System and method for managing queries pertaining to a product

Abstract

A system and method for managing queries pertaining to a product are provided. The method includes receiving, by a processing unit, a user query from a user device. Further, one or more contexts are identified from a plurality of contexts by comparing the user query with each context among the plurality of contexts. Each of the contexts includes values of one or more attributes characteristic to at least a part of the product. Further, a prompt is generated by updating the user query based on the one or more contexts identified. The prompt is provided as input to a Large Language Model to generate a natural language output, on an output device.

Description

SYSTEM AND METHOD FOR MANAGING QUERIES PERTAINING TO A PRODUCTTECHNICAL FIELD

[0001] The present embodiments relate to product database management. In particular, the present embodiments relate to a system and a method for managing queries pertaining to a product.BACKGROUND

[0002] Understanding co-relation between various artifacts associated with various stages of a product’s lifecycle is crucial for quality, functional completeness, and speed of development of a product. Such co-relations among the artifacts form a ‘digital thread.’ However, discovery of such digital threads within an industrial scenario is cumbersome & time-consuming, as it is typically done manually.

[0003] In prior art, discovery of digital threads has been attempted through implementation of joins at database level based on mapping of schema and / or attributes, through the use of semantic information or through pattern matching of names or designated keywords. Further, some prior art also reveals the use of knowledge graphs to capture the digital threads. However, attempts to construct knowledge graphs suffer from limitations of the specific semantic models employed, as the relations in knowledge graphs are generally fixed, and not dynamic in nature. Additionally, modifying knowledge graphs is typically a costly exercise.SUMMARY

[0004] In light of the above, there exists a need for an improved system and method for managing queries pertaining to a product.

[0005] The present disclosure relates to a method and system for managing queries pertaining to a product.

[0006] In an aspect, a method for managing queries pertaining to a product is disclosed. The method includes receiving, by a processing unit, a user query from a user device. The method further includes identifying one or more contexts from a plurality of contexts by comparing the user query with each context among the plurality of contexts. Each of the contexts includes values of one or more attributes characteristic to at least a part of the product. In an embodiment, identifying the one or more contexts from the plurality of contexts includes generating a query embedding corresponding to the user query using a vectorization algorithm. Further, the query embedding is compared with each context embedding of a plurality of context embeddings stored in a vector database. Each context embedding of the plurality of context embeddings correspondsto a context among the plurality of contexts. Further, one or more closest matching context embeddings from the plurality of context embeddings, in relation to the query embedding, is identified. In an embodiment, the one or more closest matching context embeddings in relation to the query embedding are identified based on a value of cosine similarity computed for the query embedding and each of the context embeddings in the vector database. The one or more closest matching context embeddings are indicative of one or more contexts associated with the user query.

[0007] In an embodiment, the plurality of context embeddings are generated by querying a product database to generate a list of the plurality of contexts associated with the product. The list of the plurality of contexts is parsed to demarcate each of the contexts. Further, a vectorization algorithm is used to generate the context embeddings corresponding to each of the contexts in the structured list of the plurality of contexts.

[0008] The method further includes generating a prompt by updating the user query based on the one or more contexts identified. In an embodiment, generating the prompt by updating the user query based on the one or more contexts identified includes transforming the one or more contexts into one or more contextual descriptions using a natural language processing technique. In an embodiment, the natural language processing technique transforms the one or more contexts into one or more contextual descriptions by selecting a description template from a plurality of description templates based on the attributes present in each of the one or more contexts identified. Further, the one or more contextual descriptions are generated by populating one or more dynamic fields in the selected description template based on the attributes in each of the one or more contexts identified. Each of the contextual descriptions is a string including a natural language description of at least one of the contexts. Further, the user query is augmented with the one or more contextual descriptions using a text data augmentation technique. The method further includes providing the prompt as input to a Large Language Model to generate a natural language output, on an output device.

[0009] In another aspect, a system for managing queries pertaining to a product is disclosed. The system includes a processing unit, and a memory unit communicatively coupled to the processing unit. The memory unit includes a query management module that, when executed by the processing unit, is configured for identifying one or more contexts from a plurality of contexts by comparing the user query with each context among the plurality of contexts. Each of the contexts includes values of one or more attributes characteristic to at least a part of the product. In an embodiment, in identifying one or more contexts from a plurality of contexts by comparing the user query with each context among the plurality of contexts, the query management module, whenexecuted by the processing unit, is further configured for generating a query embedding corresponding to the user query using a vectorization algorithm. Further, the query embedding is compared with each context embedding of a plurality of context embeddings stored in a vector database. Each context embedding of the plurality of context embeddings corresponds to a context among the plurality of contexts. Further, one or more closest matching context embeddings from the plurality of context embeddings, in relation to the query embedding, are identified. In an embodiment, the one or more closest matching context embeddings in relation to the query embedding is identified based on a value of cosine similarity computed for the query embedding and each of the context embeddings in the vector database. The one or more closest matching context embeddings are indicative of one or more contexts associated with the user query.

