification layers). Specifically, a first portion of the chemical product production and / or processing data 612, referred to herein as first chemical product production and / or processing data 612, can be generated based on embedded contextualized chemical product data 618. Second chemical product production and / or processing data 612 (as a second portion of the chemical product production and / or processing data 612, particularly a second portion of sequential chemical product production and / or processing data 612) can be generated and / or provided based on embedded contextualized chemical product data 618 and embedded first chemical product production and / or processing data 612, such as... Figure 10 The embedded first chemical product production and / or processing data 612 may refer to embedded first chemical product production and / or processing data 612 similar to embedded contextualized chemical product data 618.

[0129] Embedded contextualized chemical product data 618 and embedded first chemical product production and / or processing data 612 can be combined, for example, by cascading, to generate a production and / or processing data context tensor 620 at a second time step for generating second chemical product production and / or processing data 612. The second chemical product production and / or processing data 612 can be generated based on the production and / or processing data context tensor 620 at the second time step. This can be repeated for three or more portions of the chemical product production and / or processing data 612. Where the chemical product production and / or processing data 612 can be sentence-based, portions of chemical product data 602 and / or contextualized chemical product data 606 can refer to words and / or portions of words, particularly tokens associated with words. Portions of the chemical product production and / or processing data 612 can be generated based on previous portions of the chemical product production and / or processing data 612 until a complete chemical product production and / or processing data 612 is generated, typically indicated by an end marker.

[0130] In an embodiment, embedded contextualized chemical product data 618 can be transformed into a first production and / or processing data context tensor 620 and a second production and / or processing data context tensor 620. The first and second production and / or processing data context tensors 620 can be combined, for example, by cascading them into a production and / or processing data context tensor 620. The production and / or processing data context tensor 620 can be processed as described above to generate chemical product production and / or processing data 612. By considering a set of two or more production and / or processing data context tensors 620, generating the chemical product production and / or processing data 612 requires fewer computational resources and / or less time.

[0131] The data-driven model 624 may include one or more of a transformer decoder, a transformer encoder, a transformer encoder-decoder architecture, etc. The data-driven model 624 may be self-supervised trained based on a training dataset. The training dataset may include historical data related to chemical product data and chemical product production and / or processing data. The data-driven model may be trained to sequentially generate chemical product production and / or processing data 612 based on previously generated portions of chemical product data 602 and chemical product production and / or processing data 612. To this end, the data-driven model may be provided with contextualized chemical product data 606 to generate a first portion of chemical product production and / or processing data 612, and is provided with contextualized chemical product data 606 and the first portion of chemical product production and / or processing data 612 for generating a first portion of chemical product production and / or processing data 612 as specified by the training dataset. During training, ground truth values ​​of previous portions of chemical product production and / or processing data 612 as specified by the training dataset may be provided to the data-driven model. This self-supervised training enables data-driven models to learn independently of humans, resulting in efficient embeddings that require no human interaction. Therefore, it saves time by providing efficient learning while achieving highly accurate results correlated with chemical product production and / or processing data 612.

[0132] Figure 7A Another embodiment of contextualized chemical product data, as well as production and / or processing data, is presented.

[0133] Users may want to identify suppliers of chemical products for a specific application by providing chemical product data via a user interface. A corresponding production and / or processing data template can be selected. Chemical product data can be inserted into the production and / or processing data template, generating contextualized chemical product data 704. The data-driven model can generate chemical product production and / or processing data 706 in response to the provided contextualized chemical product data 704. Users may want further information and provide second contextualized chemical product data 708. In this example, the second contextualized chemical product data may include further queries. Based on the provided second contextualized chemical product data, the data-driven model can provide second chemical product production and / or processing data.

[0134] Figure 7B Another embodiment of contextualized chemical product data, as well as production and / or processing data, is presented.

[0135] Users may expect to identify potential consumers of Company B's chemical products via a user interface. Therefore, chemical product data can be received. Corresponding production and / or processing data templates can also be received. Inserting chemical product data into the production and / or processing data templates can generate contextualized chemical product data 712. Chemical product production and / or processing data 714 can be generated by a data-driven model in response to the provided contextualized chemical product data. Users may expect further information and be provided with second contextualized chemical product data 716. Based on the provided second contextualized chemical product data 710, the data-driven model can provide second chemical product production and / or processing data.

[0136] Figure 8 An example of obtaining an embedding layer is shown. An embedding layer can be obtained by training, for example, a continuous bag-of-words (CBOW) model or a skip-gram model.

[0137] An embedding layer can be adapted to generate embedded input data based on input data. Generating embedded input data can refer to embedding input data. Embedding input data can produce a representation associated with the input data. Therefore, embedded input 814 can be a representation associated with the input data. The input data can include one or more elements. One or more elements can be represented by input vector 806. In particular, embedded input 814 and / or input vector 806 can be machine-readable and / or processor-processable. For this purpose, embedded input 814 and / or input vector 806 can be tensors, particularly first-order tensors. Specifically, input vector 806 can be a one-hot vector, or the sum of multiple one-hot vectors. A one-hot vector can be a vector with a non-zero entry. Examples of one-hot vectors can be 808, 810, and 812. The non-zero entry in the one-hot vector and / or input vector 806 can indicate elements. For example, a lookup table can define the relationship between the position of a non-zero entry and the element indicated by the one-hot vector. The lookup table can specify multiple distinct elements. The number of distinct elements can be equal to the number of entries in the one-hot vector. The number of distinct elements can be referred to as vocabulary. In the example, elements can be represented by tags, and a sequence of elements can refer to at least a portion of a sentence. At least a portion of a sentence can be represented by multiple tags. Tags can represent at least a portion of elements and / or words. For example, in cases where an element will be associated with only one word, words such as “embeddings,” “embedding,” or “embed” will constitute distinct elements. The first tag can represent the stem “embed,” while the endings, which typically appear in multiple words, can be represented by the second, third, and fourth tags. The second, third, and fourth tags can be used to represent other words, such as “look,” “looking,” etc., preferably in conjunction with the fifth tag representing the stem “look.” Ultimately, this tagging of elements associated with multiple stems and multiple endings results in fewer tags for representing multiple elements and thus uses fewer computational resources.

