Identification of emerging needs

Abstract

A method, in particular computer-implemented method, for obtaining a chemical product with a desired property, the method comprising: - obtaining, in particular receiving one or more chemical product data set(s) related to an application of the chemical product, wherein the one chemical product data set(s) comprise one or more element(s), - selecting one or more chemical product data set(s) according to an occurrence of the one or more element(s) per chemical product data set, - determining one or more desired properties of the chemical product per selected chemical product data set, - providing the one or more desired properties of the chemical product for providing, in particular producing, the chemical product associated with the one or more desired properties.

Description

[0001] 240435

[0002] IDENTIFICATION OF EMERGING NEEDS

[0003] TECHNICAL FIELD

[0004] The disclosure relates to a method, in particular computer-implemented method, for obtaining a chemical product with a desired property, an apparatus for obtaining a chemical product with a desired property, use of a desired property, use of an indication of a chemical product associated with a desired property, a computer element.

[0005] TECHNICAL BACKGROUND

[0006] Chemical products are usually produced on large scales to ensure efficient production and used in a plurality of different applications. Tailoring the chemical production to different application is required.

[0007] SUMMARY

[0008] In an aspect, this disclosure relates to a method, in particular computer-implemented method, for obtaining a chemical product, in particular with a desired property, the method comprising: obtaining, in particular receiving one or more chemical product data set(s) related to an application of the chemical product, wherein the one chemical product data set(s) comprise one or more element(s), selecting one or more chemical product data set(s) according to an occurrence of the one or more element(s) per chemical product data set, determining one or more desired properties of the chemical product per selected chemical product data set, providing the one or more desired properties of the chemical product for providing, in particular producing the chemical product associated with the one or more desired properties.

[0009] In another aspect, it relates to an apparatus for obtaining a chemical product, in particular with a desired property, the apparatus comprising: a processor configured for performing any one of the methods as described herein.

[0010] In another aspect, it relates to use of a, in particular desired, property as determined by any one of the methods as described herein for providing, in particular producing the chemical product associated with the, in particular desired, property. 240435

[0011] 2

[0012] In another aspect, it relates to use of an indication of a chemical product, in particular associated with a desired property, determined by any one of the methods as described herein for providing, in particular producing the chemical product associated with the desired property.

[0013] EMBODIMENTS

[0014] Any disclosure, embodiments and examples described herein relate to the methods, the systems, apparatuses, uses and computer elements lined out above and below. Advantageously, the benefits provided by any of the embodiments and examples equally apply to all other embodiments and examples.

[0015] In the following, terminology as used herein and / or the technical field of the present disclosure will be outlined by ways of definitions and / or examples. Where examples are given, it is to be understood that the present disclosure is not limited to said examples.

[0016] Processors of chemical products may require chemical expertise to find chemical products for their desired application. Especially first requests from processors of chemical products are focused on the application and requirements of the processing of the chemical product. This needs chemical expertise to translate application requirements to properties the chemical products should have. Furthermore, chemical industry operates large- scale productions in production networks to produce chemical products efficiently. Chemical processes can require high energy input and thus, a high throughput is required to lower the energy cost per unit of chemical product. Other resources such as space, cooling agent, solvents or the like can be saved accordingly by scaling chemical production processes. An example of a high resource demanding process can be steam cracking. Thus, chemical producers need to bundle a variety of processor requests to allow for efficient large-scale production. As a consequence, the chemical industry is facing the challenge of serving as many processors of chemical products produced from complex production networks with high throughput.

[0017] The above-described challenge can be solved by selecting chemical product data sets according to an occurrence of elements such as words or symbols within the chemical product data sets for determining desired properties of chemical products. Using the occurrence of elements as a measure for a relevance of chemical product data sets allows to process a high number of chemical product data sets even on a real-time basis once a new chemical product data set is received. Thereby, obtaining the chemical product with the desired property can be scaled to the large number of available chemical product data sets, e.g. in literature, social media, in customer requests or the like. Furthermore, selecting the chemical product data sets for determining the desired properties according to the occurrence provides a filter to select the most relevant, i.e. the mostly desired and applied or processed chemical products to be produced by the chemical production system. As a consequence chemical production is 240435

[0018] 3 tailored to the needs for processors of chemical products and products becoming waste can be circumvented. In addition, producing chemical products on large scales is highly efficient in comparison to small or medium batch sizes. The reason for the high efficiency gains with upscaling of chemical processes is the excessive resource saving. This may include saving energy, cooling agents, solvents and the like. In the end, this contributes to reducing the CO2 footprint and waste production associated with chemical production. More than that, resource investments for establishing production facilities and developing production processes for specialized low throughput chemical products come with the danger of being related to a current trend and becoming irrelevant in the nearer future. Thus, avoiding resource invest to establish specialized chemical production processes and identify chemical products suited for large-scale production saves further resources with respect to developing chemical processes. Ultimately, by selecting chemical product data sets according to an occurrence of elements such as words or symbols within the chemical product data sets for determining desired properties of chemical products the efficiency of chemical products can be increased.

[0019] These and other objects, which become apparent upon reading the following description, are solved by the subject matters of the independent claims. The dependent claims refer to embodiments of the disclosure.

[0020] In an embodiment, chemical product data may be indicative of a production and / or processing of to the chemical product. The chemical product data may comprise chemical product producer data, chemical product production instructions, potential chemical product processing data and / or potential chemical product production data, one or more material properties associated with the 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 the chemical product, one or more input products the product based on the chemical product for producing the product based on the chemical product, one or more chemical compounds the chemical product is based on or a combination thereof. A product based on the chemical product may refer to a product resulting from processing the chemical product.

[0021] Chemical product producer data may be indicative of the producer with respect to the chemical product and / or of the production facilities associated with the production of the chemical product. For example, the chemical product producer data may be indicative of one or more producers with respect to the chemical product, quantitative production data, qualitative production data or a combination thereof. Qualitative production data may be indicative of the chemical structure associated with the chemical product to be produced. For example, the qualitative production data may be indicative of a name of the chemical product, the quality of the chemical product, the 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 the chemical product is based on, or a combination thereof. Quantitative production data may be indicative of a quantity of the chemical product. 240435

[0022] 4

[0023] Potential chemical product production data may be indicative of an expected production with respect of the chemical product based on historical production and / or production associated with alternative products to the chemical product. For example, the potential chemical product production data may be indicative of a historical production associated with the chemical product, alternatives to the chemical product, a product to be produced based on the chemical product, a field of application with respect to the chemical product or a combination thereof. In an embodiment, chemical product production data may be received in an audio format and / or a text format.

[0024] Chemical product processor data may be indicative of the chemical product processor with respect to the chemical product and / or of the processing facilities associated with the processing of the chemical product. For example, the chemical product processor data may be indicative of one or more chemical product processors with respect to the chemical product, quantitative processing data, qualitative processing data, a field of application with respect to the chemical product, the product to be produced based on the chemical product, a field of application of the product to be produced based on the chemical product or a combination thereof. Qualitative processing data may comprise a 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 being comprised in the chemical product or a combination thereof. Quantitative processing data may be indicative of quantities associated with one or more components associated with the chemical product. Chemical product processing instructions may be indicative of the processing instructions associated with the processing of the chemical product. For example, the chemical product processing instructions may be related to processing conditions such as temperature, pressure, reaction rate, catalyst or the like.

[0025] Potential chemical product processing data may be indicative of an expected processing with respect to the chemical product based on historical processing and / or processing associated with alternative products to the chemical product. For example, the potential chemical product processing data may be indicative of a historical processing associated with the chemical product, alternatives to the chemical product or a combination thereof.

[0026] In an embodiment, chemical product data may be received and / or provided in an audio format and / or a numerical format and / or text format and / or digital format. In an embodiment, the chemical product data may comprise one or more elements, preferably production and / or processing parameter values.

[0027] In an embodiment, providing input data to the data-driven model may comprise mapping the input data to a numerical representation of the input data. The numerical representation of the input data may comprise a tensor associated with the input data and / or obtained from the input data. In particular, the numerical representation of 240435

[0028] 5 the input data may be indicative of two or more elements of the input data and a relation between the two or more elements of the input data. Preferably, providing input data to the data-driven model may comprise at least one of identifying two or more elements of the input data, mapping the two or more elements of the input data to a numerical representation of the two or more elements, mapping the numerical representation of the two or more elements to a numerical representation of a predefined size related to the numerical representation of the two or more elements, mapping the numerical representation of the predefined size related to the numerical representation of the two or more elements to a numerical representation of the two or more elements and a relation between the two or more elements or a combination thereof.

