Artificial intelligence-based sustainable material design

Abstract

To provide methods for artificial intelligence (AI)-based sustainable material design.SOLUTION: A method includes: receiving a dataset including information associated with scientific literature; determining a set of materials and information associated with the set of materials based on application of neural network models to the dataset; generating embeddings for the set of materials indicative of features of each material and effect of each material on a living environment; training a generative AI model based on the embeddings; receiving a user input indicative of information associated with a queried material; generating embeddings for the queried material indicative of features of the queried material and its effect on the living environment; determining sustainability information associated with the queried material based on application of the generative AI model to the embeddings generated for the queried material; and rendering the sustainability information.SELECTED DRAWING: Figure 5

Description

Technical Field

[0001] The embodiments described in the present disclosure relate to artificial intelligence (AI)-based sustainable material design.

Background Art

[0002] Advances in the field of artificial intelligence (AI) have led to the development of numerous AI-based frameworks and related software to assist in the discovery, design, and synthesis of sustainable materials. AI-based frameworks significantly facilitate the discovery, design, and synthesis of materials or combinations of materials as compared to conventional methods and techniques that require years, if not decades, of effort for material discovery, design, and synthesis. AI-based frameworks are scalable, flexible, self-contained, and may enable the use of machine learning models (such as deep learning models). Machine learning models can be applied to scientific data with a particular focus on material information for the discovery of sustainable or environmentally compliant materials. Related software can provide solutions such as structure-based virtual screening of ultra-large chemical libraries for the discovery of sustainable materials, and acceleration of the discovery of materials (such as polypeptide materials) for solving problems such as artificial enzyme design and understanding intrinsically disordered proteins. AI-based frameworks can also be used for predictions such as automated synthesis of materials (such as synthesis of inorganic synthesis recipes), retrosynthesis (synthesis of the generated materials), and construction of element-wise graph neural networks for learning relationships between materials (entities) using knowledge graphs.

[0003] However, the generation of scientific data and subsequent extraction of relevant information from such scientific data can be challenging for deep learning models and graph neural networks. This can be due to the challenges associated with testing the quality and certainty of scientific data and determining the relevance of scientific data in assisting with the discovery or synthesis of materials. Additionally, the identified or synthesized materials may not be environmentally compliant and may have an adverse impact on the resources of the living environment.

[0004] The subject matter claimed in this disclosure is not limited to embodiments that solve any disadvantages or embodiments that operate only in an environment such as those described above. Rather, this background is provided only to illustrate an example technical field in which some of the embodiments described in this disclosure may be practiced. SUMMARY OF THE INVENTION

[0005] According to one aspect of an embodiment, the method may include a set of operations that may include receiving a dataset that includes information related to scientific literature. The set of operations may further include applying one or more neural network models to the received dataset. The set of operations may further include determining a set of materials and information related to each material in the set of materials based on the application of the one or more neural network models. The set of operations may further include generating a first set of embeddings that represent a first set of features of each material in the set of materials. The set of operations may further include generating a second set of embeddings related to text content that describes the impact of the set of materials on the resources of the living environment. The set of operations may further include training a generative artificial intelligence (AI) model based on the first set of embeddings and the second set of embeddings. The set of operations may further include receiving user input that indicates information related to the queried material. The set of operations may further include generating a third embedding that represents a second set of features of the queried material and a fourth embedding related to text content that describes the impact of the queried material on the resources of the living environment. The set of operations may further include applying the generative AI model to the third embedding and the fourth embedding. The set of operations may further include determining sustainability information related to the queried material based on the application of the generative AI model. The set of operations may further include controlling a display device to render the sustainability information related to the queried material.

[0006] The objectives and advantages of the embodiments are realized and achieved at least by the elements, features, and combinations particularly pointed out in the claims.

[0007] The above summary and the following detailed description are given by way of example and are explanatory and not restrictive of the invention claimed.

Brief Description of the Drawings

[0008] Through the use of the accompanying drawings, embodiments are described with further specificity and detail.

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

[0009] All are in accordance with at least one embodiment described in the present disclosure.

Mode for Carrying Out the Invention

[0010] Some embodiments described in this disclosure relate to methods and systems for artificial intelligence (AI)-based sustainable material design. Here, sustainable material design may include receiving a dataset that includes information related to scientific literature. Thereafter, one or more neural network models may be applied to the dataset. Based on the application of the one or more neural network models to the dataset, a set of materials and information related to each material in the set of materials may be determined. Thereafter, a first set of embeddings may be generated that represent a first set of characteristics of each material in the determined set of materials. Further, a second set of embeddings may be generated that is related to text content describing the impact of the determined set of materials on the resources of the living environment. Thereafter, a generative AI model may be trained based on the first set of embeddings and the second set of embeddings. After training, a user input indicating information related to the queried material may be received. In response to receiving the user input, a third embedding and a fourth embedding may be generated. The third embedding may represent a second set of characteristics of the queried material. The fourth embedding may be related to text content describing the impact of the queried material on the resources of the living environment. In response to the generation of the third embedding and the fourth embedding, the trained generative AI model may be applied to the third embedding and the fourth embedding. Thereafter, based on the application of the generative AI model to the third embedding and the fourth embedding, sustainability information related to the queried material may be determined. Finally, the rendering of the sustainability information related to the queried material may be controlled on a display device.

[0011] Traditional means of materials design generally require years, if not decades, of effort to discover, generate, or synthesize sustainable materials. However, with the advent of big data and advanced algorithms, AI-based frameworks have been developed to significantly facilitate the discovery, generation, or synthesis of sustainable materials. For example, scalable, flexible, and self-contained frameworks can be used for the application of deep learning models and methods to scientific data, with a particular focus on materials science for materials discovery. AI-based frameworks can function in conjunction with software to enable structure-based virtual screening of ultra-large chemical libraries and accelerate the discovery of new materials to address unknown challenges. For example, the discovery of polypeptide materials can be accelerated, enabling the addressing of challenges such as designing artificial enzymes and constructing an understanding of intrinsically disordered proteins. Graph neural networks are being used to automate materials synthesis and build an understanding of retrosynthesis. For example, element-wise graph neural networks can be used to predict the inorganic synthesis of recipes. Retrosynthesis (i.e., the synthesis of developed materials) can be useful for understanding the precursors of specific reactions. Additionally, knowledge graphs can be used to facilitate the learning of relationships between different entities through graph neural networks. However, AI-based frameworks may not include a mechanism to verify whether the discovered or synthesized materials are environmentally compliant or have adverse effects on the environment (e.g., depletion of existing resources due to increased carbon dioxide (CO 2 ) emissions from those materials, etc.).

[0012] For the discovery and synthesis of materials, natural language processing (NLP) techniques are being explored because NLP techniques facilitate the extraction of information from datasets (such as chemical tables). Based on that information, entities that make up the material can be recognized. For example, a dataset can be generated by using NLP techniques. Based on that dataset, relationships between chemical interactions that may be involved in the synthesis of organic materials can be identified. Thereafter, the organic material can be synthesized by using reactants involved in those chemical interactions. However, determining the quality and certainty of the information contained in the dataset, as well as the relevance of the information for the discovery or synthesis of sustainable or organic materials, can be challenging.

[0013] According to one or more embodiments of the present disclosure, the technical field of environmentally compliant sustainable material design can be improved by configuring a computing system (e.g., an electronic device) such that the computing system can generate environmentally compliant materials through the use of a natural language model and a generative AI model. The computing system may receive a dataset containing information related to scientific literature. The scientific literature may be published in articles, journals, books, or presentations related to materials science and chemical science. The scientific literature can be applied as input to a neural network model for the extraction of information including entities (e.g., reactants or products, etc.) that can be involved in chemical reactions and can be generated as a result of chemical reactions. The information extracted can further include information related to the entity, such as the organic structure of the entity, the category of the entity (catalyst, reactant, or product), the time and temperature at which the chemical reaction in which the entity is involved occurs, and the precursors related to the chemical reaction in which the entity is involved. Based on the information related to the entity, the computing system can classify the entity as sustainable (environmentally compliant) or non-sustainable (not environmentally compliant, or potentially harmful to the environment). Entities considered non-sustainable or potentially harmful can be filtered out, and as a result, beneficial or sustainable entities can be determined.

[0014] A computing system can determine a set of materials based on those entities. Each material in the set of materials can include one or more entities. For each material in the set of materials, an embedding can be generated. Further, another embedding can be generated based on information associated with each of the one or more entities that make up each material in the set of materials. Thus, for each material in the set of materials, two embeddings can be generated. The generation of these two embeddings can be based on the application of a natural language model (e.g., a BERT (Bidirectional Encoder Representations from Transformers) model) to the set of materials and information related to the set of materials. Using the generated embeddings related to the set of materials, a generative AI model that can include a generator model and a discriminator model can be trained. Based on the embeddings, the discriminator model can be configured to classify whether each material in the set of materials is sustainable or harmful. The generator model can be trained based on the embeddings and the input from the discriminator model. The generator can be trained to generate an output for each material in the set of materials based on the corresponding embedding related to the material. The output can be such that the receipt of the output by the discriminator model can cause the discriminator model to classify the corresponding material as sustainable. Receiving a user input indicating information related to a queried material, the computing system can generate an embedding and supply the embedding to the generative AI model. Based on the embedding, the generative AI can predict whether the queried material is sustainable and generate an explanation that outlines the basis behind the prediction. Thus, the computing system of the present disclosure can use the trained generative AI to predict whether a given material is sustainable and also provide an explanation related to the basis behind the prediction. These predictions and the provided basis can be useful for the rapid design of new sustainable and environmentally friendly materials and new compositions.

[0015] Embodiments of the present invention will be described with reference to the accompanying drawings.