[0138] A lookup table specifying a subset of the vocabulary of a language such as English can include 10,000 words or more. Embedded input 814 can be a lower-dimensional representation than input vector 806. For example, a typical embedded input 814 can include hundreds of different entries. Therefore, embedded input 814 constitutes a denser representation of one or more elements using fewer computational resources. Furthermore, embedded input 814 can represent relationships between two or more elements. For example, the words “Italy” and “Germany” can be similar or more closely related because they both define European countries, while the word “example” may be quite different from these two corresponding words. The smaller the dot product between two embedded inputs 814, the more similar the two elements associated with the embedded input 814 can be. Therefore, embedded input 814 can accurately represent one or more elements and produce accurate results based on the processing of embedded input 814.

[0139] To transform input vector 806 into embedded input 814, the embedding layer can include a number of neurons equal to the number of entries in embedded input 814. Based on embedded input 814, the output layer can generate output vector 816. The output vector can be a vector and / or can indicate one or more elements. Output vector 816 can indicate one or more elements that are different from input vector 806 and / or different from the one-hot vector associated with input vector 806. For this purpose, the output layer can include a number of neurons equal to the number of entries in input vector 806 and / or output vector 816. The output layer can apply a softmax function to embedded input 814. By doing so, the output vector can include probabilities associated with elements associated with non-zero entries in output vector 816. Therefore, one or more elements with corresponding probabilities can be obtained from output vector 816. Where input vector 806 can specify one or more sequences of elements, output vector 816 can specify one or more elements corresponding to the sequence(s) of elements(s) specified by input vector 806. Figure 8 In the example, the element associated with vector 818 corresponds to the input vector with a 71% probability. Additional or substitute elements can correspond to the input vector as indicated by the output vector with a lower probability. By defining a threshold that can be compared with the probability, the selection of corresponding elements can be customized to the user's needs. Elements generated by the model, including embedding layer 802 and output layer 804, can refer to the most probable element indicated by output vector 816. Therefore, Figure 8 The model described can generate elements associated with vector 818 with a 71% confidence level.

[0140] Figure 8The model can be a Continuous Bag-of-Words (CBOW) model. A CBOW model can be trained on a training dataset that includes multiple input vectors and corresponding output vectors. Since the training dataset may be unlabeled, training the CBOW model can be referred to as self-supervised training. Before training the CBOW model, it can be initialized with random values ​​of the weights assigned to neurons. During training, the input vectors can be used to initialize the embedding and output layers, and the loss can be determined by comparing the output vector obtained by passing the input vector 806 through the model with the output vector corresponding to the input vector 806 as specified by the training dataset. Based on the determined loss, backpropagation can be applied to determine the gradients associated with the neurons in the embedding layer 802 and the output layer 804 to reduce the loss. The weights of the neurons can be updated using a gradient descent algorithm based on the determined gradients. If the CBOW model achieves the predetermined loss, training can be terminated, and the trained CBOW model is obtained. Based on the trained CBOW model, the embedding layer 802 can be adapted to embed input data including one or more elements. This embedding layer 802 can be used in other machine learning architectures that require an embedding layer 802, such as in... Figure 9A , Figure 9B and Figure 9C The context describes the transformer encoder, transformer decoder, or transformer encoder-decoder architecture. Training these architectures may require trained embedding layers 802. Therefore, models such as the CBOW model can be trained before training the transformer encoder, transformer decoder, or transformer encoder-decoder architecture.

[0141] Figure 9A An embodiment of a transformer encoder architecture is shown. The transformer encoder includes an encoder input 978, one or more encoder blocks 974, 914, and an encoder output.

[0142] Transformer encoder architectures can be derived from, as is known in the art, and such Figure 9CThe transformer encoder-decoder architecture is shown. Specifically, the transformer encoder can be referred to as an X-former. The transformer encoder architecture can correspond to an encoder architecture associated with the transformer encoder-decoder architecture, but with additional encoder outputs instead of directly connecting the encoder block to the decoder of the transformer encoder-decoder architecture. Various transformer encoder architectures are available in the art, such as the bidirectional transformer encoder representation (BERT).

[0143] Input data can be received at encoder input 978. The input data can be contextualized chemical product data. Input embedding 902 can be applied to encoder input 978. Applying input embedding 902 can instruct the input data to pass through an embedding layer, for example, as in... Figure 8 As described in the context. Applying input embedding 902 to contextualized chemical product data can produce embedded contextualized chemical product data. Applying input embedding 902 to contextualized chemical product data can produce embedded contextualized chemical product data. This can be... Figure 6 Further description in the context.

[0144] The encoder input 978 can be equipped with position encoding 904. Applying position encoding 904 refers to adding position factors to the embedded input obtained via input embedding. Preferably, the input data can specify a sequence of elements. Position factors It can indicate the position of an element within a sequence. For example, the position factor can be obtained based on the following equation. :

[0145]

[0146]

[0147] Here, pos can refer to the position of an element within the sequence, I can refer to the dimension associated with the input embedding, and d can refer to the dimension of the model (e.g., a transformer decoder, transformer encoder, or transformer encoder-decoder). This can be referred to as absolute position embedding. Alternatively, positional encoding can be based on Rotated Position Embedding (RoPE). Positional encoding is advantageous because it enables the processing of sequential data without requiring further dimensions indicating the position of each element. Therefore, positional encoding reduces the computational resources required to embed the input data.