[0029] In an embodiment, processing the input data and / or generating the output data from the input data may comprise processing the numerical representation of the input data, in particular the numerical representation of the two or more elements and the relation between the two or more elements. Processing the numerical representation of the two or more elements and the relation between the two or more elements may comprise mapping the numerical representation of the two or more elements, and optionally the relation between the two or more elements to a numerical representation of the output data. The numerical representation of the output data may be mapped to the output data, in particular based on a relation between the numerical representation of data and the data. In particular, the data may be of a data type according to the input data. Hence, the output data may be of the data type according to the input data, e.g. of the same data type as the input data and / or of the data type specified by the input data. Preferably, processing the numerical representation of the two or more elements and the relation between the two or more elements may comprise at least one of generating two or more numerical representations of the two or more elements and the relation between the two or more elements from the numerical representation of the two or more elements and the relation between the two or more elements, modifying the two or more numerical representation of the two or more elements and the relation between the two or more elements by applying a filter to the two or more numerical representations of the two or more elements and the relation between the two or more elements, concatenating the two or more numerical representations of the two or more elements and the relation between the two or more elements, mapping the concatenated numerical representation of the two or more elements and the relation between the two or more elements to a numerical representation of the output data or a combination thereof.

[0030] In an embodiment, the one or more data-driven model may be pretrained data-driven model(s). The pretrained data-driven model(s) may be parametrized and / or trained based on data with a plurality of contexts and / or unstructured data, in particular text data and optionally numerical data such as tabular data or image data. The pretrained data-driven model(s) may be configured to perform a plurality of task and / to process data of a plurality of contexts. The pretrained data-driven model(s) may be configured to perform the task according to the provided task instruction. Hence, the pretrained data-driven model may be configured to be provided with a plurality of 240435

[0031] 6 different task instructions and / or provide a plurality of different types of output data upon receiving different task instructions.

[0032] In an embodiment, the one or more data-driven model(s) may be finetuned data-driven model(s). The finetuned data-driven model(s) may be obtained by training pretrained data-driven model(s) configured to perform a plurality of tasks according to a plurality of task instructions. The finetuned data-driven model(s) may trained additionally on a training data set comprising a plurality of task instructions of one type and corresponding output data. The fine tuned data-driven model may be trained additionally to provide output data of a predefined type according to the training data set. The finetuned data-driven model may be configured to be provided with a plurality of different task instructions and / or provide a plurality of different types of output data upon receiving different types of task instructions. Further, the finetuned data-driven model may be configured for providing one type of output data upon receiving one type of task instruction with a higher accuracy than providing other types of output data upon receiving other types of task instructions.

[0033] In an embodiment, the chemical product data set may be indicative of a desired property of the chemical product associated with the application of the chemical product. The chemical product data set may be associated with an application of the chemical product. The chemical product data set may be related to a desired application of a chemical product. The chemical product data set may characterize and / or indicate the chemical product by the desired application of the chemical product. The chemical product data set may be indicative of one or more processing steps for processing the chemical product. In an example, the chemical product data set may quantify an amount of the chemical product to be processed and / or may characterize one or more processing steps for processing the chemical product qualitatively. Additionally or alternatively, the request may be indicative of a desired property for processing the chemical product. The chemical product data sets may comprise natural language and / or may be unstructured. The chemical product data sets may comprise one or more segment(s). One segment may comprise one or more paragraph(s) of text. At least one of the one or more segments of a chemical product data set may be associated with at least one chemical product and one or more desired properties of the at least one chemical product. The chemical product data sets may be associated with points in time. Hence, the chemical product data sets may be indicative of desired properties of chemical products at different points in time. The plurality of chemical product data sets may be obtained from a database comprising the plurality of chemical product data sets. The chemical product data set may comprise text data.

[0034] In an embodiment, selecting at least one chemical product data set according to an occurrence of the one or more element(s) per chemical product data set may comprise determining an occurrence of the one or more element(s) per chemical product data set, and selecting at least one chemical product data set according to the determined occurrence. 240435

[0035] 7

[0036] In an embodiment, an indication of the chemical product may be suitable for identifying the chemical product. The indication of the chemical product may comprise at least one characteristic associated with the chemical product. The at least one characteristic associated with the chemical product may distinguish the chemical product from other chemical products. The indication of the chemical product may be related to, in particular include, a type of a chemical product, a group of chemical products, a physical entity associated with the chemical product, a name of the chemical product, one or more educt(s) for obtaining the chemical product, a CAS number or a combination thereof. The target chemical product may comprise chemical products associated with at least one target type of chemical product, a group of target chemical products, a physical entity associated with the target chemical product or a combination thereof.

[0037] In an embodiment, the properties of the chemical product may include physical properties such as aggregate state, viscosity, boiling point, density, scratch resistance or the like. Additionally or alternatively, the properties may include chemical properties such as solubility, reactivity, class of chemical compound or the like. Additionally or alternatively, the properties may include biological properties such as enzymatic activity, toxicity or the like. Additionally or alternatively, the property may be related to a class of chemical product and / or a target field of application and / or a target processing associated with the chemical product.

[0038] In an embodiment, any one of the methods may further comprise obtaining a selection of at least one of the one or more element(s), in particular via a user interface. In particular, the selection may be obtained in response to determining the occurrence associated with the one or more element(s) and / or according to the determined occurrence associated with the one or more element(s). The indication of the selection may be associated with at least of the one or more element(s) and / or the indication of the selection may be suitable for identifying at least of the one or more element(s). For example, the indication of the selection may comprise a text prompt, obtained eg via a user interface. In the example, an identification task instruction associated with the text prompt may be provided to the generative data-driven model for identifying the at least one of the one or more element(s). The selection may be obtained in response to providing the one or more element(s) and the corresponding occurrence, in particular via the user interface. Additionally or alternatively, obtaining a selection of the at least one element may comprise selecting a predefined number of elements, preferably, from the one or more element(s). Preferably, the one or more selected elements may be associated with the occurrence closer to a target occurrence than the unselected element(s). Additionally or alternatively, obtaining a selection of the at least one element may comprise selecting one or more element(s) associated with the occurrence corresponding to the a target occurrence. By selecting at least one element, non-meaningful elements and / or less frequently used elements can be excluded from determining the desired properties. Non-meaningful elements can be for example words or symbols that do not add to the context or the meaning of the data. For example, “the” or may be non- 240435

[0039] 8 meaningful elements. Further, to allow for an efficient chemical production, chemical products with a high applicability are preferred. Hence, by selecting at least one element, chemical products associated with a less efficient production due to lower expected throughput can be excluded from obtaining chemical products.

[0040] In an embodiment, any one of the methods may further comprise determining the occurrence of the one or more element(s) per chemical product data set. In an embodiment, the occurrence of the one or more element(s) may quantify a number of occurrences with respect to the one or more element(s). Determining occurrences may include counting the number of occurrences with respect to the one or more element(s). In an embodiment, element may refer to at least a part of a word, at least a part of a number, a part of a table or a combination thereof. Additionally or alternatively, the sequence of two or more elements may refer 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. A part of a table may refer to a row number, a column number and / or a table entry.

[0041] In an embodiment, the occurrence of the one or more element(s) may comprise the occurrence of one or more combination(s) of two or more elements. Optionally, the combination of the two or more elements may be indicative of an occurrence of one of the elements in relation to the at least one other element. The combination of the two or more elements may refer to a relative occurrence of two or more elements. Preferably, a relative occurrence of the two or more elements associated with each other. The occurrence of the elements can be used to determine the relevant chemical product data set, i.e. the chemical product data set associated with desired properties of chemical products that may be suitable for large-scale production. Typically, a combination of words or symbols allows for an improved summary of, i.e. more meaningful insight into, the content of the chemical product data set. Selecting relevant chemical product data sets based on more meaningful insights allows for an improved selection. Thus, determining relevant and highly desired properties of chemical products is improved. Ultimately, this materializes in an efficient processing of chemical products with desired properties that allow for tailored processing by chemical product processors.

[0042] In an embodiment, two or more chemical product data sets may be obtained, in particular received. The two or more chemical product data sets may be associated with two or more points in time. Any one of the methods may further comprise determining a change of the occurrence related to a difference of the occurrence of the one or more element(s) between the at least two different points in time. The one or more chemical product data set(s) may be selected according to the change of the occurrence. In particular, the one or more element(s) may be selected according to the change of the occurrence. Any one of the methods may further comprise identifying at least one chemical product data set comprising the at least one selected element. The desired properties of the chemical product may be determined based on and / or from the at least one identified chemical product data set. Determining the change of the occurrence allows to provide a time-resolved insight into the occurrences. The 240435

[0043] 9 change of the occurrence can be seen as a measure for a relevance associated with a certain desired property of the chemical product over time. Typically, production of chemical products needs time to be adapted.

[0044] Furthermore, desired properties of chemical products may change over time due to changing desires of end product consumers. Hence, desired long-term planning can be provided by determining the change of the occurrence to obtain chemical products efficiently.