[0016] FIG. 1 is a diagram showing an example of a network environment related to artificial intelligence (AI)-based sustainable material design according to at least one embodiment described in the present disclosure. Referring to FIG. 1, a network environment 100 is shown. The network environment 100 may include an electronic device 102 and a server 104 (which may host a database 106). The electronic device 102 and the server 104 may be communicatively coupled to each other via a communication network (such as communication network 108). The electronic device 102 may include one or more neural network models 110, a natural language model 112, and a generative AI model 114. The generative AI model 114 may include a generator model 114A and a discriminator model 114B. FIG. 1 also shows that the electronic device 102 may include information such as, for example, a set of materials 116, a first set of embeddings 118, a second set of embeddings 120, and a queried material 122. Also, the database 106 can include material information 124.

[0017] The electronic device 102 may include suitable logic, circuitry, interfaces, and / or code configured to receive a dataset that may contain information related to scientific literature. Also, the electronic device 102 can apply one or more neural network models 110 to the dataset to determine a set of materials 116 and information related to each material in the set of materials 116. Further, the electronic device 102 can generate a first set of embeddings 118 indicating a first set of characteristics of each material in the set of materials 116 and a second set of embeddings 120 related to text content describing the impact of the set of materials 116 on the resources of the living environment. Then, the electronic device 102 can train a trained generation AI model 114 based on the first set of embeddings 118 and the second set of embeddings 120. Further, the electronic device 102 can receive a user input that may indicate information related to the queried material 122. The electronic device 102 can generate a third embedding that may indicate a second set of characteristics of the queried material 122 and a fourth embedding that may be related to text content describing the impact of the queried material 122 on the resources of the living environment. The electronic device 102 can further apply the trained generation AI model 114 to the third embedding and the fourth embedding, determine sustainability information related to the queried material 122 based on the application of the generation AI model 114, and control a display device to render the sustainability information related to the queried material 122. Examples of the electronic device 102 may include, but are not limited to, computing devices, smartphones, mainframe machines, servers, consumer electronics (CE) devices, computer workstations, and / or devices having graphics processing capabilities (e.g., devices having a set of graphics processing units (GPUs), etc.).

[0018] Server 104 may include suitable logic, circuitry, and interfaces, and / or code configured to receive requests for datasets from electronic device 102. Server 104 may further be configured to retrieve a dataset from database 106 and transmit the dataset to electronic device 102. In at least one embodiment, server 104 may apply one or more neural network models 110 (stored in server 104) to a dataset (retrieved from server 104) to determine a set of materials 116 and information associated with each material in the set of materials 116. Server 104 may further be configured to generate a first set of embeddings 118 and a second set of embeddings 120 based on the set of materials 116 and the information associated with each material in the set of materials 116. Thereafter, server 104 may transmit the first set of embeddings 118 and the second set of embeddings 120 to electronic device 102. In some embodiments, server 104 may be configured to receive the first set of embeddings 118, the second set of embeddings 120, and a queried material 122 from electronic device 102. Server 104 may further be configured to train a generated AI model 114 (stored in server 104) based on the generated first set of embeddings 118 and the generated second set of embeddings 120. Thereafter, server 104 may use the generated AI model 114 to determine sustainability information associated with the queried material 122 and transmit the determined sustainability information to electronic device 102. Server 104 can be implemented as a cloud server and can execute processes through, for example, web applications, cloud applications, Hypertext Transfer Protocol (HTTP) requests, repository operations, and file transfers. Other examples of implementations of server 104 include, but are not limited to, database servers, file servers, web servers, media servers, application servers, mainframe servers, cloud computing servers, and / or any device having graph processing capabilities (e.g., a device having a set of Graphics Processing Units (GPUs)).

[0019] In at least one embodiment, server 104 can be implemented as a plurality of distributed cloud-based resources by using some techniques that may be well known to those skilled in the art. It can be understood by those skilled in the art that the scope of the present disclosure may not be limited to the implementation of server 104 and electronic device 102 as two separate entities. In certain embodiments, the functions of server 104 can be incorporated into electronic device 102 in whole or at least partially without departing from the scope of the present disclosure.

[0020] Database 106 can include suitable logic, circuitry, interfaces, and / or code configured to store material information 124. Material information 124 can correspond to a dataset that may include information related to scientific literature. Database 106 can be derived from data in a relational or non-relational database, or a set of comma-separated value (csv) files in conventional storage or big data storage. Database 106 can be stored or cached on a device such as, for example, server 104 or electronic device 102. The device storing database 106 can be configured to receive queries about material information 124. In response, the device storing database 106 can be configured to retrieve material information 124 and transmit it to electronic device 102. According to one embodiment, database 106 may be stored at the same or different locations and hosted on multiple servers. The operation of database 106 can be performed using hardware including a processor, a microprocessor (e.g., for executing or controlling the execution of one or more operations), a field programmable gate array (FPGA), or an application specific integrated circuit (ASIC). In some other examples, database 106 may be implemented using software.

[0021] The communication network 108 may include a communication medium through which the electronic device 1 (102) and the server 104 can communicate with each other. The communication network 108 can be either a wired connection or a wireless connection. Examples of the communication network 108 include, but are not limited to, the Internet, a cloud network, a cellular or wireless mobile network (such as Long-Term Evolution and 5G New Radio, etc.), a satellite network (such as a network of a set of low-earth orbit satellites, etc.), a Wireless Fidelity (Wi-Fi (registered trademark)) network, a Personal Area Network (PAN), a Local Area Network (LAN), or a Metropolitan Area Network (MAN). Various devices within the network environment 100 can be configured to connect to the communication network 108 according to various wired and wireless communication protocols. Examples of the wired and wireless communication protocols include, but are not limited to, TCP / IP (Transmission Control Protocol and Internet Protocol), UDP (User Datagram Protocol), HTTP (Hypertext Transfer Protocol), FTP (File Transfer Protocol), Zig Bee (registered trademark), EDGE, IEEE 802.11, Li-Fi (Light Fidelity), 802.16, IEEE 802.11s, IEEE 802.11g, multi-hop communication, a wireless access point (AP), device-to-device communication, a cellular communication protocol, and at least one of the Bluetooth (registered trademark) (BT) communication protocols.

[0022] According to one embodiment, each of one or more neural network models 110, a natural language model 112, and a generative AI model 114 (e.g., including each of a generator model 114A and a discriminator model 114B) can be referred to as a neural network model. The neural network model can be a computational network or system of artificial neurons arranged in multiple layers. The neural network model can be defined by its hyperparameters such as, for example, (one or more) activation functions, number of weights, cost function, regularization function, input size, and number of layers. Also, those layers can include an input layer, one or more hidden layers, and an output layer. Each layer of the multiple layers can include one or more nodes (or artificial neurons). The output of all nodes in the input layer can be coupled to at least one node of (one or more) hidden layers. Similarly, the input of each hidden layer can be coupled to the output of at least one node in another layer of the neural network. The output of each hidden layer can be coupled to the input of at least one node in another layer of the neural network model. The (one or more) nodes in the final layer can receive input from at least one hidden layer and output a result. The number of layers and the number of nodes in each layer can be determined from the hyperparameters of the neural network model. Such hyperparameters can be set before or after training of the neural network model.

[0023] Each node may correspond to a mathematical function (e.g., sigmoid function or rectified linear unit) having parameters adjustable during the training of the neural network model. The set of parameters may include, for example, weight parameters and regularization parameters. Each node may calculate an output based on one or more inputs from nodes within other (one or more) layers (e.g., (one or more) previous layers) of the neural network model using the mathematical function. All or some of the nodes of the neural network model may correspond to the same or different mathematical functions. In the training of the neural network model, one or more parameters of each node of the neural network model may be updated based on whether the output of the final layer for a given input (from the training data set) matches the correct result according to the loss function of the neural network model. The above process may be repeated for the same or different inputs until the minimum of the loss function is achieved and the training error is minimized. Some partial methods for training, such as gradient descent, stochastic gradient descent, batch gradient descent, gradient boosting, and metaheuristics, are known in the art.

[0024] In some embodiments, the neural network model may include electronic data that can be implemented as a software component of an application executable on the electronic device 102. The neural network model may rely on a library, an external script, or logic / instructions for execution by a processing device included in the electronic device 102. In at least one embodiment, the neural network model may be implemented using hardware that may include a processor, a microprocessor (e.g., for performing or controlling the execution of one or more operations), an FPGA, or an ASIC. Alternatively, in some embodiments, the neural network model may be implemented using a combination of hardware and software. Examples of neural network models may include, but are not limited to, deep neural networks (DNNs), convolutional neural networks (CNNs), artificial neural networks (ANNs), fully connected neural networks, and / or combinations of such networks.

[0025] One or more neural network models 110 may be applied to a received dataset that may include information related to scientific literature (or material information 124). In at least one embodiment, the received dataset (or material information 124) may include articles and journals related to technical fields such as, for example, the field of materials science and the field of chemical science. The received dataset or material information 124 may include one or more of entities, properties of the entities, and interactions that may occur between the entities. One or more neural network models 110 may be capable of receiving the dataset (or material information 124) as input. One or more neural network models 110 may be trained to recognize named entities and extract information related to the recognized named entities from the dataset or material information 124. The named entities to be recognized may include reactants, products, and catalysts, etc. Reactants may interact with each other for the production of products. The interaction between reactants may be included as a chemical reaction in scientific literature. The chemical reaction may be triggered under specific environmental conditions (e.g., temperature, pH level, or pressure, etc.) based on the properties of the reactants. The chemical reaction may be facilitated by a catalyst and may involve one or more precursors. The information related to the recognized named entities (extracted by the neural network) may include the properties of each reactant and catalyst, the properties of the products that may be produced based on the chemical reactions in which the reactants are involved, the properties of the chemical reactions in which the recognized named entities (i.e., reactants and catalysts) are involved, and / or the conditions under which the chemical reaction may be triggered.