[0148] By passing the input data through the encoder input, the input data can be transformed into a second-order tensor representing a sequence of elements. This second-order tensor can be called embedded input data. Embedded input data can be processed by the encoder block. Embedded input data can be provided to layer normalization 908 via residual connections. Multi-head self-attention 906 can be applied to the embedded input data. Multi-head self-attention 906 can include two components: multi-head and self-attention. Self-attention can be understood as a filter applied to the embedded input data. By applying a filter to the embedded input data, elements associated with the embedded input data that contribute to the output data to be generated can be identified to generate the output data. Therefore, a filter can represent the degree of contribution of elements associated with the embedded input data to the output data to be generated. Applying a filter can be called weighting the elements associated with the embedded input data. This is particularly advantageous for long sequences of elements. Filters can be learned and improved during training by learning to identify the contributions of elements associated with the embedded input data. For example, in a partial sentence "I went to the bakery to buy a," the last word can be generated by a data-driven model such as a transformer encoder. Self-attention allows the transformer encoder to focus primarily on the words "bakery" and "buy" to generate the word "bread." Self-attention can refer to attention generated based on input data. Therefore, filters can be determined based on the input data, preferably embedded input data. Embedded input data can be used as the query Q, key K, and value V for the self-attention operation. Self-attention can refer to attention based on the received input data. Therefore, filters can be computed based on the following formula by inserting the corresponding tensor based on the embedded input data:

[0149]

[0150] in, The dimension corresponding to the key.

[0151] To further improve the efficiency of the transformer encoder, multiple heads are used to apply filters, resulting in a multi-head self-attention 906. A multi-head self-attention 906 can include applying filters to two or more parts of the embedded input data. Therefore, a tensor can be split into two or more parts, and filters can be applied to each of these parts separately via two or more heads according to the following equation:

[0152] Where the parameter matrix Where I can refer to the number of heads. , and It can refer to the value, key, and query dimension.

[0153] The results of two or more heads can be cascaded according to the following equation:

[0154] in And h can refer to the number of heads.

[0155] Embedded input data can be transformed into a context tensor via multi-head self-attention 906. The context tensor can represent a sequence of elements in the input data and the relationships between two or more elements. Transforming the embedded contextualized chemical product data can produce a processing data context tensor 622. Transforming the embedded contextualized chemical product data 618 can produce a production and / or processing data context tensor 620. Therefore, the context tensor can be a production and / or processing data context tensor 620. The context tensor can be a second-order tensor and / or may include one or more first-order tensors. After multi-head self-attention 906, layer normalization 908 can be applied based on the context tensor and / or the embedded input data from the residual connections. Applying layer normalization 908 can refer to normalizing the context tensor.

[0156] Normalizing the context tensor reduces the value of the entries in the context tensor. This reduces the computational cost associated with processing the context tensor. After layer normalization 908, the context tensor can be passed back to the feedforward layer 910, followed by layer normalization 912 based on the residual connections to the context tensor and / or the output of the feedforward layer 910. The feedforward layer 910 can be a feedforward neural network. The feedforward neural network can include multiple fully connected neurons. Passing the context tensor through the feedforward neural network can result in a linear transformation of the context tensor. Additionally or alternatively, the neural network can include one or more activation functions, such as rectified linear units (ReLU). Therefore, the neural network can be configured to perform one or more nonlinear operations on the context tensor and / or nonlinearly transform the context tensor. After the context tensor has been transformed and / or normalized by the feedforward layer 910 and layer normalization 912, the context tensor can be provided to one or more additional encoder blocks 914. Passing the context tensor through the feedforward layer 910 allows it to be adapted for processing by the attention layers of one or more further encoder blocks 914, enabling the application of self-attention filters, preferably multi-head self-attention 906. The context vector, after being transformed by layer normalization 912 and the feedforward layer 910, can be referred to as the hidden state.

[0157] The encoder output 976 includes a linear layer 916 and a softmax layer 918. The linear layer 916 transforms the context vector into a logits vector. The linear layer can be fully connected. The logits vector obtained by passing the context tensor through the linear layer 916 can be passed through the softmax layer 918. Passing the logits vector through the softmax layer 918 refers to applying the softmax function to the logits vector. Applying the softmax function to the logits vector produces a probability distribution of one or more elements corresponding to the sequence of elements in the input data. Based on predefined selection criteria, one or more elements can be selected according to the probability distribution. The one or more selected elements can be referred to as one or more elements generated by the transformer encoder. One or more generated elements can be provided to the encoder input to generate another one or more elements corresponding to the sequence of input data and the one or more elements generated by the transformer encoder, as shown in... Figure 10 As described in the context.

[0158] The output data from encoder output 976 can be chemical product production and / or processing data. Therefore, the result of transforming the production and / or processing data context tensor 620 can be chemical product production and / or processing data.

[0159] Figure 9B An example of a transformer decoder architecture is shown.

[0160] Input data, embedded input data, context tensors, and / or output data can be as follows: Figure 9A Defined in the context of.

[0161] The transformer decoder includes a decoder input 984, one or more decoder blocks 980, 932, and a decoder output 992. The transformer decoder architecture can be derived from, as is known in the art and as... Figure 9C The transformer encoder-decoder architecture is shown. 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, but does not depend on receiving one or more hidden states from the encoder of the transformer encoder-decoder. There are various transformer decoder architectures available in the art, such as the general pre-trained transformer (GPT).

[0162] The decoder input 984 can be applied to, for example, in Figure 9AThe input embedding 902 and position encoding 904 described in the context are similar to the input embedding 920 and position encoding 922.

[0163] Decoder block 980 may include layer normalization 926, masked multi-head self-attention 924, feedforward layer 928, and / or layer normalization 930. Embedded input data generated by passing input data through decoder input 984 can be provided to layer normalization 926 via residual connections. Further, masked multi-head self-attention 924 can be applied to the embedded input data. Masked multi-head self-attention 924 corresponds to... Figure 9A The multi-head self-attention 906 described in the context further masks the portion of the embedded input data associated with elements in the sequence that are later than the elements to be generated. Additionally or alternatively, the portion of the input data associated with elements in the sequence that are later than the elements to be generated may not be received and / or transformed into the embedded input data. Therefore, the transformer decoder can be adapted to generate subsequent elements of the sequence, while the transformer encoder can be adapted to generate missing elements within a sequence and / or between two or more sequences. Thus, the transformer encoder can be configured for classification tasks. The transformer decoder can be configured for text generation.