[0045] In an embodiment, two or more chemical product data sets may be obtained, in particular received. The two or more chemical product data sets may be associated with two or more points in time. Any one of the methods may further comprise obtaining, in particular receiving at least one future point in time following the at least two points in time associated with the chemical product data sets, in particular via a user interface. The occurrence and / or the change of occurrence may be further determined associated with the at least one future point in time. The occurrence and / or the change of the occurrence can be determined associated with the at least one future point in time by extrapolating the occurrence and / or the change of the occurrence from the two or more points in time to the future point in time. For this purpose, parameters associated with a predefined mathematical function may be determined in order to achieve a distance between the mathematical function associated with the determined parameters within a target distance range. The desired property may be determined at the future point in time and / or may be provided to prepare the production of the chemical product with the desired property for the future point in time. The future point in time may be a point in time following the two or more points in time associated with the two or more chemical product data sets. Typically, production of chemical products needs time to be adapted. Furthermore, desired properties of chemical products may change over time due to changing desires of end product consumers. Hence, desired long-term planning can be provided by extrapolating the occurrence and / or the change of the occurrence to a future point in time.

[0046] In an embodiment, the one or more chemical product data set(s) may be related to two or more application(s) of two or more chemical product(s), and further comprising obtaining, in particular receiving, a request for obtaining the chemical product, wherein the request may be associated with at least one of the two or more appl ication(s) of the chemical product. Any one of the methods may further comprise selecting one or more of the two or more chemical product data set(s) according to the request. Selecting one or more of the two or more chemical product data set(s) according to the request may comprise matching at least a part of the request with the one or more chemical product data set(s). Matching at least the part of the request with the one or more chemical product data set(s) may comprise matching elements comprised by the one or more chemical product data set(s) and at least the part of the request. Matching elements may comprise performing a keyword search and selecting one or more chemical product data set(s) with the highest matching score. The matching score may be indicative of a number of matching elements. Optionally, the matching score may be obtained by weighting matching of different elements differently. Additionally or alternatively, matching the elements may comprise obtaining a numerical 240435

[0047] 10 representation per chemical product data set and a numerical representation of the request. Obtaining a numerical representation may include providing the one or more chemical product data set(s) and / or the request to one or more embedding layer(s). Further, matching the elements may comprise determining a distance between the numerical representations of the one or more chemical product data set(s) and the numerical representation of the request. The request may comprise text data. By doing so, the chemical product data sets relevant to the query can be identified from a plurality of available chemical product data sets. As a consequence, less computational resources are used for obtaining the chemical product with the desired properties.

[0048] In an embodiment, the one or more chemical product data set(s) may be associated with two or more two or more points in time. Any one of the methods may further comprise obtaining, in particular receiving, an indication of one or more point(s) in time, and selecting one or more of the two or more chemical product data set(s) according to the indication of the one or more point(s) in time. Selecting one or more of the two or more chemical product data set(s) according to the indication of the one or more point(s) in time may comprise matching at least a part of the indication of the one or more point(s) in time with the one or more chemical product data set(s). Matching at least the part of the indication of the one or more point(s) in time with the one or more chemical product data set(s) may comprise matching elements comprised by the one or more chemical product data set(s) and at least the part of the indication of the one or more point(s) in time. Matching elements may comprise performing a keyword search and selecting one or more chemical product data set(s) with the highest matching score. The matching score may be indicative of a number of matching elements. Optionally, the matching score may be obtained by weighting matching of different elements differently. Additionally or alternatively, matching the elements may comprise obtaining a numerical representation per chemical product data set and a numerical representation of the indication of the one or more point(s) in time. Obtaining a numerical representation may include providing the one or more chemical product data set(s) and / or the indication of the one or more point(s) in time to one or more embedding layer(s). Further, matching the elements may comprise determining a distance between the numerical representations of the one or more chemical product data set(s) and the numerical representation of the indication of the one or more point(s) in time. By doing so, the chemical product data sets relevant at a predefined point in time can be identified from a plurality of available chemical product data sets. As a consequence, less computational resources are used for obtaining the chemical product with the desired properties. Furthermore, chemical production can be improved at the predefined point in time by tailoring production of chemical products to desired applications for chemical products.

[0049] In an embodiment, the one or more chemical product data set(s) may be associated with two or more two or more points in time. Any one of the methods may further comprise obtaining, in particular receiving, an indication of one or more location(s) and further selecting the one or more chemical product data set(s) according to the indication of one or more location(s). Selecting one or more of the two or more chemical product data set(s) according to the 240435

[0050] 11 indication of one or more location(s) may comprise matching at least a part of the indication of one or more location(s) with the one or more chemical product data set(s). Matching at least the part of the indication of one or more location(s) with the one or more chemical product data set(s) may comprise matching elements comprised by the one or more chemical product data set(s) and at least the part of the indication of one or more location(s). Matching elements may comprise performing a keyword search and selecting one or more chemical product data set(s) with the highest matching score. The matching score may be indicative of a number of matching elements. Optionally, the matching score may be obtained by weighting matching of different elements differently.

[0051] Additionally or alternatively, matching the elements may comprise obtaining a numerical representation per chemical product data set and a numerical representation of the indication of one or more location(s). Obtaining a numerical representation may include providing the one or more chemical product data set(s) and / or the indication of one or more location(s) to one or more embedding layer(s). Further, matching the elements may comprise determining a distance between the numerical representations of the one or more chemical product data set(s) and the numerical representation of the indication of one or more location(s). By doing so, the chemical product data sets relevant at a predefined location can be identified from a plurality of available chemical product data sets. Chemical products with different properties may be desired at different locations, e.g. due to local specifications for processing chemical products. As a consequence, less computational resources are used for obtaining the chemical product with the desired properties. Furthermore, chemical production can be improved at the predefined location by tailoring production of chemical products to desired applications for chemical products.

[0052] In an embodiment, any one of the methods may further comprise obtaining an indication on a quantity associated the application of the chemical product, in particular together with the request and / or extracted from the one or more chemical product data set(s). The one or more chemical product data set(s) may be selected further according the quantity associated with the chemical product. The indication of the quantity associated with the application may comprise the quantity associated with the application. The quantity associated with the application of the chemical product may refer to a desired quantity of the chemical product and / or a quantity of one or more material(s) desired to be processed with the chemical product and / or a quantity of one or more material(s) desired to be obtained by processing the chemical product. Selecting the one or more chemical product data set(s) according to the quantity associated with the application and the occurrence of the one or more element(s) may comprise. Chemical processes can require high energy input and thus, a high throughput is required to lower the energy cost per unit of chemical product. Other resources such as space, cooling agent, solvents or the like can be saved accordingly by scaling chemical production processes. An example of a high resource demanding process can be steam cracking. Thus, taking the quantity associated with the application into account for obtaining the chemical product with the desired property allows to focus on high-quantity production processes. 240435

[0053] 12

[0054] In an embodiment, any one of the methods may further comprise obtaining a selection of at least a part of the one or more chemical product data set(s) and selecting at least the part of the one or more chemical product data set(s) according to the obtained selection. The obtained selection may specify a part of at least one of the chemical product data sets. For example, at least one of the plurality of chemical product data sets may comprise two or more parts. The parts may be related to different applications and / or at least one of the two or more parts may comprise more details related to the application than the at least other part of the two or more parts. By selecting at least a part of the chemical product data sets, the relevant and / or more detailed part of the data set can be selected for obtaining the chemical product with the one or more desired properties.

[0055] In an embodiment, the one or more chemical product data set(s) may comprise two or more element(s), in particular per chemical product data set. Any one of the methods may further comprise eliminating at least one element from the one or more chemical product data set(s), in particular per chemical product data set. NLP models may get distracted from less meaningful and context-determining words or symbols such as “the”, or the like. Hence, the elements independent from the context of the chemical product data set may be removed. Consequently, accuracy of results from NLP models operating on the processed data can be improved. Ultimately, this improves data extraction from chemical product data sets.

[0056] In an embodiment, determining the one or more desired properties may comprise providing the one or more selected chemical product data set(s), in particular via a user interface and receiving the desired property associated with the one or more selected chemical product data set(s) in response to providing the one or more selected chemical product data set(s), in particular via a user interface. In an example, chemical expertise from a human expert may be required to extract the one or more desired properties from the chemical property data set. By providing the chemical product data set, input from the human expert can be received. Often, said human experts can be considered as benchmark for obtaining chemical products with desired properties. Hence, providing the product to the human experts, provides a very accurate determination of the desired properties. Ultimately, this increases reliability of obtaining chemical products with desired properties on large scales.

[0057] In an embodiment, determining the one or more desired properties may comprise providing a task instruction for determining the desired property associated with the one or more selected chemical product data set(s) to a generative data-driven model. The task instruction may be related to, in particular may include, the one or more selected chemical product data set(s). The generative data-driven model may be configured to follow task instructions provided. The chemical product data sets may comprise text data. By doing so, complex content of the chemical product data can be obtained by the generative data-driven model, e.g. a natural language processing (NLP) model. By doing so, no human input is required for interpreting such complex data. In turn, this decreases time at scale. Followingly, using generative data-driven model for determining the desired property 240435

[0058] 13 allows to operate a higher number of chemical product data sets in a small time frame. As a consequence, real time adaptations of desired properties upon receiving further chemical product data sets or locally and / or timely resolved determination of desired properties are enabled. Ultimately, this contributes to tailoring the chemical production more efficiently to the processing and / or the application of the chemical products.