[0026] One or more neural network models 110 can also be trained to filter named entities based on whether the named entity is sustainable (environmentally friendly or harmless) or non-sustainable (environmentally bad or harmful). In at least one embodiment, the filtering of the recognized named entities (i.e., reactants) can be based on the results (i.e., products) that can be produced by the interactions (i.e., chemical reactions) between the recognized named entities and the conditions under which the interactions can be triggered. For example, if the product produced is environmentally bad or harmful, the reactants that promote the production of that product (by the relevant interactions) can be filtered out. On the other hand, if the reactants involved in a chemical reaction lead to the production of a sustainable product, such reactants can be identified as useful or important.

[0027] The natural language model 112 may receive, as input, a set of materials 116 and information related to the set of materials 116. The set of materials 116 may include both sustainable materials and non-sustainable materials. Each material in the set of materials 116 may include one or more reactants. Also, the information related to each material in the set of materials 116 may be obtained based on recognized named entities, i.e., information related to the one or more reactants that make up the corresponding material. The natural language model 112 may be trained to generate, as output, a first set of embeddings 118 and a second set of embeddings 120. The first set of embeddings 118 may include embeddings generated for each material in the set of materials 116. The embeddings may indicate the characteristics of the corresponding material. The second set of embeddings 120 may include embeddings that may be generated based on the information related to each material in the set of materials 116. The embeddings may indicate the impact of the set of materials 116 on the environment. Examples of the natural language model 112 may include, but are not limited to, the BERT (Bidirectional Encoder Representations from Transformers) model, the GPT (Generative Pre-trained Transformers) model, the RoBERTa (Robustly Optimized BERT Pretraining Approach) model, or a large language model (LLM).

[0028] The generative AI model 114 (i.e., each of the generator model 114A and the discriminator model 114B) can be trained based on the first set of embeddings 118, the second set of embeddings 120, and information related to the set of materials 116. The discriminator model 114B can be trained using embeddings related to both sustainable and non-sustainable materials included in the set of materials 116. This training can be such that the discriminator model 114B can classify whether the output generated by the generator model 114A is related to sustainable materials. The generator model 114A can be trained to generate an output for the queried material 122 such that the discriminator model 114B can accurately predict whether the queried material 122 is sustainable. Thus, based on the training, the generative AI model 114 can be configured to predict whether the queried material 122 is sustainable. Examples of the generative AI model 114 can include, but are not limited to, a Generative Adversarial Network (GAN) model, a Variational Autoencoder (VAE) model, an autoregressive model, a Generative Pre-trained Transformers (GPT) model, or a Large Language Model (LLM).

[0029] In operation, the electronic device 102 can be configured to receive a dataset that can include information related to scientific literature. According to one embodiment, the dataset can be received from the server 104 or the database 106 (via the server 104). In such an embodiment, the dataset can correspond to the material information 124. The scientific literature can be related to a technical field such as, for example, the field of materials science or chemistry. The information related to the scientific literature can include technical content. Details regarding the reception of the dataset are further described, for example, in FIG. 3 (302).

[0030] The electronic device 102 may further be configured to apply one or more neural network models 110 to a dataset. The one or more neural network models 110 may include a first neural network model and a second neural network model. Based on receiving a dataset as input (i.e., information related to scientific literature), the first neural network model may generate an output. The output may include named entities such as reactants, products, catalysts, etc. that may be involved in chemical reactions triggered in different situations. The output may further include information related to each of the named entities. The named entities, and the information related to each of the named entities, can be extracted from the information related to scientific literature.

[0031] The second neural network model may receive, as input, the extracted named entities and the extracted information related to each of the named entities. The second neural network model may generate a classification result indicating whether the named entity is sustainable or non - sustainable. This classification may be based on the information related to each of the named entities.

[0032] The electronic device 102 may further be configured to determine a set of materials 116 and information related to each material in the set of materials 116. This determination can be based on the application of one or more neural network models 110 to the dataset. Each material in the set of materials 116 may include one or more named entities extracted by the first neural network model. The information related to each material in the set of materials 116 can be determined based on the information related to the one or more named entities (extracted by the first neural network model) that may be included in the corresponding material.

[0033] According to one embodiment, the determined set of materials 116 can include sustainable materials and non-sustainable materials. The sustainable materials can include one or more sustainable named entities (so classified by the second neural network model), while the non-sustainable materials can include one or more non-sustainable named entities (so classified by the second neural network model). Details regarding the determination of the set of materials and the material information are further described, for example, in FIG. 3 (304).

[0034] The electronic device 102 can further be configured to generate a first set of embeddings 118 that can indicate a first set of features of each material of the set of materials 116. According to one embodiment, the generation of the first set of embeddings 118 can be based on the application of the natural language model 112 to the set of materials 116. Each embedding of the first set of embeddings 118 can be generated for each material of the set of materials 116 and can indicate a first set of features for the corresponding material.

[0035] The electronic device 102 may further be configured to generate a second set 120 of embeddings that may be related to text content describing the impact of the set 116 of materials on the resources of the living environment. According to one embodiment, the generation of the second set 120 of embeddings may be based on the application of the natural language model 112 to the information related to each material of the set 116 of materials. Each embedding of the second set 120 of embeddings may be generated for each material of the set 116 of materials and may indicate the impact of the corresponding material on the resources of the living environment. Among the embeddings of the second set 120 of embeddings, some generated for the sustainable materials (which may include one or more named sustainable entities) included in the set 116 of materials may indicate the compatibility of each of the sustainable materials with the environment. On the other hand, among the embeddings of the second set 120 of embeddings, other embeddings generated for the non-sustainable materials (which may include one or more named non-sustainable entities) included in the set 116 of materials may indicate the incompatibility of each of the non-sustainable materials with the environment, or the harmful impact of each of the non-sustainable materials on the environment. Details regarding the generation of embeddings (e.g., including the first set of embeddings and the second set of embeddings) will be further described, for example, in FIG. 3 (306).

[0036] The electronic device 102 may further be configured to train a generative AI model 114 based on the first set 118 of embeddings and the second set 120 of embeddings. The generative AI model 114, i.e., the generator model 114A and the discriminator model 114B, may be trained to predict whether a particular material is sustainable based on the information related to the material. This prediction may be based on the first set 118 of embeddings and the second set 120 of embeddings. The generative AI model 114 may further generate information indicating the basis behind the prediction.

[0037] According to one embodiment, the generator model 114A can be trained to generate an output about a material based on a first set of embeddings 118 and a second set of embeddings 120. The output can be such that the discriminator model 114B classifies the material (for which the output was generated) as sustainable (i.e., environmentally friendly). Similarly, the discriminator model 114B is trained based on the first set of embeddings 118 and the second set of embeddings 120 such that the discriminator model 114B receives an output (generated by the generator model 114A for that material) and is trained to accurately classify the material as sustainable (i.e., environmentally friendly) or non - sustainable (i.e., environmentally harmful). Details regarding the training of the generative AI model are further described, for example, in FIG. 3 (308).

[0038] The electronic device 102 can further be configured to receive a user input that can indicate information related to the queried material (such as the queried material 122, etc.). The received user input can include instructions for determining whether the queried material 122 is sustainable. Using the information indicated in the received user input, the characteristics related to the queried material 122 can be determined. Based on the determined characteristics, one or more features of the queried material 122 can be determined, and the impact of the queried material 122 on the environment can also be determined. Note that the queried material 122 can be a material among the set of materials 116. Details regarding the reception of the user input are further described, for example, in FIG. 3 (310).

[0039] The electronic device 102 may further be configured to generate a third embedding indicating a second set of features of the queried material 122 and a fourth embedding related to text content describing the impact of the queried material 122 on the resources of the living environment. The third embedding and the fourth embedding can be generated by using the natural language model 112. According to one embodiment, the queried material 122 may be supplied as an input to the natural language model 112. Based on the application of the natural language model 112 to the queried material 122, a second set of features of the queried material 122 may be generated as an output of the natural language model 112 (shown in the third embedding). Further, the application of the natural language model 112 to information related to the queried material 122 may generate a fourth embedding related to text content describing the impact of the queried material 122 on the resources of the living environment. The third embedding and the fourth embedding can be used to determine the sustainability of the queried material 122. Details regarding the generation of embeddings for the queried material (e.g., including the third embedding and the fourth embedding) will be further described, for example, in FIG. 3 (312).

[0040] The electronic device 102 may further be configured to apply the trained generative AI model 114 to the third embedding and the fourth embedding. The third embedding and the fourth embedding may be supplied as inputs to the generative AI model 114. Additionally, the generator model 114A may receive, as inputs, input noise and received user input, i.e., information related to the queried material 122. Based on the input, the generator model 114A can generate an output for the queried material 122. The discriminator model 114B may receive, as inputs, (in addition to the generated third embedding and the generated fourth embedding) the output generated by the generator model 114A for the queried material 122.

[0041] The electronic device 102 may further be configured to determine sustainability information related to the queried material 122 (based on the application of the generative AI model 114). The sustainability information may include at least one of a classification result obtained as an output of the discriminator model 114B and an explanation of the classification result. The output of the discriminator model 114B (i.e., the classification result) may correspond to the prediction of the generative AI model 114. The classification result can indicate whether the queried material 122 is classified as sustainable or non-sustainable. The discriminator model 114B can classify the queried material 122 as sustainable when the queried material 122 is likely to be compatible with or friendly to the environment. On the other hand, the discriminator model 114B can classify the queried material 122 as non-sustainable when the queried material 122 is unlikely to be compatible with or harmful to the environment. Also, the explanation of the classification result can indicate the basis behind the classification of the queried material 122 as sustainable or non-sustainable. Details regarding the determination of the sustainability information are further described, for example, in FIG. 3 (314).

[0042] The electronic device 102 may further be configured to control a display device to render sustainability information related to the queried material 122. According to one embodiment, the electronic device 102 can send a control command to the display device, as a result of which the display device can receive the control command from the electronic device 102 and render the content that can be received from the electronic device 102. The display device can receive information including the sustainability information (e.g., the output of the discriminator model 114B) and control commands for rendering the sustainability information on the screen of the display device. Upon receiving the information, the received sustainability information related to the queried material 122 can be rendered on the display screen of the display device.