[0164] Similar to Figure 9A The transformer encoder described in the context can generate a context tensor by applying masked multi-head self-attention 924 and layer normalization 926. The context tensor can be provided to layer normalization 930 via residual connections. Further, the feedforward layer 928 and layer normalization 930 can be similar to... Figure 9A The context tensor is described in the feedforward layer 910 and layer normalization 912. The context tensor can be provided to one or more additional decoder blocks 932.

[0165] The decoder output 992 may include a linear layer 934 and a softmax layer 936. The linear layer 934 and softmax layer 936 can be similar to... Figure 9A The linear layer 916 and the softmax layer 918 are described in the context of this study.

[0166] Figure 9C An example of a transformer encoder-decoder architecture is shown.

[0167] Input data, embedded input data, context tensors, and / or output data can be as follows: Figure 9A Defined in the context of.

[0168] The transformer encoder-decoder may include an encoder input 988, one or more encoder blocks 986, 964, a decoder input 994, a decoder block 990, and a decoder output 992. The encoder input 988 may correspond to... Figure 9A The encoder input terminal 978. One or more encoder blocks 986, 964 can correspond to Figure 9A One or more encoder blocks 974, 914. Decoder input 994 can correspond to... Figure 9B The decoder input is 984.

[0169] Decoder block 990 may include, as in Figure 9B The masking multi-head self-attention 924, layer normalization 926, feedforward layer 928, and layer normalization 930 described in the context are similar to the masking multi-head self-attention 970, layer normalization 972, feedforward layer 938, and layer normalization 940. Decoder block 990 may further include multi-head self-attention 950 and layer normalization 948. Similar to... Figure 9B The description suggests that the context tensor can be obtained from masked multi-head self-attention 970 and layer normalization 972. This can be combined with... Figure 9A Multi-head self-attention 906, similar to multi-head self-attention 950, is applied to the context vector obtained from layer normalization 972 and the hidden states of one or more encoder blocks 986, 964. Layer normalization 948 can be applied to the context vector obtained from multi-head self-attention 950 and the context vector provided via residual connections from layer normalization 972. Similar to... Figure 9B The description suggests that the context vector generated by layer normalization 948 can be processed via feedforward layer 938 and layer normalization 940. The context vector generated by layer normalization 940 can be provided to another decoder block 942 similar to decoder block 990. The context vector obtained from one or more decoder blocks 990, 942 can be provided to decoder output 992. Decoder output 992 can correspond to... Figure 9B The decoder output is 982.

[0170] Using the above architecture, the transformer encoder-decoder can receive and process input data at encoder input 988 and one or more encoder blocks 986, 964, as well as decoder block 990 and decoder output 992. Based on the input data, the transformer encoder-decoder can generate output data partially or sequentially. Sequentially generated output data can be provided to decoder input 994, one or more decoder blocks 990, 942, and decoder output 992, and / or processed by them. Preferably, the sequence can be provided to encoder input 988, and decoder input 994 can be provided with at least a portion of the elements of the generated output data after at least a portion of the output data has been generated. By doing so, by considering both the input data and the generated output data, subsequent elements of the output data can be generated with higher accuracy because the transformer encoder-decoder can receive more data over time.

[0171] Due to the transformer encoder-decoder architecture, the transformer encoder-decoder can be configured to transform a sequence into another representation of the sequence. An example of transforming a sequence into another representation could be translating a sentence into another language. Various transformer encoder-decoders are available in the art, such as BART, T5, etc.

[0172] In an embodiment, layer normalization 908, 912 can be applied before masking multi-head self-attention 924, multi-head self-attention 906, and / or feedforward layer 910 in the transformer decoder, transformer encoder, and / or transformer encoder-decoder. By doing so, computational resources used to apply multi-head self-attention 906 and / or feedforward layer 910 to the embedded input data and / or context tensors can be reduced, since the number of entries for the corresponding tensors may be lower after normalization.

[0173] In an embodiment, the decoder output 992 may include a classification neural network, additional feedforward layers, convolutional layers, fully connected layers, etc. For example, the transformer encoder-decoder can be configured to select among multiple options. To this end, the transformer encoder-decoder can be provided with three different input datasets and can classify the context vectors obtained from one or more decoder blocks 990 via one or more linear layers. Therefore, the architecture can be extended according to the use case to be solved.

[0174] Figure 10Examples of training and / or deploying transformer encoders, transformer decoders, and / or transformer encoder-decoders are shown.

[0175] The encoder / decoder / encoder-decoder architecture 1002 can correspond to, for example, in... Figures 9A to 9C The transformer decoder, transformer encoder, and / or transformer encoder-decoder are described in the context. Input data, embedded input data, context tensors, and / or output data can be as follows: Figure 9A Defined in the context of.

[0176] The output data generated by the encoder / decoder / encoder-decoder architecture 1002 may include one or more elements, specifically a sequence of elements. Previously generated elements of the output data may be provided as input for generating the next element in the output data sequence.

[0177] exist Figure 10In the example, the input data may include N elements, specifically input tokens. Input tokens may be tokens specifically designed for input into a data-driven model such as a transformer decoder, transformer encoder, or transformer encoder-decoder. The output data to be generated may include M elements. The encoder / decoder / encoder-decoder architecture 1002 may generate one element of the output data in one time step based on the received input data and optionally previously generated output data elements. Therefore, M time steps are required to generate M elements. The time steps include providing inputs 1010, 1012, and 1014 to the encoder / decoder / encoder-decoder architecture 1002 and receiving output data 1004, 1008, and 1006 from the encoder / decoder / encoder-decoder architecture 1002. In the first time step, input 1010 may include N input tokens. The N input tokens may, for example, be associated with N words, stems, or endings. Preferably, the N input tokens may specify a question. One or more input tokens may specify the start and / or end of a token sequence. Input 1010 can be processed by encoder / decoder / encoder-decoder architecture 1002. Based on input 1010, at least a portion of output data 1004 can be generated. At least a portion of the output data may include a first output marker. In the next time step, the generated first output marker may be provided together with input 1012. Specifically, if input 1012 can be received by transformer encoder-decoder, the input marker may be received at encoder input 988, and the first output marker may be received at decoder input 994. If input 1012 can be received by transformer encoder, input 1012 may be received at encoder input 978, and the same applies to transformer decoder and decoder input 984. Based on input 1012, output data 1008 including a first output marker and a second output marker can be generated. Generating output data 1008 based on input 1012 may refer to generating the second marker based on the first marker and N input markers, where the first marker may have already been generated based on the N input markers. This process can be repeated until the last marker in the sequence of output data 1006 can be generated. Preferably, the last marker can be an end marker. The end marker can terminate the generation of further output markers.