[0059] In an embodiment, determining the one or more desired properties may comprise providing the one or more selected chemical product data set(s) to a classification data-driven model for classifying at least one part of the one or more chemical product data set(s) related to the desired property. The classification data-driven model may be configured to classify whether the parts of chemical product data sets provided to the classification data-driven model are related to the desired property. The at least one part of the one or more chemical product data set(s) may comprise and / or may be the desired property. The classification data-driven model may comprise a representation generating engine configured to generate a numerical representation of data provided to the representation generating engine. The numerical representation generated by the representation generating engine may be indicative of the relation between two or more elements of the data provided to representation generating engine. An example of a representation generating engine may be the pretrained text representation model as described in the context of the Figures. The representation generating engine may be already available and may be trained based on a large amount of datapoints from different contexts. Thus, it is advantageous to use the available representation generating engine. Further, the classification model may comprise a custom model. The custom model may be configured to receive the numerical representation of the data from the representation generating engine and provide an indication of whether the numerical representation provided to the custom model may be related to the desired property. An example of the custom model may be the custom model trained on several hundred manual annotations as described in the context of the Figures. By using the classification data-driven model for determining the desired property, the exact language of the chemical product data set is reproduced when determining the desired properties. This avoids hallucination and allows an error-free, accurate and fast determination of the desired properties. The desired properties determined may be provided for producing the chemical product associated with the desired properties.

[0060] In an embodiment, any one of the methods may further comprise obtaining an indication of a chemical product associated with the desired property by providing a request for receiving an indication of the chemical product associated with the desired property to a database. The request for obtaining the indication of the chemical product may be related to, in particular may comprise the desired property. The database may comprise a plurality of indications of chemical products and corresponding properties. The database may be configured to provide the indication of the chemical product associated with the desired property in response to receiving the request for obtaining the indication of the chemical product. Any one of the methods may further comprise providing the indication of the chemical product for providing, in particular producing the chemical product associated with the desired property.

[0061] In an embodiment, an element may refer to one or more datapoint(s) comprises by the data and / or data set(s). Element may be a part of the chemical product data set(s).

[0062] In an embodiment, two or more desired properties, in particular different properties may be determined from the selected chemical product data set(s). Any one of the methods described herein may further comprise obtaining, in particular receiving, an indication of a selection of at least one of the desired properties, e.g. via a user interface, and selecting at least one of the desired properties according to the obtained, in particular received, indication of the selection. The indication of the selection may be suitable for identifying at least one of the desired properties. The indication of the selection may comprise a text prompt and / or may be associated with at least one of the desired properties.

[0063] In an embodiment, two or more desired properties may be determined from the selected chemical product data set(s). Any one of the methods described herein may further comprise obtaining, in particular receiving, an indication of a group of desired properties comprising at least two of the determined desired properties. Providing the one or more desired properties of the chemical product may comprise providing the group of desired properties, e.g. via a user interface.

[0064] In an embodiment, two or more chemical product data sets may be obtained, in particular received. The two or more chemical product data sets may be associated with two or more points in time. At least two of the chemical product data sets may be selected. At least one chemical product data set may be selected per point in time. Any one of the methods may further comprise further comprising obtaining an indication of a chemical product associated with the desired property per point in time by providing a request for receiving an indication of the chemical product associated with the desired property to a database. The request for obtaining the indication of the chemical product may be related to, in particular may comprise the desired property. The database may comprise a plurality of indications of chemical products and corresponding properties. The database may be configured to provide the indication of the chemical product associated with the desired property in response to receiving the request for obtaining the indication of the chemical product. Any one of the methods may further comprise providing the indication of the chemical products for providing, in particular producing the chemical product associated with the desired property. 240435

[0065] 15

[0066] BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS

[0067] In the following, the present disclosure is further described with reference to the enclosed figures. The same reference numbers in the drawings and this disclosure are intended to refer to the same or like elements, components, and / or parts.

[0068] FIG. 1 illustrates an aspect of the subject matter in accordance with one embodiment.

[0069] FIG. 2 illustrates an embodiment of matching chemical product data sets related to an application of one or more chemical product(s) with at least one chemical product having one or more desired properties.

[0070] FIG. 3 illustrates an embodiment of matching requests with chemical products having desired properties.

[0071] FIG. 4 illustrates embodiment of a method for obtaining a chemical product with a desired property.

[0072] FIG. 5 illustrates an embodiment of obtaining a database comprising one or more chemical product data set(s) related to an application associated one or more desired properties of one or more chemical product(s).

[0073] FIG. 6 illustrates an embodiment of determining an occurrence of one or more element(s) within one chemical product data set.

[0074] FIG. 7 illustrates an embodiment of occurrence s at different points in time.

[0075] FIG. 8 illustrates an embodiment of a classification data-driven model configured to classify whether the elements comprised by the at least one identified chemical product data set correspond to the desired property.

[0076] FIG. 9 illustrates an embodiment of a computing apparatus for obtaining a chemical product with a desired property.

[0077] FIG. 10 illustrates an embodiment of a transformer architecture. 240435

[0078] 16

[0079] DETAILED DESCRIPTION

[0080] The following embodiments are mere examples for implementing the method, the system or application device disclosed herein and shall not be considered limiting.

[0081] FIG. 1 illustrates an embodiment of producing chemical products 110 by one or more chemical production facilities 102.

[0082] Chemical products 110 may be produced by the one or more chemical production facilities 102. The chemical products 110 produced by the one or more chemical production facilities 102 may be associated with one or more properties 106. The chemical products 110 may be processed by one or more product processing facilities 108. For processing the chemical products operating system 112 by the one or more product processing facilities 108 the chemical products 1 10 may be required to have to one or more properties 106. The one or more properties 106 may be necessary for producing end products with a target quality. Said end products may be versatile and may range from parts of cars to packaging. End products with a lower quality than the target quality may become waste immediately as such end products may not be suited for the intended application at all or may degrade faster than the end products with the target quality. Hence, ensuring sufficient quality of the chemical products produced by the one or more chemical production facilities 102 is important to reduce resource waste. Thereby, the waste of chemical production is further lowered, i.e. the efficiency of producing chemical products 1 10 is increased, while the production of the chemical products by the one or more chemical production facilities 102 may be tailored to the properties of the end products.

[0083] In chemical industry, said chemical products 110 are produced on a ton-scale and one chemical product can serve for different purposes. Furthermore, chemical products 1 10 are obtained from chains of several chemical production step typically including a plurality of chemical reactions. For example, the production system may be a Verbund system. In such a system, the production of chemical products 1 10 may start with converting naphtha to smaller molecules and may be continued with further reaction steps to end at more complex molecular structures and / or chemical compositions. At all different stages of such a complex production system, chemical products are produced for different application purposes. Per application purpose, one or more processing entities e.g. at different sizes and / or at different locations may be available. Therefore, an efficient matching between chemical products 110 produced by large-scale and highly complex production systems and a variety of properties of chemical products required by chemical product processing entities at scale is required. This matching can be achieved by applying the subject-matter of this disclosure.

[0084] FIG. 2 illustrates an embodiment of matching chemical product data sets related to an application of one or more chemical product(s) with at least one chemical product having one or more desired properties. 240435

[0085] 17

[0086] A plurality of chemical products with a plurality of distinct properties may be produced, in particular by a plurality of chemical production facilities. A plurality of applications may be associated with the plurality of chemical products. The target and / or desired application of a chemical product depends on the properties of the chemical product. The properties of the chemical product may include physical properties such as aggregate state, viscosity, boiling point, density, scratch resistance or the like. Additionally or alternatively, the properties may include chemical properties such as solubility, reactivity, class of chemical compound or the like. Additionally or alternatively, the properties may include biological properties such as enzymatic activity, toxicity or the like. Additionally or alternatively, the property may be related to a class of chemical product and / or a target field of application and / or a target processing associated with the chemical product.

[0087] A plurality of desired applications of the chemical product may be associated with a plurality of different processing steps. Different processing steps may be associated with at least partially same or different chemical products. To bundle and match a high number of different intended applications with the produced chemical products, scalable solutions are required that can operate on real time data.

[0088] Processors of chemical products may require chemical expertise to find chemical products for their desired application. Especially first requests from processors of chemical products are focused on the application and requirements of the processing of the chemical product. This needs chemical expertise to translate application requirements to properties the chemical products should have. Furthermore, chemical industry operates large- scale productions in production networks to produce chemical products efficiently. Chemical processes can require high energy input and thus, a high throughput is required to lower the energy cost per unit of chemical product. Other resources such as space, cooling agent, solvents or the like can be saved accordingly by scaling chemical production processes. An example of a high resource demanding process can be steam cracking. Thus, chemical producers need to bundle a variety of processor requests to allow for efficient large-scale production. As a consequence, the chemical industry is facing the challenge of serving as many processors of chemical products produced from complex production networks with high throughput. The herein described subject-matter may be intended to overcome the above-described challenge.