[0043] FIG. 1 may be modified, added to, or omitted without departing from the scope of the present disclosure. For example, the network environment 100 may include more or fewer elements than those illustrated and described in the present disclosure. In some embodiments, the functions of each of the server 104 and the database 106 may be incorporated into the electronic device 102 without departing from the scope of the present disclosure.

[0044] FIG. 2 is a block diagram showing an exemplary electronic device for AI-based sustainable material design according to at least one embodiment described in the present disclosure. FIG. 2 will be described in relation to the elements from FIG. 1. Referring to FIG. 2, a block diagram 200 of a system 202 including the electronic device 102 is shown. The electronic device 102 may include a processor 204, a memory 206, a persistent data storage 208, an input / output (I / O) device 210, and a network interface 212. In at least one embodiment, the memory 206 may store one or more neural network models 110, a natural language model 112, and a generative AI model 114. In at least one embodiment, the I / O device 210 may include a display device 210A.

[0045] Processor 204 may include suitable logic, circuitry, and interfaces configured to execute a set of instructions stored in memory 206. Processor 204 may be configured to execute program instructions related to various operations performed by electronic device 102. Processor 204 may be configured to receive a dataset that may include information related to scientific literature. Processor 204 may further be configured to apply one or more neural network models 110 to the dataset. Processor 204 may further be configured to determine a set of materials 116 and information related to each material in the set of materials 116 based on the application of one or more neural network models 110. Processor 204 may further be configured to generate a first set of embeddings 118 that may indicate a first set of characteristics of each material in the set of materials 116. Processor 204 may further be configured to generate a second set of embeddings 120 that may be related to text content describing the impact of the set of materials 116 on the resources of the living environment. Processor 204 may further be configured to train a generative AI model 114 based on the first set of embeddings 118 and the second set of embeddings 120. Processor 204 may further be configured to receive user input indicating information related to a queried material 122. Processor 204 may further be configured to generate a third embedding that may indicate a second set of characteristics of the queried material 122, and may further be configured to generate a fourth embedding that may be related to text content describing the impact of the queried material 122 on the resources of the living environment. Processor 204 may further be configured to apply the generative AI model 114 to the third embedding and the fourth embedding. Processor 204 may further be configured to determine sustainability information related to the queried material 122 based on the application of the generative AI model 114. Processor 204 may further be configured to control display device 210A to render the sustainability information related to the queried material 122. Processor 204 may be implemented based on several processor technologies known in the art.Examples of processor technologies can include, but are not limited to, a central processing unit (CPU), an X86-based processor, a reduced instruction set computing (RISC) processor, an application specific integrated circuit (ASIC) processor, a complex instruction set computing (CISC) processor, a graphics processing unit (GPU), a coprocessor, or a combination thereof.

[0046] Although shown as a single processor in FIG. 2, processor 204 can include any number of processors configured to individually or collectively perform or direct the performance of any number of operations of the electronic device 102 described in this disclosure. Additionally, one or more of the processors may be present on one or more different electronic devices, such as different servers, for example. In at least one embodiment, processor 204 can be configured to interpret and / or execute program instructions and process data that may be stored in memory 206 or persistent data storage 208. In some embodiments, processor 204 can be configured to fetch program instructions from persistent data storage 208 and load the program instructions into memory 206. After the program instructions are loaded into memory 206, processor 204 can execute the program instructions.

[0047] Memory 206 may include suitable logic, circuitry, and interfaces configured to store one or more instructions to be executed by processor 204. Execution of the one or more instructions stored in memory 206 by processor 204 may perform various operations of processor 204 (and electronic device 102). Memory 206 may store a dataset (e.g., material information 124) that may include information related to scientific literature, information related to each material of material set 116 and material set 116, first embedded set 118, second embedded set 120, third embedding, fourth embedding, and sustainability information related to queried material 122. Examples of implementations of memory 206 may include, but are not limited to, CPU cache, hard disk drive (HDD), solid state drive (SSD), random access memory (RAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), and / or secure digital (SD) card.

[0048] The persistent data storage 208 may include suitable logic, circuitry, and / or interfaces configured to store program instructions executable by the processor 204. The persistent data storage 208 may include a computer-readable storage medium carrying or storing computer-executable instructions or data structures. Such a computer-readable storage medium may include any available medium that can be accessed by a general-purpose or special-purpose computer, such as, for example, the processor 204. By way of example and not limitation, such a computer-readable storage medium may include compact disc read-only memory (CD-ROM) or other optical disk storage, magnetic disk storage or other magnetic storage devices (e.g., hard disk drive (HDD)), flash memory devices (e.g., solid state drive (SSD), secure digital (SD) card, other solid state memory devices), or any other storage medium that can be used to carry or store particular program code in the form of computer-executable instructions or data structures and that can be accessed by a general-purpose or special-purpose computer, including tangible or non-transitory computer-readable storage media. Combinations of the above may also be included within the scope of computer-readable storage media. The computer-executable instructions may include, for example, instructions and data configured to cause the processor 204 to perform specific operations or groups of operations associated with the electronic device 102.

[0049] The I / O device 210 may include suitable logic, circuitry, and interfaces configured to receive inputs and render outputs based on the received inputs. For example, the I / O device 210 may receive an input that can trigger the reception of a dataset including information related to scientific literature. The I / O device 210 may further receive a user input indicating information related to the queried material 122. Additionally, the I / O device 210 may render outputs such as, for example, the determined set of materials 116, information related to each material in the set of materials 116, and sustainability information related to the queried material 122. The I / O device 210, which may include various input and output devices, may be configured to communicate with the processor 204. Examples of the I / O device 210 may include, but are not limited to, a touch screen, a keyboard, a mouse, a joystick, the display device 210A, a microphone, and a speaker.

[0050] The display device 210A may include suitable logic, circuitry, and interfaces configured to render sustainability information related to the queried material 122. The display device 210A may be a touch screen that enables a user to provide user input via the display device 210A. The touch screen may be at least one of a resistive touch screen, a capacitive touch screen, or a thermal touch screen. The display device 210A may be implemented through several known technologies such as, but not limited to, for example, a liquid crystal display (LCD) display, a light emitting diode (LED) display, a plasma display, or an organic LED (OLED) display technology, or other display devices. According to one embodiment, the display device 210A may refer to the display screen of a head-mounted device (HMD), a smart glass device, a see-through display, a projection-based display, an electrochromic display, or a transparent display.

[0051] The network interface 212 may include suitable logic, circuitry, and interfaces configured to facilitate communication between the processor 204 (i.e., the electronic device 102) and the server 104 via the communication network 108. The network interface 212 may be implemented by using various known techniques for supporting wired or wireless communication of the electronic device 102 with the communication network 108. The network interface 212 may include, but is not limited to, an antenna, a radio frequency (RF) transceiver, one or more amplifiers, a tuner, one or more oscillators, a digital signal processor, a coder-decoder (CODEC) chipset, a subscriber identity module (SIM) card, or a local buffer circuit. The network interface 212 may be configured to communicate wirelessly with a network such as the Internet, an intranet, or a wireless network such as, for example, a cellular telephone network, a wireless local area network (LAN), and a metropolitan area network (MAN). The wireless communication may be configured to use one or more of a plurality of communication standards, protocols, and technologies such as, for example, EDGE (Enhanced Data GSM Environment), wideband code division multiple access (W-CDMA), LTE (Long Term Evolution), 5G (5th Generation), NR (New Radio), GSM (registered trademark) (Global System for Mobile Communications), code division multiple access (CDMA), time division multiple access (TDMA), Bluetooth (registered trademark), Wi-Fi (Wireless Fidelity) (e.g., IEEE 802.11a, IEEE 802.11b, IEEE 802.11g, or IEEE 802.11n, etc.), VoIP (Voice over Internet Protocol), Li-Fi (light fidelity), Wi-MAX (Worldwide Interoperability for Microwave Access), a protocol for electronic mail, instant messaging, and SMS (Short Message Service).

[0052] Examples of the electronic device 102 may be modified, added, or omitted without departing from the scope of the present disclosure. For example, in some embodiments, examples of the electronic device 102 may include any number of other components that may not be explicitly illustrated or described for the sake of brevity.

[0053] FIG. 3 is a diagram showing an exemplary execution pipeline for AI-based sustainable material design according to an embodiment of the present disclosure. FIG. 3 will be described in relation to the elements from FIGS. 1 and 2. Referring to FIG. 3, an execution pipeline 300 is shown. The exemplary execution pipeline 300 may include a series of operations that may be performed by the processor 204 of the electronic device 102 in FIG. 1 for the discovery, design, and synthesis of sustainable materials, i.e., materials that are environmentally compliant or friendly. A series of operations that may start at 302 and end at 314 are shown in the execution pipeline 300.

[0054] At 302, a dataset 302A including information related to scientific literature may be received. In at least one embodiment, the processor 204 may be configured to receive a dataset 302A that may include information related to scientific literature. The information related to scientific literature can be in text format or multimedia format (e.g., images, audio, or video, etc.). The scientific literature may be related to the field of materials science, biology, chemistry, biochemistry, or environmental science. For example, the received dataset 302A may constitute information included in articles, textbooks, books, and presentations, etc. published by scientific or engineering societies on topics such as, for example, materials science, chemistry, environmental science, or combinations thereof.

[0055] At 304, based on the received data set 302A, a set of materials 304A and material information 304B related to the set of materials 304A can be determined. In at least one embodiment, the processor 204 can be configured to determine the set of materials 304A and the material information 304B related to the set of materials 304A based on the received data set 302A. The material information 304B can include information related to each material of the set of materials 304A determined from the received data set 302A. The determination of the set of materials 304A and the material information 304B can be based on the application of a neural network-based analysis to the data set 302A (i.e., information related to scientific literature).