[0178] Similar to the data processing during the deployment of the encoder / decoder / encoder-decoder architecture 1002, the encoder / decoder / encoder-decoder architecture 1002 can be trained. The training dataset can include multiple sequences containing multiple elements. These sequences can be associated with input data and / or output data. Additionally or alternatively, these sequences can be independent of the input data and / or output data. For example, where the input data and output data can refer to chemical compositions represented via text, the training dataset can include sequential text data independent of chemical composition. In this example, the training dataset can include word sequences derived from a dialogue. In embodiments, the training dataset can at least partially include the input dataset and / or the output dataset.

[0179] Training can 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 can be initialized randomly. Additionally or alternatively, training can be performed as described in... Figure 8 The CBOW model or skip-gram model described in the context is used to obtain the input embeddings of the encoder / decoder / encoder-decoder architecture 1002. The trained embedding layer can be used during training. The parameters associated with the embedding layer can remain constant and / or can be updated after a predefined number of training epochs. By doing so, fewer parameters need to be updated, resulting in faster training with less computational resource consumption. Furthermore, the accuracy associated with the embedding layer can remain constant and / or can be improved by avoiding error compensation associated with the newly initialized encoder / decoder / encoder-decoder architecture 1002.

[0180] During training of the encoder / decoder / encoder-decoder architecture 1002, at least a portion of the sequence of the training dataset can be provided to the encoder / decoder / encoder-decoder architecture 1002 one after another, and one or more elements can be generated one after another based on the sequence of the training dataset. Elements generated based on the sequence can follow elements of the sequence portions that may be provided to the encoder / decoder / encoder-decoder architecture 1002. The generated one or more elements can be compared with one or more elements as specified by the training dataset that follow at least a portion of the sequence provided to the encoder / decoder / encoder-decoder architecture 1002. Therefore, during training, the encoder / decoder / encoder-decoder architecture 1002 can generate a guess about the next element, and the guess about the next element in the sequence can be compared with the ground truth value of the actual next element specified according to the training dataset. Based on the guess about the next element and the ground truth value, a loss can be determined. The loss can define the similarity between the guess about the next element and the ground truth value. The loss can be determined by forming a vector dot product between the labels associated with one or more elements and the labels associated with the ground truth value. When the loss is not zero, it may be necessary to update the parameters associated with the encoder / decoder / encoder-decoder architecture 1002. Preferably, the parameters associated with the encoder / decoder / encoder-decoder architecture 1002 can be independent of the embedding layer. For example, the parameters associated with the encoder / decoder / encoder-decoder architecture 1002 can be the weights of the neurons in the encoder / decoder / encoder-decoder architecture 1002.

[0181] Based on the determined loss, backpropagation can be applied to determine the gradients associated with the parameters of the encoder / decoder / encoder-decoder architecture 1002 to reduce the loss. Based on the determined gradients, the parameters associated with the encoder / decoder / encoder-decoder architecture 1002, preferably the weights of the neurons associated with the encoder / decoder / encoder-decoder architecture 1002, can be updated using a gradient descent algorithm.

[0182] The training dataset can be unlabeled. The sequence of elements in the training dataset can inherently include ground truth values ​​for determining the loss based on one or more elements generated during the training of the encoder / decoder / encoder-decoder architecture 1002. Therefore, the encoder / decoder / encoder-decoder architecture 1002 can be self-supervised trained. This is advantageous because it saves time and resources used to create labeled training datasets. Furthermore, it enables the use of large training datasets associated with several megabytes in size. Thus, the data-driven model can be accurate in generating the elements of the sequence. Additionally, the large training dataset enables few-shot or even zero-shot predictions. Therefore, the data-driven model trained as described above is general-purpose and helps save resources required to train and / or host multiple goal-oriented models (such as CNNs). This training can be referred to as pre-training. The data-driven model can be configured to perform few-shot or even zero-shot predictions for multiple use cases after pre-training. The performance of the data-driven model can be further improved through additional training, referred to as fine-tuning.

[0183] Figure 11 An embodiment for input embedding for embedding numbers and / or tables is illustrated. Input data, embedded input data, context tensors, and / or output data can be as follows: Figure 9A Defined in the context of.

[0184] Where the sequence of elements associated with the input data (preferably included in the input data) can be of one type, it is possible to apply, as in Figures 9A to 9C The input embeddings described in the context are 902, 920, 952, and 966. For example, the type of input data can be text, where elements can be associated with at least a portion of a word, punctuation characters, a start marker indicating the beginning of one or more sequences associated with the input data, and / or an end marker. In another example, the input data can be at least partially numeric. Therefore, the input data can include multiple numbers. Numeric input data can be, for example, tabular data. Tabular data can specify one or more rows and / or one or more columns. Therefore, tabular data can include one or more cells, where these cells can be associated with one or more numeric values.

[0185] Numeric input data may require different embeddings than text input data. Input embeddings for numeric input data can include token embeddings, positional embeddings, column embeddings, row embeddings, or combinations thereof.