[0089] In an embodiment, chemical product data sets related to an application associated one or more desired properties of one or more chemical product(s) may be available, e.g. in a database and / or provided via an interface. In an example, a plurality of chemical product data sets relating to applications of a plurality of chemical products with a plurality of properties may be available and a selection of chemical product data sets related to an application requiring one or more desired properties may be obtained. By performing 408 to 418, the desired properties relevant to one or more application(s) may be determined. By rating the relevance, chemical production can be 240435

[0090] 18 focused on producing the chemical products with the desired properties efficiently with a lower resource invest than small-scale production.

[0091] FIG. 3 illustrates an embodiment of matching requests with chemical products having desired properties.

[0092] In an embodiment, request from one or more processor(s) can be received and can be clustered according to the desired properties associated with the request by performing at least parts of 402 to 418. By doing so, the desired properties relevant to one or more processor(s) may be determined. By rating the relevance, chemical production can be focused on producing the chemical products with the desired properties efficiently with a lower resource invest than small-scale production.

[0093] FIG. 4 illustrates embodiment of a method for obtaining a chemical product with a desired property.

[0094] A plurality of chemical product data sets associated with at least two different points in time may be obtained, in particular provided 404. The chemical product data sets may be indicative of a plurality of chemical products and desired properties of the chemical product. The chemical product data sets may comprise natural language and / or may be unstructured. The chemical product data sets may comprise one or more segment(s). One segment may comprise one or more paragraph(s) of text. At least one of the one or more segments of a chemical product data set may be associated with at least one chemical product and one or more desired properties of the at least one chemical product. The chemical product data sets may be associated with points in time. Hence, the chemical product data sets may be indicative of desired properties of chemical products at different points in time. The plurality of chemical product data sets may be obtained from a database comprising the plurality of chemical product data sets. The chemical product data sets associated with the database may be obtained by processing as described in the context of FIG. 4.

[0095] A request for obtaining a chemical product may be obtained, in particular received 402. The request may be provided via a user interface and / or initialized and / or triggered by a user. The request may be associated with an application of the chemical product. The request may be related to a desired application of a chemical product. The request may characterize the chemical product by the desired application of the chemical product. The request may be indicative of one or more processing steps for processing the chemical product. In an example, the request may quantify an amount of the chemical product to be processed and / or may characterize one or more processing steps for processing the chemical product qualitatively. Additionally or alternatively, the request may be indicative of a desired property for processing the chemical product. 240435

[0096] 19

[0097] At least one chemical product data set per point in time may be selected by matching at least a part of the request with the plurality of chemical product data sets 406. Matching at least the part of the request with the plurality of chemical product data sets may comprise matching elements comprised by the chemical product data sets and at least the part of the request. One element may refer to one word, an ensemble of words, a number, an ensemble of numbers or the like. Matching elements may comprise performing a keyword search and selecting one or more chemical product data set(s) with the highest matching score. The matching score may be indicative of a number of matching elements. Optionally, the matching score may be obtained by weighting matching of different elements differently. Additionally or alternatively, matching the elements may comprise performing an embedding search. This may include obtaining a numerical representation per chemical product data set and a numerical representation of the request. Obtaining a numerical representation may include providing the chemical product data sets and / or the request to one or more embedding layer(s) as described in the context of FIG. 10. Further, performing the embedding search may comprise determining a distance between the numerical representations of the chemical product data sets and the numerical representation of the request. By doing so, the chemical product data sets relevant to the query can be identified and less computational resources are used for obtaining the chemical product with the desired properties.

[0098] An occurrence of one or more element(s) within the at least one selected chemical product data set per point in time may be determined 408. In an embodiment, the one or more elements may comprise a combination of two or more elements. Optionally, the combination of the two or more elements may be indicative of an occurrence of one of the elements in relation to the at least one other element. For example, the combination of the two or more elements may be a topic. The topic may be obtained via topic modelling. For further details, see FIG. 6. Hence, the combination of the two or more elements may refer to a relative occurrence of two or more words or terms. Preferably, a relative occurrence of the two or more words or terms associated with each other. In an embodiment, an occurrence of the one or more element(s) may quantify a number of occurrences with respect to the one or more element(s). Determining occurrences may include counting the number of occurrences with respect to the one or more element(s).

[0099] A selection of the at least one element may be obtained 410. The selection may be received via a user interface. For example, the one or more element(s) comprised by the at least one selected chemical product data set and the corresponding occurrence may be provided via a user interface. The selection may be received via the user interface, in particular in response to providing the one or more element(s) and the corresponding occurrence. Additionally or alternatively, a predefined number of elements may be selected according to the determined occurrence. For example, the predefined number of elements with the highest occurrences determined may be selected. 240435

[0100] 20

[0101] A change of the occurrence related to a difference of the occurrence of the selected at least one element between the at least two different points in time may be determined 412. The change of the occurrence may be related to a quantitative and / or a qualitative difference of the occurrence of the selected at least one element between the at least two different points in time. A qualitative difference of the occurrence may be obtained by subtracting the occurrence s at a point in time preceding an earlier point in time by the occurrences at the earlier point in time. A negative difference of the occurrences may be indicative of a negative change of the occurrence, analogous for a positive difference. A difference equal to 0 may be indicative of a steady occurrence. The value of the difference of the occurrence may be an example of a quantitative difference. Determining the change of the occurrence allows to provide a time-resolved insight into the occurrences. The change of the occurrence can be seen as a measure for a relevance associated with a certain desired property of the chemical product over time. Based on this, extrapolation to future points in time can be obtained. Typically, production of chemical products needs time to be adapted. Furthermore, desired properties of chemical products may change over time due to changing desires of end product consumers. Hence, long-term planning is desired. Extrapolations to future points in time provides the opportunity to tailor chemical production to future developments. An example of different occurrences over time can be seen in FIG. 7.

[0102] At least one element may be selected according to the change of the occurrence associated with the at least one element 414. This may include selecting a predefined number of elements associated with the highest and / or lowest change of the occurrence and / or selecting the at least one element with a change of the occurrence within a target occurrence range.

[0103] At least one chemical product data set comprising the at least one selected element may be identified 416. The chemical product data set comprising the at least one selected element may be highly related to the selected topic. Therefore, the chemical product data set comprising the at least one selected element may be relevant for obtaining the chemical product with the desired property.

[0104] One or more desired properties associated with the at least one identified chemical product data set may be determined 418. From the data sets corresponding to the request and identified as relevant, the desired properties associated with the request may be determined. The combination of the relevant data sets and the requests allows to extract the context needed for determining the desired properties of the chemical product. Examples of matching request with desired properties can be seen in FIG. 3. Determining the desired properties allows to match the request with suited chemical products.

[0105] Determining the one or more desired properties may comprise querying a generative data-driven model such as a large language model and / or any other natural language processing (NLP) model for providing the desired 240435

[0106] 21 property associated with the identified at least one chemical product data set. Optionally, at least a part of the request may be provided together with the at least one identified chemical product data set to the generative model. The generative data-driven model may be prompted by one or more task instruction(s) to provide the desired property associated with the identified at least one chemical product data set. By doing so, the generative data-driven model may obtain the chemical product data set as context in order to extract the desired property. This allows to identify desired properties beyond language barriers such as synonyms and requires no human intervention for determining the desired properties. Further, using a generative data-driven model can accelerate determining the one or more desired properties. To avoid hallucination by the generative data-driven model a human oversight may be implemented. This may include a human checking the determination of desired properties by the generative data-driven model at predefined points in time and / or if a confidence score associated with determining the desired properties by the generative data-driven model may be within a evaluation range. An example of the generative data-driven model may be found in the context of FIG. 11 A to FIG. 12.

[0107] Additionally or alternatively, the at least one identified chemical product data set may be provided to a classification data-driven model. The classification data-driven model may be configured to classify whether the elements comprised by the at least one identified chemical product data set correspond to the desired property. The classification data-driven model may comprise a pretrained text representation models 804 and a custom model trained on several hundred manual annotations 806. The pretrained text representation model 804 may be configured to generate a numerical representation of the data provided to the pretrained text representation model 804. The numerical representation generated by the pretrained text representation model 804 may be indicative of the relation between two or more elements of the data provided to pretrained text representation model 804. In an example, the pretrained text representation model 804 may comprise one or more encoder input(s), one or more embedding layer(s) and / or one or more encoder block(s). The custom model trained on several hundred manual annotations 806 may be configured to obtain the numerical representation generated by the pretrained text representation model 804 and classify if the numerical representation may be related to the desired property. The classification data-driven model is further described in the context of FIG. 8. By doing so, the exact language of the chemical product data set is reproduced when determining the desired properties. This avoids hallucination and allows an error-free, accurate and fast determination of the desired properties. The desired properties determined may be provided for producing the chemical product associated with the desired properties.