[0056] According to one embodiment, the processor 204 can apply one or more neural network models 110 to the received data set 302A for the determination of the set of materials 304A and the material information 304B. For example, the processor 204 can apply a first neural network model among the one or more neural network models 110 to the received data set 302A. The first neural network model can be a language model configured to perform a named entity recognition task upon receiving the data set 302A. The named entity recognition task can include identifying named entities from the information related to the scientific literature included in the received data set 302A. As an output of the first neural network model, named entities can be generated. The identified named entities can include one or more of reactants, catalysts, and products. Accordingly, the processor 204 can identify a set of reactants based on the application of the first neural network model to the received data set 302A. The processor 204 can further identify a set of products and a set of catalysts.

[0057] The named entity recognition task can further include extracting information related to each of the named entities. Information related to a named entity can include characteristics of the named entity, interactions between named entities, and conditions under which interactions between named entities can be triggered. Interactions can include chemical reactions that can occur between reactants of a set of reactants. The occurrence of a chemical reaction can be facilitated by catalysts of a set of catalysts, and the chemical reaction can lead to the production of products of a set of products. A chemical reaction may be triggered in a particular scenario or condition that has a particular impact on the environment. Thus, information related to a named entity can include characteristics of each reactant of a set of reactants, characteristics of each product of a set of products, and characteristics of each catalyst of a set of catalysts. Information related to a named entity can further include a chemical reaction in which a set of reactants is involved, a set of products produced by the chemical reaction, a set of catalysts that facilitate the chemical reaction, and conditions under which the chemical reaction is triggered.

[0058] Accordingly, the processor 204 can extract information related to each identified reactant of the identified set of reactants based on the application of the first neural network model to the received dataset 302A. Information related to each reactant of the identified set of reactants can include the organic structure of the corresponding reactant, the decay rate associated with the corresponding reactant, the biodegradability associated with the corresponding reactant, one or more catalysts that can facilitate a chemical reaction in which the corresponding reactant is involved, one or more products produced by the chemical reaction, temperature requirements for triggering the chemical reaction, or one or more precursors that can be involved in the chemical reaction. The processor 204 can also extract information related to each product of the set of products and information related to each catalyst of the set of catalysts. Information related to each product of the set of products can include the organic structure, decay rate, or biodegradability of the corresponding product.

[0059] According to one embodiment, the processor 204 can apply a second neural network model to an input that includes a specified named entity and extracted information related to the specified named entity. The second neural network model can be a language-based classifier model that can be trained to perform a classification task. The classification task can include classifying a named entity as sustainable or harmful based on the information related to the named entity. A named entity can be classified as sustainable if the named entity is environmentally friendly or compatible with the environment. On the other hand, a named entity can be classified as harmful if the named entity is not compatible with the environment or appears to be harmful to the environment. To determine whether the reactants of a specified set of reactants are sustainable, the processor 204 can apply the second neural network model among one or more neural network models 110 to the specified set of reactants. Based on the application of the second neural network model, the processor 204 can select a subset of reactants from the specified set of reactants. The selection of the subset of reactants can be based on the classification of each reactant in the subset of reactants as sustainable. The selected subset of reactants can be referred to as a subset of sustainable reactants. Other reactants (i.e., reactants not included in the selected subset of reactants) in the specified set of reactants can be classified as harmful. These reactants can constitute a subset of harmful reactants.

[0060] For example, a reactant in a specified set of reactants can be classified as harmful or non-sustainable if the reactant is not biodegradable, has a low decay rate (or a high half-life), and is involved in a chemical reaction that can lead to the production of environmentally harmful products. On the other hand, a reactant in a specified set of reactants can be classified as sustainable if the reactant is biodegradable or is involved in a chemical reaction that leads to the production of sustainable or desirable products.

[0061] According to one embodiment, the set of materials 304A can be determined based on the identification of the set of reactants and the selection of a subset of the reactants. Each material in the set of materials 304A can include one or more of the reactants in the set of reactants. Also, the set of materials 304A can include a subset of sustainable materials and a subset of harmful materials. Each sustainable material in the subset of sustainable materials can include one or more of the selected subset of reactants (the subset of sustainable reactants). However, each harmful material in the subset of harmful materials can include one or more of the reactants classified as harmful among the identified set of reactants. In some embodiments, the harmful materials can be filtered out and the subset of sustainable materials can be determined to be important or useful.

[0062] According to one embodiment, the material information 304B can include information related to each material in the set of materials 304A. Thus, the material information 304B can include information related to each material in the subset of sustainable materials and information related to each material in the subset of harmful materials. The information related to each material in the set of materials 304A can be determined based on the extracted information related to each of the one or more reactants (among the identified set of reactants) that can be included in the corresponding material.

[0063] At 306, an embedding related to the determined set of materials 304A can be generated. In at least one embodiment, the processor 204 can be configured to generate an embedding related to the determined set of materials 304A. The generated embedding can include a first set of embeddings 306A and a second set of embeddings 306B. Each embedding in the first set of embeddings 306A and each embedding in the second set of embeddings 306B can be generated for each material in the set of materials 304A. The first set of embeddings 306A can be generated based on the application of the natural language model 112 to the determined set of materials 304A. The first set of embeddings 306A can represent a set of features of each material in the determined set of materials 304A. The second set of embeddings 306B can be generated based on the application of the natural language model 112 to the material information 304B. The second set of embeddings 306B can be related to the text content describing the impact of the determined set of materials 304A on the resources of the living environment. Examples of the natural language model 112 can include, but are not limited to, the BERT (Bidirectional Encoder Representations from Transformers) model, the GPT (Generative Pre-trained Transformers) model, the RoBERTa (Robustly Optimized BERT Pretraining Approach) model, or a large language model (LLM).

[0064] According to one embodiment, the processor 204 can apply the natural language model 112 to each material in the determined set of materials 304A. Based on this application, an embedding of a first set of embeddings 306A for the corresponding material can be generated. The generated embeddings can indicate the characteristics of the corresponding material. The characteristics can indicate whether the corresponding material is sustainable or harmful based on the classification of one or more reactants constituting the corresponding material as sustainable or harmful. The processor 204 can further apply the natural language model 112 to the information related to each material in the determined set of materials 304A. Based on this application, an embedding of a second set of embeddings 306B for the corresponding material can be generated. The generated embeddings can indicate the impact of the corresponding material on the environment. The impact can indicate whether the corresponding material is useful (e.g., when the efficiency of the product containing the corresponding material is improved), or harmful to the environment (e.g., when the corresponding material releases toxic gases under certain conditions or reduces the efficiency of the product that may contain the corresponding material).

[0065] At 308, the AI model 114 generated based on the first set of embeddings 306A and the second set of embeddings 306B can be trained. In at least one embodiment, the processor 204 can be configured to train the AI model 114 generated based on the first set of embeddings 306A and the second set of embeddings 306B. Examples of the generative AI model 114 can include, but are not limited to, a Generative Adversarial Network (GAN) model, a Variational Autoencoder (VAE) model, an autoregressive model, a Generative Pre-trained Transformers (GPT) model, or a Large Language Model (LLM). After training, the generative AI model 114 can predict sustainability information related to materials and generate an explanation including the basis behind the prediction. In one embodiment, the generative AI model 114 can correspond to a conditional generative adversarial network (GAN) model including a generator model 114A and a discriminator model 114B. In one embodiment, the generator model 114A can be trained simultaneously while the discriminator model 114B is in an idle state (i.e., not being trained). Also, the discriminator model 114B can be trained before or after the training of the generator model 114A such that the generator model 114A can be in an idle state (i.e., not being trained) during the training time interval of the discriminator model 114B. The generator model 114A can be trained to generate an output for each material in the set of materials 304A such that the discriminator model 114B classifies each material in the set of materials 304A as sustainable. On the other hand, the discriminator model 114B can be trained to accurately classify each material in the set of materials 304A as sustainable or harmful.

[0066] According to one embodiment, the processor 204 can train the generator model 114A based on information related to each material of the first set of embeddings 306A, the second set of embeddings 306B, and the set of materials 304A (i.e., the material information 304B). The generator model 114A can be further trained based on random input and generator loss. The random input can be random noise (e.g., white noise) generated using a Gaussian noise model. For the materials in the set of materials 304A, the generator model 114A can be applied to the random input, the first embedding generated for the material in the first set of embeddings 306A, and the second embedding generated for the material in the second set of embeddings 306B. Based on the application of the generator model 114A, an output for the material can be generated. For the generation of the generator loss, the discriminator model 114B can be applied to the generated output. The generator model 114A can be further trained based on the generator loss to generate an updated output for the material, and the application of the discriminator model 114B to the updated output can result in the minimization of the generator loss and the classification of the material as sustainable. The generator model 114A can be similarly trained to generate an output for each of the other materials in the set of materials 304A.

[0067] Similarly, the processor 204 can train the discriminator model 114B based on information related to each material of the first set of embeddings 306A, the second set of embeddings 306B, and the set of materials 304A (i.e., the material information 304B). The discriminator model 114B can be further trained based on the output that can be generated by the generator for each material of the set of materials 304A and the discriminator loss. The discriminator model 114B is applied to the output generated by the generator model 114A for the material (from the set of materials 304A), the first embedding generated for the material (from the first set of embeddings 306A), and the second embedding generated for the material (from the second set of embeddings 306B), and the discriminator model 114B can generate a classification result as the output. The classification result can be a value indicating whether the material is sustainable or harmful and the degree to which the material is sustainable or harmful. Based on the classification result, the discriminator model 114B can generate a discriminator loss. The discriminator model can be further trained based on the discriminator loss so that the material is accurately classified as sustainable or harmful. The discriminator model 114B can be similarly trained to accurately classify each of the other materials of the set of materials 304A as sustainable or harmful.

[0068] At 310, a user input 310A indicating information related to the queried material can be received. In at least one embodiment, the processor 204 can be configured to receive a user input 310A indicating information related to the queried material. The reception of the user input 310A can be based on requirements for determining whether the queried material is sustainable or harmful. The information related to the queried material can be extracted from scientific literature using a language-based model (such as a first neural network model). The queried material can include one or more reactants, or a combination of reactants and products. In some embodiments, the queried material can be a material among the set of materials 304A.