[0186] Applying token embeddings to one or more elements (specifically, tokens associated with input data) can produce a machine-processable representation associated with those elements (specifically, tokens). Applying token embeddings to one or more elements can instruct those elements through an embedding layer, for example, as in... Figure 8 As described in the context of [the previous section], a tag embedding can specify one or more elements, specifically tags, in a machine-processable representation. For example, a tag embedding can transform a numerical value into a vector. This is advantageous because such a representation can be enriched with further information, such as the position of the tags within a sequence and / or within a table associated with the tag sequence. Positional embeddings can be similar to [the previous section]... Figure 8 , Figures 9A to 9C The location embedding described in the context of [the table / table] is as follows. When the input data can be tabular data, column embedding can be applied. Applying column embedding to one or more elements (particularly tags associated with the input data) can produce a machine-processable representation specifying the position of one or more elements within Table 1102 (preferably within columns of Table 1102). Applying column embedding can refer to adding column factors to input data embedded via tag embedding, particularly embedded input data. Column factors can be the same for elements associated with the same column, and / or may differ between two or more elements associated with different columns. Similarly, when the input data can be tabular data, row embedding can be applied. Applying row embedding to one or more elements (particularly tags associated with the input data) can produce a machine-processable representation specifying the position of one or more elements within Table 1102 (preferably within rows of Table 1102). Applying row embedding can refer to adding column factors to input data embedded via tag embedding, particularly embedded input data. Row factors can be the same for elements associated with the same row, and / or may differ between two or more elements associated with different rows.

[0187] In this embodiment, the input data may be at least partially numerical and at least partially text. Therefore, the input data may include two or more types of data. The data type may refer to a modality. Therefore, different embeddings can be applied to the input data. For the text-containing portions of the input data, embeddings can be applied... Figure 8 , Figures 9A to 9CThe input embeddings mentioned above. For the portion of input data that is embedded as numeric tokens, positional embedding, column embedding, and row embedding can be applied. Furthermore, segmented embeddings can be applied to the input data independently of its type. Segmented embeddings can specify the type of input data that can be associated with one or more elements. For example, if the input data includes text and numbers, the input data can include both types of input data. Applying segmented embeddings to input data can refer to adding segmentation factors to the input data, preferably embedding the input data and / or the input data after tokenized embeddings have been applied. Segmentation factors can specify the type of data associated with one or more elements. Segmentation factors can 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.

[0188] Applying marker embedding, position embedding, segment embedding, column embedding, row embedding, or combinations thereof can generate embedded input data and / or can be the output of any of encoder inputs 978, 984, 988 or decoder inputs 984, 994. Data obtained by applying marker embedding, position embedding, segment embedding, column embedding, row embedding, or combinations thereof can be processed by encoder blocks 974, 986, decoder blocks 980, 990, encoder output 976, and decoder outputs 992, 982.

[0189] Figure 12 An example of the Mamba architecture is shown. The Mamba architecture can be used as a data-driven model. The Mamba architecture can improve inference speed associated with transformer-based models.

[0190] The layered Mamba architecture can be similar to the one about Figure 12 The transformer decoder architecture is discussed. However, instead of stacking decoder blocks, Mamba blocks 1232 and 1204 are stacked. Mamba block 1232 can be based on a selective spatial state sequence model (S6).

[0191] The input labels can be linearly projected into an expanded latent space via linear layers 1212 and 1220 (this allows more information to be captured during processing in the selective state space layer 1210), followed by convolution via convolutional layer 1214 and a nonlinear function (e.g., a sigmoid linear unit (SiLu) or a swish activation function). The convolution preceding the selective state space layer 1210 prevents independent label computation. The selective state space layer 1210 performs selective state space operations. Furthermore, learnable skip connections can be provided via linear layer 1220, which can map the input to the output using a linear transformation, similar to residual connections in a transformer model, which can help mitigate the vanishing gradient effect.

[0192] The selective state-space layer 1210 can be a linear recurrent network that selectively processes data based on input labels, allowing focus on relevant data and discarding irrelevant data. For example, in each step, a separate weight vector can be determined based on the corresponding input labels. The determined weight vector can then be used in a selective scan.

[0193] The selective state-space layer 1210 can be used, for example, in convolutional mode for parallelizable training, and in recurrent mode for generating output data in near-constant time. State-space operations can be based on solving state equations and output equations, where the state equations describe how the state changes based on the influence of the input, and the output equations describe how the state is transformed into the output. Furthermore, how the input influences the output can be represented by a learnable linear transformation (e.g., matrix D) used in learnable skip connections.

[0194] The state equation for the hidden state can be (in discretized form):

[0195]

[0196] The output can be represented by the following formula (in discretized form):

[0197]

[0198] This discretized state-space model can be expanded into a recurrent form similar to a recurrent network, illustrating that a selective state-space model can be or includes a linear recurrent model. However, here, matrices A, B, and C can also be used as the kernel for the convolution of the state-space model. The kernel K used for this can be, for example:

[0199]

[0200] This allows the output to be determined:

[0201]

[0202] Therefore, in this representation of the state-space model, training can be performed in parallel, just like in a convolutional neural network.

[0203] Matrix A can be a matrix that represents the most recent tags well and decays the older tags, and can be initialized using HiPPO:

[0204]

[0205] In this configuration, each entry below the diagonal is set to 0. This allows for the creation of long-term memory for selective state-space models.

[0206] For Mamba block 1232, matrices B and C, as well as the stride Δ used for matrix discretization, can depend on the input labels and can be trained during training to determine different matrices B and C for each input label. This enhances content awareness and can function similarly to multi-head self-attention in transformer models. However, unlike spatial state models with fixed matrices A, B, and C, the convolutional representation here may not be easily determined. Therefore, to operate the selective state space layer 1210 in convolutional mode, selective scanning can be applied using the associative properties of hidden state computation, allowing the sequences to be determined in parallel in parts and iteratively combined, enabling parallel training. Further read and write operations can be reduced by kernel fusion of the described stride, selective scanning, and multiplication with C.

[0207] Linear layer 1202 can project the generated output back to the same dimension as the input.

[0208] Mamba blocks can be used with transformer decoder blocks or hybrid expert blocks (e.g., decoder blocks where feedforward layers are replaced by gated networks and multiple parallel feedforward layers, with the gated networks switching between feedforward layers based on the input), which allows the advantages of different architectures to be taken advantage of.