[0108] An indication of the chemical product associated with the desired properties may be determined 420. This may include providing an indication of the desired property to a property database. The property database may comprise a plurality of chemical products and corresponding properties associated with the chemical product. The property database may be configured to provide an indication of the chemical product associated with the desired 240435

[0109] 22 property in response to receiving the indication of the chemical product. Additionally or alternatively, the indication of the chemical product may be received via a user interface, e.g. in response to providing the indication of the desired property via the user interface.

[0110] The indication of the chemical product may be provided for producing the chemical product associated with the desired property 422, e.g. via a user interface to a worker of one or more chemical production facilities 102 and / or a specification for producing the chemical product with the desired property may be retrieved for producing the chemical product and / or the specification for producing the chemical product with the desired property may be provided to a control engine 118 as it can be seen in FIG. 1.

[0111] In an embodiment, at least one future point in time preceding the at least two points in time associated with the chemical product data sets may be obtained, in particular received, e.g. via a user interface. An occurrence and / or a change of occurrence may be determined with respect to the at least one future point in time, e.g. by extrapolating the occurrence and / or the change of the occurrence to the future point in time. For this purpose, parameters associated with a predefined mathematical function may be determined in order to achieve a distance between the mathematical function associated with the determined parameters within a target distance range. Examples may include minimizing chi squared and / or R squared. The desired property may be determined at the future point in time and / or may be provided to prepare the production of the chemical product with the desired property for the future point in time.

[0112] In an embodiment, an indication on the quantity of resources associated with the application of the chemical product may be obtained, for example together with the request and / or extracted from the at least one identified chemical product data set. Extracting the indication of the quantity from the chemical product data set may comprise providing at least one task instruction including the at least one identified chemical product data set as described in the context of 418.

[0113] FIG. 5 illustrates an embodiment of obtaining a database comprising one or more chemical product data set(s) related to an application associated one or more desired properties of one or more chemical product(s).

[0114] A plurality of chemical product data sets related to a plurality of applications associated with a plurality of desired properties of a plurality of chemical products, e.g. from one or more database(s) and / or via an interface such as a user interface 504.

[0115] A selection of at least a part of at least one of the plurality of chemical product data sets may be obtained 506. The obtained selection may specify a part of at least one of the chemical product data sets to be processed 240435

[0116] 23 further. For example, at least one of the plurality of chemical product data sets may comprise two or more parts. The parts may be related to different applications and / or at least one of the two or more parts may comprise more details related to the application than the at least other part of the two or more parts. By selecting at least a part of the chemical product data sets, the relevant and / or more detailed part of the data set can be selected for obtaining the chemical product with the one or more desired properties.

[0117] One or more element(s) may be eliminated from the at least one selected part of the plurality of chemical product data sets. NLP models may get distracted from less meaningful and context-determining words or symbols such as “the”, or the like. Hence, the elements independent from the context of the chemical product data set may be removed. Consequently, accuracy of results from NLP models operating on the processed data can be improved. Ultimately, this improves data extraction from chemical product data sets.

[0118] A selection of at least one of the plurality of chemical product data set may be obtained 510. Obtaining the selection of the at least one chemical product data set may comprise obtaining a request associated with an application of one or more chemical product(s) having at least one desired property and matching at least a part of the request with the plurality of chemical product data sets. Matching at least the part of the request with the plurality of chemical product data sets may comprise matching elements comprised by the chemical product data sets and at least the part of the request. One element may refer to one word, an ensemble of words, a number, an ensemble of numbers or the like. Matching elements may comprise performing a keyword search and selecting one or more chemical product data set(s) with the highest matching score. The matching score may be indicative of a number of matching elements. Optionally, the matching score may be obtained by weighting matching of different elements differently. Additionally or alternatively, matching the elements may comprise performing an embedding search. This may include obtaining a numerical representation per chemical product data set and a numerical representation of the request. Obtaining a numerical representation may include providing the chemical product data sets and / or the request to one or more embedding layer(s) as described in the context of FIG. 10. Further, performing the embedding search may comprise determining a distance between the numerical representations of the chemical product data sets and the numerical representation of the request. By doing so, the chemical product data sets relevant to the query can be identified and less computational resources are used for obtaining the chemical product with the desired properties.

[0119] One or more keyword(s) per selected chemical product data set may be obtained 512. The keywords may allow to identify and / or retrieve at least one chemical product data set. The keywords may be related to the context of the selected chemical product data set(s). 24

[0120] Additionally or alternatively to 512, one or more numerical representation(s) of the one or more selected chemical product data set(s) may be obtained 514. The one or more numerical representation(s) may allow to identify and / or retrieve at least one chemical product data set via semantic search, i.e. determining a distance between the numerical representation(s) of the chemical product data set(s) and for example the request as described in the context of FIG. 4. In an example, numerical representation of the one or more chemical product data set(s) may comprise an embedding of the one or more chemical product data set(s).

[0121] The one or more selected chemical product data set(s), the one or more keyword(s) and / or the one or more numerical representations may be provided via a database. Preferably, the one or more selected chemical product data set(s) provided may be associated with at least two different points in time. Additionally or alternatively, the database may be updated with one or more chemical product data set(s) associated with another point in time.

[0122] FIG. 6 illustrates an embodiment of determining an occurrence of one or more element(s) within a chemical product data set.

[0123] Determining the occurrence of the one or more element(s) within a chemical product data set may comprise determining one or more topic(s) associated with the chemical product data set, preferably via topic modelling.

[0124] Topic modeling may be used in natural language processing and machine learning to discover and extract hidden themes or topics from a large corpus of text documents. It may be particularly useful for analyzing unstructured text data, such as news articles, social media posts, customer reviews, or academic papers. The goal of topic modeling may be to automatically identify and assign topics to each document in a given collection, without any prior knowledge of the topics. It helps in understanding the main themes or subjects that are present in the text data, and provides a way to organize and categorize documents based on these topics. This can be beneficial determining the context of a chemical product data set.

[0125] In topic modeling, a topic may refer to a particular theme or subject associated with at least one text document. The topic may be related to and / or may comprise a set of elements such as words or symbols occuring associated with each other within the document. Each topic may be characterized by a distribution of elements. Element associated with a higher occurrence may be associated with a higher contribution to the distribution of the elements.

[0126] An example of at least a part of a chemical product data set and corresponding topics may be seen in FIG. 6. Topic modelling may be performed by applying Latent Dirichlet Allocation (for example LatentDirichletAllocation implementation in python, scikit learn 0.17 or later), Non-Negative Matrix Factorization (for example NMF implementation in scikit learn 0.17 or later), Top2Vec (for example Top2Vec algorithm top2vec 1.0.34 in python) or BERTopic (for example BERTopic algorithm bertopic 0.16.3 in python).

[0127] FIG. 7 illustrates an embodiment of occurrences at different points in time.

[0128] In particular, the temporal evolution of occurrences of two topics can be seen, occurrences of topic 1 increased within the shown time frame while the occurrences of topic 2 decreased. An extrapolation of the shown occurrences may provide an even larger difference between the occurrences of topic 1 in relation to the occurrences of topic 2. As a consequence, topic one may be more relevant for large-scale chemical production. Hence, topic 1 may be selected over topic 2 according to the occurrences and / or the change of occurrence.

[0129] FIG. 8 illustrates an embodiment of a classification data-driven model configured to classify whether the elements comprised by the at least one identified chemical product data set correspond to the desired property.

[0130] The classification data-driven model may comprise a pretrained text representation models 804 and a custom model trained on several hundred manual annotations 806. The pretrained text representation model 804 may be configured to generate a numerical representation of the data provided to the pretrained text representation model 804. The numerical representation generated by the pretrained text representation model 804 may be indicative of the relation between two or more elements of the data provided to pretrained text representation model 804. In an example, the pretrained text representation model 804 may comprise one or more encoder input(s), one or more embedding layer(s) and / or one or more encoder block(s). The custom model trained on several hundred manual annotations 806 may be configured to obtain the numerical representation generated by the pretrained text representation model 804 and classify if the numerical representation may be related to the desired property.

[0131] The pretrained text representation model 804 may be trained on a plurality of data sets, preferably comprising text data related to different contexts. The pretrained text representation model 804 may be trained for generating a numerical representation of the data provided to the pretrained text representation model 804.

[0132] The custom model trained on several hundred manual annotations 806 may comprise one or more connected layer(s). The output of the one or more connected layer(s) may be a numerical value per provided digital representation. The provided numerical value may be indicative of a classification of the provided digital representation. The custom model trained on several hundred manual annotations 806 may be trained based on at least parts of chemical product data sets and corresponding indications on the classification of at least parts of 240435

[0133] 26 the chemical product data set. In particular, the custom model trained on several hundred manual annotations 806 may be trained based on a plurality of numerical representations of at least parts of a plurality of chemical product data sets and a plurality of indications on the classification, preferably one indication of the classification per numerical representation of at least parts of the plurality of chemical product data sets.