[0069] At 312, an embedding related to the queried material can be generated based on the information related to the queried material. In at least one embodiment, the processor 204 can be configured to generate an embedding related to the queried material based on the information related to the queried material. The generated embedding can include a third embedding and a fourth embedding. The third embedding can indicate a set of features of the queried material, and the fourth embedding can be related to text content describing the impact of the queried material on the resources of the living environment.

[0070] According to one embodiment, the processor 204 can apply the natural language model 112 to the queried material for generating the third embedding. The processor 204 can further apply the natural language model 112 to the information related to the queried material for generating the fourth embedding.

[0071] In some embodiments, the queried material can be a material among the set of materials 304A. In such embodiments, the generated third embedding can be an embedding among the set of first embeddings 306A, and the generated fourth embedding can be an embedding among the set of second embeddings 306B.

[0072] At 314, based on the third embedding and the fourth embedding, sustainability information related to the queried material can be determined. In at least one embodiment, the processor 204 can be configured to determine sustainability information related to the queried material based on the third embedding and the fourth embedding. The sustainability information related to the queried material can correspond to a first indication that defines whether the queried material is sustainable or harmful, and a second indication that explains the basis behind the first indication. The sustainability information can be determined based on the application of the generative AI model 114 to the third embedding and the fourth embedding.

[0073] According to one embodiment, the generator model 114A can receive a first set of inputs. The first set of inputs can include random inputs, a generated third embedding (i.e., a set of features of the queried material), and a generated fourth embedding (i.e., the impact of the queried material on the environment). Based on the first set of inputs, the generator model 114A can generate an output. Thereafter, the discriminator model 114B can receive a second set of inputs. The second set of inputs can include the output generated by the generator model 114A, the third embedding, and the fourth embedding. Based on the second set of inputs, the discriminator model 114B can generate a classification result indicating whether the queried material is sustainable or harmful. The generated classification result can be a prediction of the generative AI model 114 corresponding to the first indication. The discriminator model 114B can further generate the basis behind the first indication (i.e., the generated classification or prediction).

[0074] For example, the queried material can be a material that can be produced based on the oxidation of oil molecules by oxygen molecules. Information related to the material can indicate that when the material is used in a device deployed within a water mass and in contact with water, the material can release harmful chemicals into the water mass. Based on the application of the natural language model 112 to the queried material and the information related to the material, embeddings can be generated for the material and the information related to the material. Based on the application of the generative AI model 114 to the generated embeddings, the discriminator model 114B can classify the material as harmful and generate a rationale behind the classification that "the material is associated with a risk of releasing harmful chemicals into the water mass".

[0075] In another example, the queried material can be a synthetic fertilizer, and the information related to the queried material can indicate that the queried material releases excessive ammonia. The queried material can be classified as harmful, and the generated rationale can be that "synthetic fertilizers are associated with a risk of releasing harmful chemicals into the atmosphere". In another example, the queried material can be a synthetic drug, and the information related to the queried material can indicate that the queried material releases dioxins. The queried material can be classified as harmful, and the generated rationale can be that "synthetic drugs are associated with a risk of side effects for patients".

[0076] In another example, the queried material can be a material that can be used to evaluate industrial waste, and the information related to the queried material can indicate that the queried material is produced from plant-derived gum. The queried material can be classified as sustainable, and the generated rationale can be that "the material is supplied from plant-derived materials". In another example, the queried material can be a synthetic fertilizer, and the information related to the queried material can indicate that the queried material is produced from natural waste. The queried material can be classified as sustainable, and the generated rationale can be that "synthetic fertilizers are produced from natural waste".

[0077] In another example, the queried material can be a synthetic drug, and the information related to the queried material can indicate that the queried material is produced from vegetable oil. The queried material can be classified as sustainable, and the basis for the production can be that "synthetic drugs are supplied from plant herbs and do not cause side effects".

[0078] Embodiments of the present disclosure can promote the production of materials while taking into account the impact of the produced materials on the resources of the living environment. For the production of sustainable materials that are compatible with nature and environmentally friendly, the embodiments can utilize natural language models and generative AI models. The use of AI-based models enables accurate classification of materials regarding the sustainability of the materials with the environment, thereby protecting the environment from potentially harmful effects that non-sustainable materials may have on the environment. The AI-based models further facilitate providing an explanation behind the prediction indicating the sustainability of the materials. The explanation includes the basis behind each of the predictions regarding the impact that the materials may have on the environment. The generation of the basis behind the predictions can enable the creation of a knowledge base related to the characteristics of the materials and the use of the knowledge base to create training data for training the AI-based models to improve the accuracy of future predictions by the AI-based models.

[0079] Embodiments enable the design of new materials in an ethical manner that minimizes or nullifies any harmful impact of the materials on the environment or climate. Embodiments provide a conditional GAN model that can be trained based on embeddings indicating the characteristics of the materials and the impact of the materials on the resources of the living environment. The training enables the filtering of materials that are undesirable, incompatible, or harmful to the environment and facilitates the generation of new materials that can be environmentally friendly. The use of embeddings can facilitate the incorporation of time-varying environmental constraints into the conditional GAN model by encoding the dynamic reactions between chemical substances (i.e., reactants) to assist in the discovery, design, and synthesis of sustainable materials. The generation of materials can be aligned with the United Nations (UN) Sustainable Development Goals (SDGs).

[0080] Figure 4 is a diagram showing an exemplary architecture including a language model and a generative AI model for AI-based sustainable material design according to an embodiment of the present disclosure. Figure 4 will be described in relation to the elements from Figures 1, 2, and 3. Referring to Figure 4, an exemplary architecture 400 is shown. The exemplary architecture 400 may include the natural language model 112 and the generative AI model 114 of Figure 1. The natural language model 112 may receive as inputs a dataset 302A (represented as "X"), a set of materials 304A, and material information 304B. As described in Figure 3, the set of materials 304A and the material information 304B can be determined based on the application of one or more neural network models 110 to the dataset 302A. The set of materials 304A can be determined based on named entities (e.g., reactants or products) identified from the dataset 302A. The material information 304B may include information related to each material in the set of materials 304A (e.g., time-varying constraints related to the environment).

[0081] The natural language model 112 can be a transformer-based model that may include an encoder 402A and a decoder 402B. The encoder 402A can generate a first set of embeddings 306A based on a set of materials 304A. For each material in the set of materials 304A, an embedding (represented as "C") of the first set of embeddings 306A can be generated. The first set of embeddings 306A can indicate the characteristics of each material in the set of materials 304A. The encoder 402A can further generate a second set of embeddings 306B (represented as "C'") based on the material information 304B. Based on the information related to each material in the set of materials 304A (contained in the material information 304B), an embedding of the second set of embeddings 306B can be generated. The second set of embeddings 306B can indicate the influence of each material in the set of materials 304A on the resources of the living environment.

[0082] The generative AI model 114, i.e., each of the generator model 114A and the discriminator model 114B, may receive as inputs a first set of embeddings 306A and a second set of embeddings 306B. The generative AI model 114 may be trained based on the inputs. The generator model 114A may further receive a random input 404 (represented as “Z”) that may be generated by use of a set of Gaussian noise models. For a queried material (e.g., a material in the set of materials 304A or any other material), the generator model 114A may generate an output (represented as “G(Z,C,C’)”). The discriminator model 114B may further receive material information 304B and the generated output (i.e., “G(Z,C,C’)”) as inputs. Based on the application of the discriminator model 114B to the inputs, a result (represented as “R”) may be generated. The processor 204 may be able to determine a labeler loss 406A (e.g., a generator loss) based on the result and the output. Additionally, the processor 204 may be able to determine an anti-labeler loss 406B. The generator model 114A may be further trained based on the labeler loss 406A. Based on the training, the output generated by the generator model 114A (i.e., “G(Z,C,C’)”) may be updated. The discriminator model 114B may be able to generate a prediction (represented as “P”) indicating whether the queried material is sustainable or harmful. The discriminator model 114B may further be able to indicate the basis (represented as “R’”) behind the prediction. The prediction and the basis behind the prediction may constitute sustainability information related to the queried material.

[0083] Note that the exemplary architecture 400 of FIG. 4 is for illustrative purposes and should not be construed as limiting the scope of the present disclosure.

[0084] FIG. 5 is a diagram showing a flowchart of an example method for AI-based sustainable material design according to an embodiment of the present disclosure. FIG. 5 will be described in relation to the elements from FIGS. 1, 2, 3, and 4. Referring to FIG. 5, a flowchart 500 is shown. The method shown in flowchart 500 can start at 502 and can be executed by any suitable system, device, or apparatus, such as, for example, the example of the electronic device 102 of FIG. 1 or the processor 204 of FIG. 2. Although shown as individual blocks, the steps and operations associated with at least one block of flowchart 500 can be divided into further blocks, combined into fewer blocks, or deleted, depending on the particular embodiment.

[0085] At block 502, a dataset including information related to scientific literature can be received. In one embodiment, the processor 204 can be configured to receive a dataset including information related to scientific literature. Details of the reception of the dataset including information related to scientific literature are further provided, for example, in FIGS. 1 and 3.

[0086] At block 504, one or more neural network models 110 can be applied to the dataset. In one embodiment, the processor 204 can be configured to apply one or more neural network models 110 to the dataset. Details of the application of one or more neural network models 110 to the received dataset are further provided, for example, in FIGS. 1 and 3.

[0087] At block 506, based on the application of one or more neural network models 110, a set of materials and information related to each material in the set of materials can be determined. In one embodiment, the processor 204 can be configured to determine a set of materials and information related to each material in the set of materials based on the application of one or more neural network models 110. Details of the determination of the set of materials and the information related to each material in the set of materials are further provided, for example, in FIGS. 1 and 3.