[0209] An architectural example of the Mamba block can be found in Albert Gu and Tri Dao’s “Mamba: Linear-TimeSequence Modeling with Selective State Spaces” arXiv:2312.00752v2 [cs.LG] May 31, 2024, which is incorporated herein by reference.

[0210] The prior art disclosure in the publication (Issue 684; paragraphs

[1000] to

[8005] ; ISSN: 2198-4786; publication date: February 12, 2024) shall be regarded as reference RF1, which is incorporated herein by reference in its entirety. Preferably, the (chemical) product is the product as described in reference RF1; paragraphs

[1000] to

[8005] . Preferably, the methods / processes described herein are further methods / processes for producing the product.

[0211] The conversion step for obtaining the product preferably includes one or more steps as described below, and can be performed by conventional methods well known to those skilled in the art. The conversion step preferably includes one or more steps selected from the following:

[0212] Recovery, preferably through depolymerization, gasification, pyrolysis, and / or steam cracking; and / or

[0213] Purification, preferably crystallization, (solvent) extraction, distillation, evaporation, hydrogenation, absorption, adsorption, and / or ion exchange; and / or

[0214] Assembly, preferably foaming, synthesis, chemical transformation, chemical conversion, polymerization and / or compounding; and / or

[0215] Forming, preferably foaming, extrusion and / or molding; and / or

[0216] Finishing, preferably coating and / or smoothing.

[0217] Additionally, one or more steps are described in detail in reference RF1; paragraphs

[1000] through

[8005] .

[0218] This disclosure has also been described in conjunction with various preferred embodiments and examples. However, by studying the accompanying drawings, this disclosure, and the claims, those skilled in the art, as well as those practicing the claimed subject matter, will understand and implement other variations. It is particularly noteworthy that any steps presented can be performed in any order; that is, this disclosure is not limited to a specific order of these steps. Furthermore, it is not required that different steps be performed at a specific location or node in a distributed system; that is, each step can be performed on different nodes using different devices / data processing.

[0219] The order of all method steps presented above is not mandatory, and alternative orders are possible. However, the specific order of method steps shown as examples in the accompanying drawings should be considered as a possible order of method steps, for example, of the corresponding embodiments described in the corresponding drawings or of embodiments including at least some of the steps described in the corresponding drawings.

[0220] In this specification, any connections presented in the described embodiments should be understood in a manner that the components involved are operatively coupled. Therefore, these connections may be direct or indirect, have any number or combination of intermediate elements, and there may be only a functional relationship between the components.

[0221] As used herein, "determine" may also include "initiate or cause determination," "generate" may also include "initiate and / or cause generation," "provide" may also include "initiate or cause determination, generation, selection, sending, and / or transmission," and "acquire" may also include "initiate or cause determination, generation, selection, retrieval, and / or reception." "Initiating or causing an action" may include any processing signal that triggers a computing node or device to perform the corresponding action.

[0222] The term "comprising" or "including" should be understood in an open-ended sense, meaning that the object "including element A" may also include other elements besides element A. Furthermore, the term "comprising" or "including" may be limited to "consisting of," meaning consisting only of the specified element.

[0223] The indefinite article "a" or "an" should not be interpreted as "one," meaning that the use of the expression "an element" does not exclude the presence of other elements. A single element or other unit can perform the function of several entities or items recited in the claims. The fact that certain measures are recited only in mutually different dependent claims does not indicate that a combination of these measures cannot be used in advantageous implementations or can include additional elements.

[0224] The expressions “A and / or B” and “at least one of A or B” are considered interchangeable and are intended to include any of the following three cases: (i) A, (ii) B, (iii) A and B. More generally, the expressions “at least one of the following: ” and “at least one of ” and similar terms (where a list of two or more elements is connected by “and” or “or”) mean at least one element, or at least two or more elements, or at least all elements.

[0225] Within the scope of this disclosure, provision may include any interface configured to provide data. This may include application programming interfaces, human-machine interfaces (such as displays), and / or software module interfaces. Provision may include transmitting or submitting data to the interface, particularly displaying data to a user or having data used by a receiving entity.

[0226] Within the scope of this disclosure, acquisition may include any interface configured to acquire or receive data. This may include application programming interfaces, human-machine interfaces (such as displays), and / or software module interfaces. Acquisition may include transferring or submitting data from the interface, particularly data used by the receiving entity. Any acquisition of data, data structures, datasets, etc., may include receiving such data, data structures, datasets, etc., from a server that provides (e.g., hosts) a database containing such data, data structures, datasets, etc.

[0227] Various units, circuits, entities, nodes, or other computing components may be described as being “configured to” perform one or more tasks. “Configured to” should be described as meaning “having a circuit system that performs one or more tasks during operation.” Units, circuits, entities, nodes, or other computing components may be configured to perform tasks even when the unit / circuit / component is not in operation. Units, circuits, entities, nodes, or other computing components forming a structure corresponding to “configured to” may include hardware circuitry and / or memory storing executable program instructions to perform operations. For convenience in the description, units, circuits, entities, nodes, or other computing components may be described as performing one or more tasks. This description should be interpreted as including the phrase “configured to.” Any statement of “configured to” is not expressly intended to invoke 35 USC. Explanation of 112(f).

[0228] Generally, the methods, apparatuses, systems, computer elements, nodes, or other computing components described herein may include memory, software components, and hardware components. Memory may include volatile memory (such as static or dynamic random access memory) and / or non-volatile memory (such as optical or magnetic disk storage devices, flash memory, programmable read-only memory, etc.). Hardware components may include any combination of the following: combinational logic circuit systems, clock storage devices (such as flip-flops, registers, latches, etc.), finite state machines, memory (such as static random access memory or embedded dynamic random access memory), custom-designed circuit systems, programmable logic arrays, etc.

[0229] In this specification, any connections presented in the described embodiments should be understood in a manner that the components involved are operatively coupled. Therefore, these connections may be direct or indirect, have any number or combination of intermediate elements, and there may be only a functional relationship between the components.