[0134] In an embodiment, the custom model trained on several hundred manual annotations 806 may be trained to identify a trigger. The trigger may indicate that one or more datapoint(s) related to the desired property may be close to, in particular following, the datapoint(s) related to the trigger.

[0135] FIG. 9 illustrates an embodiment of a computing apparatus for obtaining a chemical product with a desired property.

[0136] The computing apparatus may comprise an intake interface 904 in connection with a chemical product data set selection engine 906 and an occurrence determining engine 908. The intake interface may be configured to obtain, in particular receive a plurality of chemical product data sets associated with at least two different points in time as described in the context of 404 and / or obtain, in particular receive a request for obtaining a chemical product as described in the context of 402. The chemical product data set selection engine 906 may be configured to the request from the intake interface 904. Additionally or alternatively, the chemical product data set selection engine 906 may be configured to select at least one chemical product data set per point in time by matching at least a part of the request with the plurality of chemical product data sets as described in the context of 406. The chemical product data set selection engine 906 may be in communication with a database comprising chemical product data sets 902. The database comprising the chemical product data sets 902 may be configured to provide the chemical product data sets to the chemical product data set selection engine 906. Further, the chemical product data set selection engine 906 may be in communication with the occurrence determining engine 908. The occurrence determining engine 908 may be configured to determine an occurrence of one or more element(s) within the at least one selected chemical product data set per point in time as described in the context of 408. The occurrence determining engine 908 may be in communication with the intake interface 904. The intake interface 904 may be further configured to obtain, in particular receive a selection of the at least one element. Additionally or alternatively, the occurrence determining engine 908 may be in communication with a preselection engine. The preselection engine may be configured to select at least one element according to the determined occurrence associated with the element as described in the context of 410. The occurrence determining engine 908 may be further in communication with a change of occurrence determining engine 912. The change of occurrence determining engine 912 may be configured to determine a change of the occurrence related to a difference of the occurrence of the selected at least one element between the at least two different points in time. 240435

[0137] 27

[0138] The change of occurrence determining engine 912 may be in communication with an element selection engine 910. The element selection engine 910 may be configured to select least one element according to the change of the occurrence associated with the at least one element as described in the context of 414. The element selection engine 910 may be in communication with a chemical product data set identification engine 914. The chemical product data set identification engine 914 may be configured to identify at least one chemical product data set comprising the at least one selected element as described in the context of 416. The chemical product data set identification engine 914 may be in communication with desired property determining engine 916. The desired property determining engine 916 may be configured to determine one or more desired properties associated with the at least one identified chemical product data set as described in the context of 418. The desired property determining engine 916 may be in communication with chemical product identification engine 920. The chemical product identification engine 920 may be configured to determine an indication of the chemical product associated with the desired properties as described in the context of 420. The chemical product identification engine 920 may be in communication with a database comprising chemical products and corresponding properties 918 and an output interface 922. The database comprising chemical products and corresponding properties 918 may be configured to provide the indication of the chemical product associated with the desired property upon receiving an indication of the desired property. The output interface 922 may be configured to provide the indication of the desired property and / or the indication of the chemical product associated with the desired property as described in the context of FIG. 4.

[0139] In an embodiment, the intake interface 904 may be further configured to obtain, in particular receive at least one future point in time. The intake interface 904 may be in communication with an extrapolation engine. The extrapolation engine may receive the change of occurrence associated with at least two different points in time from the occurrence determining engine 908. Further, the extrapolation engine may be configured to determine the occurrence at the future point in time as described in the context of FIG. 4.

[0140] In an embodiment, the intake interface 904 may be further configured to obtain, in particular receive an indication on the quantity of resources associated with the application of the chemical product. Additionally or alternatively, the at least one chemical product data set may be provided to an extraction engine by the chemical product identification engine 920. The extraction engine may be configured to extract the indication of the quantity of resources from the chemical product data set.

[0141] FIG. 10 illustrates an embodiment of a data-driven model, i.e. a transformer encoder comprising an encoder input 1088, one or more encoder block(s) 1086 and an encoder output 1076 and / or a transformer decoder comprising a decoder input 1094, one or more decoder block(s) 1090 and a decoder output 1092 and / or a transformer encoderdecoder. 240435

[0142] 28

[0143] The transformer encoder comprises an encoder input 1078, one or more encoder blocks 1074, 1014 and an encoder output. A plurality of transformer encoder architectures are available in the art such as the bi-directional encoder representations from transformers (BERT). The input data may be received at the encoder input 1078. The input data may comprise at least one of text data, numerical data, tabular data, image data or the like. Where the input data may comprise one of text data, numerical data, tabular data, image data or the like, input embedding of a type corresponding to the type of input data may be applied. Hence, the input embedding may be configured to map text data, numerical data, tabular data, image data or a combination thereof to a numerical representation of the input data. An example of input embedding associated with text data may be a continuous bag-of-words-model (CBOW). Additionally or alternatively, Word2Vec may be used for representing input data. Upon receiving the input data, the input data may be tokenized via a vocabulary associated with a preselection of elements of expected input data. Applying the input embedding may comprise mapping the input data, in particular the two or more elements of the input data to a numerical represenation of the input data, preferably of a predefined size e.g. via padding. Further, the encoder input 1078 may apply positional encoding 1004. For example, the positional factor pposmay be obtained based on the following equation:

[0144] 1000077 where pos may refer to the position of the element within the sequence, I may refer to the dimension associated with the input embedding and d may refer to the dimension of the data-driven model. Alternatively, the positional encoding may be based on rotary positional embeddings (RoPE). The embedded input data may be processed by the encoder block. The embedded input data may be provided to the layer normalization 1008 by a residual connection. Multi-head self-attention 1006 may be applied to the embedded input data. The embedded input data may serve as query Q, key K and value V with respect to the self-attention operation. For improving the efficiency, multiple heads are used to apply the filter according to the following equation: head i = Attention(QWtQ, KWtK, 7Wi') with parameter matrices may refer to the number of heads, dv, dKand dQmay refer to the dimensions of the value, key and query.

[0145] The result of the two or more head may be concatenated according to the following equation: MultiHead(Q, K, 7) = Concat(head 1, . . . , headh)W(>where IVOe ^hdvxdanc|may numberof heads. This may result in a context tensor. After the multi-head self-attention 1006 layer normalization 1008 may be applied based on the context tensor and / or the embedded input data from the residual connection. The so-obtained tensor may be passed to a feed-forward layer 1010 again followed by layer normalization 1012 based on the residual connection to the context tensor and / or the output of the feed-forward layer 1010. 240435

[0146] 29

[0147] The encoder output 1076 may comprise a linear layer 1034 and a softmax layer 1036. The encoder output may be configured to map the concatenated numerical representation of the two or more elements and the relation between the two or more elements to a numerical representation of the output data. Additionally or alternatively, a decoding model may be used to map the concatenated numerical representation of the two or more elements and the relation between the two or more elements to a numerical representation of the output data. The decoding model may be trained to relate a numerical represenation of data of a type according to the input data.

[0148] The transformer decoder comprises a decoder input 1084, one or more decoder blocks 1080, 1028 and a decoder output 1092. A plurality of transformer decoder architectures are available in the art such as the generative pretrained transformers (GPT). In contrast to the transformer encoder, the transformer decoder may perform masked multi-head self-attention 1020 by additionally masking a part of the embedded input data associated with elements later in the sequence than the element to be generated. Additionally or alternatively, the part of the input data associated with elements later in the sequence than the element to be generated may not be received and / or transformed into the embedded input data.

[0149] The transformer encoder-decoder may comprise a combination of the transformer encoder and transformer decoder wherein the context tensor obtained from the encoder block may be used for the multi-head self-attention 1064 operation in at least one decoder block 1090.

[0150] Input data to the data-driven model may comprise image data. The encoder input and / or decoder input may comprise one or more linear projection layer(s) for a linear projection of a sequence of two or more partial images. This may result in changing the dimension of the one or more received images. Furthermore, positional embedding may be applied to the sequence, preferably by passing the sequence of one or more images and / or partial images through the one or more linear projection layer(s). Additionally or alternatively, the input data may comprise tabular data. Input embeddings for tabular input data may comprise a token embedding, a positional embedding, a column embedding, a row embedding or a combination thereof.