[0088] At block 508, a first set of embeddings indicating a first set of features of each material in the set of materials can be generated. In one embodiment, the processor 204 can be configured to generate a first set of embeddings indicating a first set of features of each material in the set of materials. Details of the generation of the first set of embeddings are further provided, for example, in FIGS. 1, 3, and 4.

[0089] At block 510, for the set of materials, a second set of embeddings related to text content describing the impact of the set of materials on the resources of the living environment can be generated. In one embodiment, the processor 204 can be configured to generate a second set of embeddings indicating the impact of the set of materials on the environment for the set of materials. Details of the generation of the second set of embeddings are further provided, for example, in FIGS. 1, 3, and 4.

[0090] At block 512, based on the first set of embeddings and the second set of embeddings, the generative AI model 114 can be trained. In one embodiment, the processor 204 can be configured to train the generative AI model 114 based on the first set of embeddings and the second set of embeddings. Details of the training of the generative AI model 114 are further provided, for example, in FIGS. 1 and 3.

[0091] At block 514, a user input indicating information related to the queried material can be received. In one embodiment, the processor 204 can be configured to receive a user input indicating information related to the queried material. Details of the reception of the user input indicating information related to the queried material are further provided, for example, in FIGS. 1, 3, and 4.

[0092] At block 516, a third embedding indicating a second set of features of the queried material and a fourth embedding related to text content describing the impact of the queried material on the resources of the living environment can be generated. In one embodiment, the processor 204 can be configured to generate a third embedding indicating a second set of features of the queried material and a fourth embedding related to text content describing the impact of the queried material on the resources of the living environment. Details of the generation of the third embedding and the fourth embedding are further provided, for example, in FIGS. 1, 3, and 4.

[0093] At block 518, the generative AI model 114 can be applied to the third embedding and the fourth embedding. In one embodiment, the processor 204 can be configured to apply the generative AI model 114 to the third embedding and the fourth embedding. Details of the application of the generative AI model 114 to the third embedding and the fourth embedding are further provided, for example, in FIGS. 1, 3, and 4.

[0094] At block 520, based on the application of the generative AI model 114, sustainability information related to the queried material can be determined. In one embodiment, the processor 204 can be configured to determine sustainability information related to the queried material based on the application of the generative AI model 114. Details of the determination of the sustainability information related to the queried material are further provided, for example, in FIGS. 1, 3, and 4.

[0095] At block 522, a display device (such as display device 210A, etc.) can be controlled to render sustainability information related to the queried material. In one embodiment, processor 204 can be configured to control display device 210A to render sustainability information related to the queried material. Details of the control of the display device for rendering sustainability information are further provided, for example, in FIG. 1.

[0096] Flowchart 500 is shown as individual operations such as 502, 504, 506, 508, 510, 512, 514, 516, 518, 520, and 522, but the present disclosure is not so limited. However, in a particular embodiment, such individual operations can further be divided into additional operations, combined into fewer operations, or deleted, depending on a particular implementation, without detracting from the essence of the disclosed embodiments.

[0097] Various embodiments of the present disclosure may provide one or more non-transitory computer-readable storage media configured to store instructions that, in response to being executed, cause a system (such as an example of the electronic device 102) to perform operations. The operations may include receiving a dataset that may include information related to scientific literature. The operations may further include applying one or more neural network models (such as one or more neural network models 110) to the dataset. The operations may further include determining a set of materials (such as a set of materials 116) and information related to each material in the set of materials 116 based on the application of the one or more neural network models 110. The operations may further include generating a first set of embeddings (such as a first set of embeddings 118) that indicate a first set of features of each material in the set of materials 116. The operations may further include generating a second set of embeddings (such as a second set of embeddings 120) related to text content that describes the impact of the set of materials 116 on the resources of the living environment. The operations may further include training a generative AI model (such as a generative AI model 114) based on the first set of embeddings 118 and the second set of embeddings 120. The operations may further include receiving a user input that indicates information related to a queried material (such as a queried material 122). The operations may further include generating a third embedding that indicates a second set of features of the queried material 122 and a fourth embedding related to text content that describes the impact of the queried material 122 on the resources of the living environment. The operations may further include applying the trained generative AI model 114 to the third embedding and the fourth embedding. The operations may further include determining sustainability information related to the queried material 122 based on the application of the generative AI model 114. The operations may further include controlling a display device (i.e., the display device 210) to render the sustainability information related to the queried material 122.

[0098] As used in this disclosure, the terms “module” or “component” can refer to a specific hardware implementation configured to perform the operations of that module or component, and / or a software object or software routine stored in and / or executed by general-purpose hardware of a computing system (e.g., a computer-readable medium, a processing device, etc.). In some embodiments, the various components, modules, engines, and services described in this disclosure can be implemented as objects or processes that execute on a computing system (e.g., as separate threads). Some of the systems and methods described in this disclosure are generally described as being implemented in software (stored in and / or executed by general-purpose hardware), but specific hardware implementations or combinations of specific hardware and software are also possible and contemplated. In this description, a “computing entity” can be any computing system as previously defined in this disclosure, or any module or combination of modules that execute on a computing system.

[0099] The terms used in this disclosure, and particularly those used in the appended claims (e.g., the body of the appended claims), are generally intended to be “open” terms (e.g., the term “including” should be interpreted as “including but not limited to,” the term “having” should be interpreted as “having at least,” the term “comprising” should be interpreted as “comprising but not limited to,” etc.).

[0100] Also, if a specific number of claim recitations is intended, such intent shall be expressly recited in the claims, and in the absence of such recitation, no such intent exists. For example, for purposes of illustration, the following appended claims may include the use of introductory phrases “at least one” and “one or more” to introduce claim recitations. However, the use of such phrases should not be construed to limit a particular claim including a claim recitation introduced by an indefinite article “a” or “an” to embodiments that include only one of what is so recited, and this is true even where the same claim includes both an introductory phrase “one or more” or “at least one” and an indefinite article such as “a” or “an” (e.g., “a” and / or “an” should be construed to mean “at least one” or “one or more”), and the same is true for the use of definite articles used to introduce claim recitations.

[0101] Also, even where a specific number of claim recitations being introduced is expressly recited, one of ordinary skill in the art will recognize that such recitation should be construed to mean at least the recited number (e.g., a bare recitation of “two recitations” without other modifiers means at least two recitations, or two or more recitations). Also, where conventions similar to “at least one of A, B, and C” or “one or more of A, B, and C” are used, generally such constructs are intended to include only A, only B, only C, A and B together, A and C together, B and C together, or A, B, and C together, and the like.

[0102] Also, any disjunctive word or phrase presenting two or more alternative terms, whether in the specification, claims, or drawings, should be understood to contemplate the possibility of including one of those terms, any of those terms, or both terms. For example, the phrase “A or B” should be understood to include the possibility of “A” or “B” or “A and B”.

[0103] All examples and conditional language described in this disclosure are intended for educational purposes to help the reader understand the concepts contributed by the inventors to promote this disclosure and the relevant technical field, and should not be construed as being limited to such specifically described examples and conditions. Although embodiments of the present disclosure have been described in detail, various changes, substitutions, and modifications can be made without departing from the spirit and scope of the present disclosure.