[0230] Furthermore, any methods, processes, and actions described or illustrated herein may be implemented using executable instructions in a general-purpose or special-purpose processor, and these executable instructions are stored on a computer-readable storage medium (e.g., a disk, memory, etc.) for execution by such processor. The reference to 'computer-readable storage medium' should be understood to include special-purpose circuitry, such as signal processing devices, and other devices.

[0231] Any disclosures and embodiments described herein relate to the methods, systems, devices, and computer program elements listed above, and vice versa. Advantageously, the benefits provided by any embodiment and example also apply to all other embodiments and examples, and vice versa.

[0232] All terms and definitions used in this document are to be understood in a broad sense and, unless otherwise stated, have their general meaning.

[0233] It should be understood that all presented embodiments are merely examples, and any feature presented in a particular example embodiment may be used alone with any aspect, or in combination with any feature presented in the same or another particular example embodiment, and / or in combination with any other feature not mentioned. In particular, the example embodiments presented in this specification should also be understood as being disclosed in all possible combinations with each other, provided that such combinations are technically reasonable and the example embodiments are not alternatives to each other. It should be further understood that any feature presented in example embodiments of a particular category (method / apparatus / computer program / system) may also be used in example embodiments of any other category in a corresponding manner. It should also be understood that the presence of a feature in a presented example embodiment does not necessarily mean that the feature forms an essential feature and cannot be omitted or substituted.

Claims

1. A computer-implemented method for generating chemical product production and / or processing data related to the production and / or processing of chemical products, the method comprising: Provide chemical product data related to the production and / or processing of this chemical product. Provide multiple production and / or processing data templates indicating the structure associated with at least a portion and / or the associated elements of the chemical product data. At least one production and / or processing data template is selected based on at least a portion of the chemical product data. Contextualized chemical product data is generated by combining at least a portion of the chemical product data with at least one selected production and / or processing data template. The contextualized chemical product data is provided to a data-driven model to generate production and / or processing data for the chemical product, wherein the data-driven model is parameterized and / or trained to provide the production and / or processing data for the chemical product in response to the provision of the contextualized chemical product data. Provide production and / or processing data for the chemical product.

2. The method of claim 1, further comprising: Receive at least a portion of the second contextualized chemical product data based on the chemical product's production and / or processing data, and The second contextualized chemical product data is provided to the data-driven model to generate production and / or processing data for the second chemical product, and Provide production and / or processing data for the second chemical product.

3. The method as described in claim 1 or 2, wherein, The plurality of production and / or processing data templates include multiple types of production and / or processing data templates, and wherein selecting at least one production and / or processing data template based on at least a portion and / or associated elements of the chemical product data means selecting the at least one production and / or processing data template based on the type of the production and / or processing data template, wherein the type of the production and / or processing data template specifies that the chemical production and / or processing data includes at least one of the following: chemical product manufacturer data, chemical product production instructions, chemical product processor data, chemical product processing instructions, one or more material properties associated with the chemical product, and one or more material properties associated with the product based on the chemical product.

4. The method according to any one of claims 1 to 3, wherein, The data-driven model is a pre-trained data-driven model, wherein the pre-trained data-driven model is trained on general data to provide output data based on the provided input data.

5. The method according to any one of claims 1 to 4, wherein, The data-driven model is a fine-tuned data-driven model, wherein the fine-tuned data-driven model is preferably a pre-trained data-driven model that is further trained in a second training based on historical data input data to provide chemical product production and / or processing data based on provided contextualized chemical product data, wherein the pre-trained data-driven model is trained based on general data to provide output data based on provided input data.

6. The method of any one of claims 1 to 5, further comprising selecting the data-driven model from a plurality of data-driven models based on the chemical product data, and / or providing an indication for selecting the data-driven model from a plurality of data-driven models and selecting the data-driven model from a plurality of data-driven models based on the indication.

7. The method according to any one of claims 1 to 6, wherein, The plurality of production and / or processing data templates are provided via a database, and wherein selecting at least one production and / or processing data template based on at least a portion and / or associated elements of the chemical product data may include providing a query to the database based on the production and / or processing data and / or receiving the at least one production and / or processing data template based on the query.

8. The method according to any one of claims 1 to 7, wherein, The plurality of production and / or processing data templates are provided via a database, and wherein selecting at least one production and / or processing data template based on at least a portion and / or associated elements of the chemical product data includes generating embedded chemical product data and generating one or more embedded production and / or processing data templates, and selecting the production and / or processing data template based on determining that the tensor product between the tensor associated with the embedded chemical product data and the tensor associated with at least one of the one or more production and / or processing data templates is within a predefined numerical range.

9. The method according to any one of claims 1 to 8, wherein, The chemical product data includes one or more elements associated with the chemical product data, and wherein at least one of the plurality of production and / or processing data templates includes one or more elements.

10. The method of claim 9, wherein, The contextualized chemical product data includes two or more elements that are equal to one or more elements associated with the chemical product data and / or one or more elements of at least one production and / or processing data template.

11. The method of claim 10, wherein, The contextualized chemical product data includes a sequence of two or more elements that are equal to one or more elements associated with the chemical product data and / or one or more elements of at least one production and / or processing data template.

12. The method according to any one of claims 8 to 10, wherein, The element refers to at least a part of a word, at least a part of a number, a part of a table, or a combination thereof, and / or wherein the sequence of two or more elements refers to at least a part of a sentence, at least a part of a number sequence, at least a part of a table, or a combination thereof.

13. An apparatus and / or system comprising: processor; as well as A memory storing instructions that, when executed by the processor, configure the device to perform the steps of any one of the methods as claimed in any one of claims 1 to 2, 4 to 10, or 12.

14. The system of claim 13, further comprising a user interface, wherein, The user interface is configured to receive chemical product processing data and / or chemical product production data from the user, and / or is configured to provide the chemical product processing data and / or chemical product production data to the processor, in particular.

15. Use of production and / or processing data related to the production and / or processing of a chemical product, obtained by any one of the methods according to any one of claims 1 to 12, for the control and / or monitoring of the production and / or processing of the chemical product.

Images