[0151] In an example, the data-driven model may comprise a Mamba block. A Mamba architecture may enhance inference speed in relation to a transformer based model. Mamba block may be based on a selective space state sequence model. A selective state space layer may be a linear recurrent network that selectively process data based on the input token, which may allow to focus on relevant data and discard irrelevant data. For instance in each step a separate weight vector may be determined based on the respective input token. The determined weight vector may then be used in a selective scan. A selective state space layer may be used in a convolutional mode e.g. for parallelizable training and a recurrent mode for near-constant time generation of output data. A state space operation may be based on solving the state and output equations, wherein a state equation may describe how a state changes based on how the input influences the state and an output equation may describe how the state is translated to the output. Further how the input influences the output may be represented by a learnable linear transformation used in a learnable skip connection. An example of the architecture of a mamba block may be found in “Mamba: Linear-Time Sequence Modeling with Selective State Spaces” by Albert Gu and Tri Dao arXiv:2312.00752v2 [cs.LG] 31 May 2024, , which is incorporated herein by reference.

[0152] In an example, any one of the data-driven models may further comrpise a mixture of experts block. The mixture of experts block may be decoder blocks wherein the feed-forward layer may be exchanged for a gating network and a number of parallel feed-forward layers, wherein the gating network may switch between the feed-forward layers depending on the input. This may allow leveraging advantages of the different architectures.

[0153] The data-driven model may generate one token per timestep upon processing the input data and optionally all output tokens already produced. The training data set may comprise a plurality of sequences comprising a plurality of elements. During the training of the data-driven model, sequences of the training data set may be provided to the data-driven model and one or more elements may be generated based on the sequences of the training data set one by another. The elements generated based on the sequences may follow the elements of the parts of sequences the data-driven model may have been provided with. The generated one or more elements may be compared to the one or more elements following the at least a part of the sequences provided to the data- driven model in order to adapt the parameters of the data-driven model depending on a deviation of the predefined token and the generated token.

[0154] The present disclosure has been described in conjunction with preferred embodiments and examples as well. However, other variations can be understood and effected by those persons skilled in the art and practicing the claimed invention, from the studies of the drawings, this disclosure and the claims.

[0155] Any steps presented herein can be performed in any order. The methods disclosed herein are not limited to a specific order of these steps. It is also not required that the different steps are performed at a certain place or in a certain computing node of a distributed system, i.e. each of the steps may be performed at different computing nodes using different equipment / data processing.

[0156] As used herein ..determining" also includes ..initiating or causing to determine", “generating" also includes ..initiating and / or causing to generate" and “providing” also includes “initiating or causing to determine, generate, select, send and / or receive”. “Initiating or causing to perform an action” includes any processing signal that triggers a computing node or device to perform the respective action.

[0157] In the claims as well as in the description the word “comprising” or “including” or similar wording does not exclude other elements or steps and shall not be construed limiting to the elements or steps lined out. The indefinite article “a” or “an” does not exclude a plurality. A single element or other unit may fulfill the functions of several entities or items recited in the claims. The mere fact that certain measures are recited in the mutual different dependent claims does not indicate that a combination of these measures cannot be used in an advantageous implementation or further elements may be included. 240435

[0158] 31

[0159] Providing in the scope of this disclosure may include any interface configured to provide data. This may include an application programming interface, a human-machine interface such as a display and / or a software module interface. Providing may include communication of data or submission of data to the interface, in particular display to a user or use of the data by the receiving node, entity or interface.

[0160] Various units, circuits, entities, nodes or other computing components may be described as “configured to” perform a task or tasks. Configured to shall recite structure meaning “having circuitry that” performs the task or tasks on operation. The units, circuits, entities, nodes or other computing components can be configured to perform the task even when the unit / circuit / component is not operating. The units, circuits, entities, nodes or other computing components that form the structure corresponding to “configured to” may include hardware circuits and / or memory storing program instructions executable to implement the operation. The units, circuits, entities, nodes or other computing components may be described as performing a task or tasks, for convenience in the description. Such descriptions shall be interpreted as including the phrase “configured to.” Any recitation of “configured to” is expressly intended not to invoke 35 U.S.C. § 112(f) interpretation.

[0161] In general, the methods, apparatuses, systems, computer elements, nodes or other computing components described herein may include memory, software components and hardware components. The memory can include volatile memory such as static or dynamic random-access memory and / or nonvolatile memory such as optical or magnetic disk storage, flash memory, programmable read-only memories, etc. The hardware components may include any combination of combinatorial logic circuitry, clocked storage devices such as flops, registers, latches, etc., finite state machines, memory such as static random-access memory or embedded dynamic random-access memory, custom designed circuitry, programmable logic arrays, etc.

[0162] Any disclosure and embodiments described herein relate to methods, systems, apparatuses, devices, chemicals, materials, computer program elements lined out above and vice versa. Advantageously, the benefits provided by any of the embodiments and examples equally apply to all other embodiments and examples and vice versa.

[0163] All terms and definitions used herein are understood broadly and have their general meaning.

Claims

24043532CLAIMSWhat is claimed is:1 . A method, in particular computer-implemented method, for obtaining a chemical product, the method comprising: obtaining, in particular receiving one or more chemical product data set(s) related to an application of the chemical product, wherein the one chemical product data set(s) comprise one or more element(s), selecting one or more chemical product data set(s) according to an occurrence of the one or more element(s) per chemical product data set, determining one or more desired properties of the chemical product per selected chemical product data set, providing the one or more desired properties of the chemical product for providing, in particular producing the chemical product associated with the one or more desired properties.

2. The method of claim 1 , further comprising obtaining an indication of a selection of at least one of the one or more element(s), in particular via a user interface.

3. The method of claim 1 or 2, wherein the occurrence of the one or more element(s) comprises the occurrence of one or more combination(s) of two or more elements.

4. The method of any one of claims 1 to 3, wherein two or more chemical product data sets are obtained, in particular received, and wherein the two or more chemical product data sets are associated with two or more points in time, and further comprising determining a change of the occurrence related to a difference of the occurrence of the one or more element(s) between the at least two different points in time, and wherein the one or more chemical product data set(s) are selected according to the change of the occurrence.

5. The method of any one of claims 1 to 4, wherein two or more chemical product data sets are obtained, in particular received, and wherein the two or more chemical product data sets are associated with two or more points in time, and further comprising obtaining, in particular receiving, at least one future point in time following the at least two points in time associated with the chemical product data sets, in particular via a user interface, wherein the occurrence and / or the change of occurrence is further determined associated with the at least one future point in time.

336. The method of any one of claims 1 to 5, wherein the one or more chemical product data set(s) may be associated with two or more two or more points in time, and further comprising obtaining, in particular receiving, an indication of one or more point(s) in time, and further comprising selecting one or more of the two or more chemical product data set(s) according to the indication of the one or more point(s) in time.

7. The method of any one of claims 1 to 6, further comprising obtaining, in particular receiving, an indication of one or more location(s) and further selecting the one or more chemical product data set(s) according to the indication of one or more location(s).

8. The method of any one of claims 1 to 7, wherein determining the one or more desired properties comprises providing the one or more selected chemical product data set(s), in particular via a user interface, and receiving the desired property associated with the one or more selected chemical product data set(s) in response to providing the one or more selected chemical product data set(s), in particular via a user interface.

9. The method of any one of claims 1 to 8, wherein determining the one or more desired properties comprises providing a task instruction for determining the desired property associated with the one or more selected chemical product data set(s) to a generative data-driven model, and wherein the task instruction is related to, in particular includes, the one or more selected chemical product data set(s), and wherein the generative data-driven model is configured to follow task instructions provided.

10. The method of any one of claims 1 to 9, wherein determining the one or more desired properties comprises providing the one or more selected chemical product data set(s) to a classification data-driven model for classifying at least one part of the one or more chemical product data set(s) related to the desired property, wherein the classification data-driven model is configured to classify whether the parts of chemical product data sets provided to the classification data-driven model are related to the desired property.11 . The method of any one of claims 1 to 10, wherein determining the one or more desired properties comprises providing the one or more selected chemical product data set(s) to a classification data-driven model for classifying at least one part of the one or more chemical product data set(s) related to the desired property, wherein the classification data-driven model comprises a representation generating engine configured to generate a numerical representation of data provided to the representation generating engine and a custom model configured to receive the numerical representation of the data from the representation generating engine and provide an indication of whether the numerical representation provided to the custom model is related to the desired property.2404353412. The method of any one of claims 1 to 1 1 , further comprising obtaining an indication of a chemical product associated with the desired property by providing a request for receiving an indication of the chemical product associated with the desired property to a database, wherein the request for obtaining the indication of the chemical product is related to, in particular comprises the desired property, and wherein the database comprises a plurality of indications of chemical products and corresponding properties, and wherein the database is configured to provide the indication of the chemical product associated with the desired property in response to receiving the request for obtaining the indication of the chemical product, and further comprising providing the indication of the chemical product for providing, in particular producing the chemical product associated with the desired property.

13. An apparatus for obtaining a chemical product, the apparatus comprising: a processor configured for performing any one of the methods according to any one of claims 1 to 12.

14. Use of a property as determined by any one of claims 1 to 12 for providing, in particular producing the chemical product.

15. Use of an indication of a chemical product associated with a property determined by any one of claims 1 to 12 for providing, in particular producing the chemical product associated with the desired property.

Images