[0104] Regarding the above description, the following additional remarks are disclosed. (Appendix 1) A method executed by a processor, comprising: Receiving a dataset including information related to scientific literature; Applying one or more neural network models to the dataset; Determining a set of materials and information related to each material in the set of materials based on the application of the one or more neural network models; Generating a first set of embeddings indicating a first set of features of each material in the set of materials; Generating a second set of embeddings related to text content describing the impact of the set of materials on the resources of the living environment; Training an artificial intelligence (AI) model generated based on the first set of embeddings and the second set of embeddings; Receiving a user input indicating information related to the queried material; Generating a third embedding indicating a second set of features of the queried material and a fourth embedding related to text content describing the impact of the queried material on the resources of the living environment; Applying the generated AI model to the third embedding and the fourth embedding; Determining sustainability information related to the queried material based on the application of the generated AI model; Controlling a display device to render the sustainability information related to the queried material. A method having the above. (Appendix 2) The method further applies a first neural network model among the one or more neural network models to the dataset, identifies a set of reactants based on the application of the first neural network model, applies a second neural network model among the one or more neural network models to the set of reactants, selects a subset of reactants from the set of reactants based on the application of the second neural network model, and has the determination of the set of materials is further based on the identification of the set of reactants and the selection of the subset of reactants, the method described in Appendix 1. (Appendix 3) The method further extracts information related to each reactant in the set of reactants based on the application of the first neural network model, and has the information related to each material in the set of materials includes the information related to each reactant among one or more reactants included in the corresponding material in the set of reactants, the method described in Appendix 2. (Appendix 4) The determined information related to each reactant in the set of reactants is the organic structure of the corresponding reactant, the decay rate related to the corresponding reactant, the biodegradability related to the corresponding reactant, one or more catalysts that promote the chemical reaction involved in the corresponding reactant, one or more products generated by the chemical reaction, the temperature requirement for causing the chemical reaction, or one or more precursors involved in the chemical reaction, and includes at least one of the above, the method described in Appendix 3. (Appendix 5) The method further Applying a natural language model to each material of the set of materials and the information associated with each material of the set of materials comprising the generation of the set of first embeddings and the generation of the set of second embeddings are further based on the application of the natural language model The method according to Appendix 1 (Appendix 6) The generation AI model corresponds to a conditional generative adversarial network (GAN) model including a generator model and a discriminator model, the method according to Appendix 1 (Appendix 7) The generator model is trained to generate an output for each material of the set of materials such that the discriminator model classifies each material of the set of materials as sustainable, the method according to Appendix 6 (Appendix 8) The output is generated based on at least one of a random input, the set of first embeddings, the set of second embeddings, a generator loss received from the discriminator model based on a previously generated output, or the information associated with each material of the set of materials, the method according to Appendix 7 (Appendix 9) The discriminator model is configured to classify the material as sustainable or harmful, the method according to Appendix 7 (Appendix 10) The classification by the discriminator model is based on at least one of the output generated for each material of the set of materials, the set of first embeddings, the set of second embeddings, a discriminator loss generated based on a previous classification result, and the information associated with each material of the set of materials, the method according to Appendix 9 (Appendix 11) The generator model receives a first set of inputs including a random input, the third embedding, and the fourth embedding The generator model generates an output based on the first set of inputs The discriminator model receives a second set of inputs including the output, the third embedding, and the fourth embedding classify the queried material as sustainable or harmful based on the second set of inputs by the discriminator model The method according to appendix 6, further comprising this (Appendix 12) The determined sustainability information related to the queried material corresponds to a first indication specifying whether the queried material is sustainable or harmful, and a second indication explaining the basis behind the first indication. The method according to appendix 1 (Appendix 13) One or more non-transitory computer-readable storage media configured to store instructions, which, in response to being executed, cause an electronic device to perform operations including receiving a dataset including information related to scientific literature applying one or more neural network models to the dataset determining a set of materials and information related to each material in the set of materials based on the application of the one or more neural network models generating a first set of embeddings indicating a first set of features of each material in the set of materials generating a second set of embeddings related to text content describing the impact of the set of materials on the resources of the living environment training a generated artificial intelligence (AI) model based on the first set of embeddings and the second set of embeddings receiving a user input indicating information related to the queried material generating a third embedding indicating a second set of features of the queried material and a fourth embedding related to text content describing the impact of the queried material on the resources of the living environment applying the generated AI model to the third embedding and the fourth embedding determining sustainability information related to the queried material based on the application of the generated AI model Controlling a display device to render the sustainability information related to the queried material A non-transitory computer-readable storage medium having the above. (Appendix 14) The operation is Applying a first neural network model among the one or more neural network models to the dataset Identifying a set of reactants based on the application of the first neural network model Applying a second neural network model among the one or more neural network models to the set of reactants Selecting a subset of reactants from the set of reactants based on the application of the second neural network model Having the above, The determination of the set of materials is further based on the identification of the set of reactants and the selection of the subset of reactants The non-transitory computer-readable storage medium according to Appendix 13 (Appendix 15) The operation is Extracting information related to each reactant in the set of reactants based on the application of the first neural network model Having the above, The information related to each material in the set of materials includes the information related to each reactant among one or more reactants included in the corresponding material in the set of reactants The non-transitory computer-readable storage medium according to Appendix 14 (Appendix 16) The determined information related to each reactant in the set of reactants is The organic structure of the corresponding reactant The decay rate related to the corresponding reactant The biodegradability related to the corresponding reactant One or more catalysts that promote the chemical reaction involving the corresponding reactant One or more products generated by the chemical reaction The temperature requirement for causing the chemical reaction, or One or more precursors involved in the chemical reaction, The non-transitory computer-readable storage medium according to appendix 15, comprising at least one of them. (Appendix 17) The operation is Applying a natural language model to each material of the set of materials and the information related to each material of the set of materials, Having, The generation of the first set of embeddings and the generation of the second set of embeddings are further based on the application of the natural language model, The non-transitory computer-readable storage medium according to appendix 13. (Appendix 18) The generated AI model corresponds to a conditional generative adversarial network (GAN) model including a generator model and a discriminator model, and is the non-transitory computer-readable storage medium according to appendix 13. (Appendix 19) The determined sustainability information related to the queried material corresponds to a first indication that defines whether the queried material is sustainable or harmful, and a second indication that explains the basis behind the first indication, and is the non-transitory computer-readable storage medium according to appendix 13. (Appendix 20) A memory configured to store instructions, A processor coupled to the memory and configured to execute the instructions to execute a process, the process comprising: Receiving a dataset including information related to scientific literature, Applying one or more neural network models to the dataset, Determining a set of materials and information related to each material of the set of materials based on the application of the one or more neural network models, Generating a first set of embeddings indicating a first set of characteristics of each material of the set of materials, Generating a second set of embeddings related to text content describing the impact of the set of materials on the resources of the living environment. Train a generative artificial intelligence (AI) model based on the first set of embeddings and the second set of embeddings, Receive user input indicating information related to the queried material, Generate a third embedding representing a second set of features of the queried material and a fourth embedding related to text content describing the impact of the queried material on the resources of the living environment, Apply the generative AI model to the third embedding and the fourth embedding, Determine sustainability information related to the queried material based on the application of the generative AI model, Control a display device to render the sustainability information related to the queried material, A processor having: An electronic device having.

Claims

1. 1. A processor-implemented method comprising: receiving a dataset containing information related to scientific literature; applying one or more neural network models to the data set; determining a set of ingredients and information associated with each ingredient in the set of ingredients based on said applying the one or more neural network models; generating a first set of embeddings indicative of a first set of features for each material in the set of materials; generating a second set of embeddings associated with textual content describing an impact of the set of materials on a living environment resource; training a generative artificial intelligence (AI) model based on the first set of embeddings and the second set of embeddings; receiving user input indicating information related to the queried material; generating a third embedding indicative of a second set of features of the queried material and a fourth embedding associated with textual content describing an impact of the queried material on the resources of the living environment; applying the generative AI model to the third embedding and the fourth embedding; determining sustainability information associated with the queried material based on the application of the generative AI model; controlling a display device to render the sustainability information related to the queried material. How to have that.

2. The method further comprises: applying a first neural network model of the one or more neural network models to the data set; identifying a set of reactants based on said applying of said first neural network model; applying a second neural network model of the one or more neural network models to the set of reactants; selecting a subset of reactants from the set of reactants based on the applying of the second neural network model. Having said that, the determining of the set of materials is further based on the identifying of the set of reactants and the selecting of the subset of reactants. The method of claim 1.

3. The method further comprises: extracting information associated with each reactant of the set of reactants based on the applying of the first neural network model. Having said that, the information associated with each ingredient of the set of ingredients includes information associated with each of one or more reactants included in a corresponding ingredient of the set of reactants; The method of claim 2.

4. The determined information associated with each reactant of the set of reactants comprises: The organic structures of the corresponding reactants, the decay rates associated with the corresponding reactants, Biodegradability associated with the corresponding reactants, one or more catalysts that promote a chemical reaction involving the corresponding reactants; one or more products produced by said chemical reaction; the temperature requirements for causing said chemical reaction; or one or more precursors involved in said chemical reaction; The method of claim 3 , comprising at least one of:

5. The method further comprises: applying a natural language model to each ingredient in the set of ingredients and to the information associated with each ingredient in the set of ingredients; Having said that, the generating of the first set of embeddings and the generating of the second set of embeddings are further based on applying the natural language model. The method of claim 1.

6. 2. The method of claim 1, wherein the generative AI model corresponds to a conditional generative adversarial network (GAN) model that includes a generator model and a discriminator model.

7. The method of claim 6 , wherein the generator model is trained to generate an output for each material in the set of materials such that the discriminator model classifies each material in the set of materials as sustainable.

8. 8. The method of claim 7, wherein the outputs are generated based on at least one of a random input, the first set of embeddings, the second set of embeddings, generator losses received from the discriminator model based on previously generated outputs, or the information associated with each material in the set of materials.

9. The method of claim 7 , wherein the discriminator model is configured to classify the materials as sustainable or hazardous.

10. 10. The method of claim 9, wherein the classification of the discriminator model is based on at least one of the outputs generated for each material in the set of materials, the first set of embeddings, the second set of embeddings, a discriminator loss generated based on previous classification results, and the information associated with each material in the set of materials.

11. receiving, by the generator model, a first set of inputs including a random input, the third embedding, and the fourth embedding; generating outputs based on the first set of inputs with the generator model; receiving, by the discriminator model, a second set of inputs including the output, the third embedding, and the fourth embedding; classifying the queried material as sustainable or hazardous based on the second set of inputs by the discriminator model. The method of claim 6 further comprising:

12. 2. The method of claim 1, wherein the determined sustainability information related to the queried material corresponds to a first indication defining whether the queried material is sustainable or harmful and a second indication explaining the rationale behind the first indication.

13. One or more non-transitory computer-readable storage media configured to store instructions that, in response to being executed, cause an electronic device to perform an operation, including: receiving a dataset containing information related to scientific literature; applying one or more neural network models to the data set; determining a set of ingredients and information associated with each ingredient in the set of ingredients based on said applying the one or more neural network models; generating a first set of embeddings indicative of a first set of features for each material in the set of materials; generating a second set of embeddings associated with textual content describing an impact of the set of materials on a living environment resource; training a generative artificial intelligence (AI) model based on the first set of embeddings and the second set of embeddings; receiving user input indicating information related to the queried material; generating a third embedding indicative of a second set of features of the queried material and a fourth embedding associated with textual content describing an impact of the queried material on the resources of the living environment; applying the generative AI model to the third embedding and the fourth embedding; determining sustainability information associated with the queried material based on the application of the generative AI model; controlling a display device to render the sustainability information related to the queried material. A non-transitory computer readable storage medium having

14. A memory configured to store instructions; a processor coupled to the memory and configured to execute the instructions to perform a process, the process comprising: receiving a dataset containing information related to scientific literature; applying one or more neural network models to the data set; determining a set of ingredients and information associated with each ingredient in the set of ingredients based on said applying the one or more neural network models; generating a first set of embeddings indicative of a first set of features for each material in the set of materials; generating a second set of embeddings associated with textual content describing an impact of the set of materials on a living environment resource; training a generative artificial intelligence (AI) model based on the first set of embeddings and the second set of embeddings; receiving user input indicating information related to the queried material; generating a third embedding indicative of a second set of features of the queried material and a fourth embedding associated with textual content describing an impact of the queried material on the resources of the living environment; applying the generative AI model to the third embedding and the fourth embedding; determining sustainability information associated with the queried material based on the application of the generative AI model; controlling a display device to render the sustainability information related to the queried material. a processor; An electronic device having the

Images