[0010] In an embodiment, the query management module, when executed by the processing unit, is configured to generate the plurality of context embeddings by querying a product database to generate a list of the plurality of contexts associated with the product. The list of the plurality of contexts is parsed to demarcate each of the contexts. Further, a vectorization algorithm is used to generate the context embeddings corresponding to each of the contexts in the structured list of the plurality of contexts.

[0011] The query management module, when executed by the processing unit, is further configured for generating a prompt by updating the user query based on the one or more contexts identified. In an embodiment, generating the prompt by updating the user query based on the one or more contexts identified includes transforming the one or more contexts into one or more contextual descriptions using a natural language processing technique. In an embodiment, the natural language processing technique transforms the one or more contexts into one or more contextual descriptions by selecting a description template from a plurality of description templates based on the attributes present in each of the one or more contexts identified. Further, the one or more contextual descriptions are generated by populating one or more dynamic fields in the selected description template based on the attributes in each of the one or more contexts identified. Each of the contextual descriptions is a string including a natural language description of at least one of the contexts. Further, the user query is augmented with the one or more contextual descriptions using a text data augmentation technique. The query management module, when executed by the processing unit, is further configured for providing the prompt as input to a Large Language Model to generate a natural language output, on an output device.

[0012] In yet another aspect, a non-transitory computer-readable storage medium having machine- readable instructions stored therein, that when executed by a system, cause the system to perform a method of managing queries pertaining to a product, as described above.

[0013] This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the following description. The summary is not intended to identify features or essential features of the claimed subject matter. Further, the claimed subject matter is not limited to implementations that solve any or all disadvantages noted in any part of this disclosure.BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is a block diagram of an example system for managing queries pertaining to a product, according to one embodiment;

[0015] Figure 2 is a schematic representation of a system for managing queries pertaining to a product, according to another embodiment;

[0016] Figure 3 illustrates a block diagram of a system for managing queries pertaining to a product, according to yet another embodiment;

[0017] Figure 4 is a process flowchart depicting an example method for generating the plurality of context embeddings, in accordance with an embodiment;

[0018] Figure 5 is a process flowchart depicting an example method of managing queries pertaining to a product, according to one embodiment; and

[0019] Figure 6 shows structure of a Generative Pre-trained Transformer.DETAILED DESCRIPTION

[0020] The present disclosure relates to a method and system for managing queries pertaining to a product. Various embodiments are described with reference to the drawings, where like reference numerals are used in reference to the drawings. Like reference numerals are used to refer to like elements throughout. In the following description, numerous specific details are set forth to provide a thorough understanding of embodiments. These specific details need not be employed to practice embodiments. In other instances, well known materials or methods have not been described in detail to avoid unnecessarily obscuring embodiments. While the disclosure is susceptible to various modifications and alternative forms, specific embodiments thereof are shown by way of example in the drawings and will herein be described in detail. There is no intent to limit the disclosure to the particular forms disclosed. Instead, the disclosure is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the present disclosure.

[0021] Figure 1 is a block diagram of an example system 100 for managing queries pertaining to a product, according to one embodiment. The system 100 may be a personal computer, workstation, laptop computer, tablet computer, and the like. In Figure 1, the system 100 includes a processing unit 102, a memory unit 104, a storage unit 106, a bus 108, an input unit 110, and a display unit 112.

[0022] The processing unit 102, as used herein, may be any type of computational circuit, such as, but not limited to, a microprocessor, microcontroller, complex instruction set computing microprocessor, reduced instruction set computing microprocessor, very long instruction word microprocessor, explicitly parallel instruction computing microprocessor, graphics processor, digital signal processor, or any other type of processing circuit. The processing unit 102 may also include embedded controllers, such as generic or programmable logic devices or arrays, application specific integrated circuits, single-chip computers, and the like.

[0023] The memory unit 104 may be non-transitory volatile memory and non-volatile memory. The memory unit 104 may be coupled for communication with the processing unit 102, such as being a computer-readable storage medium. The processing unit 102 may execute instructions and / or code stored in the memory unit 104. A variety of computer-readable instructions may be stored in and accessed from the memory unit 104. The memory unit 104 may include any suitable elements for storing data and machine-readable instructions, such as read only memory, random access memory, erasable programmable read only memory, electrically erasable programmable read only memory, a hard drive, a removable media drive for handling compact disks, digital video disks, diskettes, magnetic tape cartridges, memory cards, and the like.

[0024] In the present embodiment, the memory unit 104 includes a query management module 114 stored in the form of machine-readable instructions on any of the above-mentioned storage media and may be in communication to and executed by the processing unit 102. The query management module 114 causes the processing unit 102 to manage queries pertaining to the product. For example, the query management module 114 manages queries pertaining to the product by receiving a user query from a user device, identifying one or more contexts from a plurality of contexts by comparing the user query with each context among the plurality of contexts, generating a prompt by updating the user query based on the one or more contexts identified, and providing the prompt as input to a Large Language Model to generate a natural language output, on an output device. Each of the contexts includes values of one or more attributes characteristic to at least a part of the product.

[0025] The storage unit 106 may be a non-transitory storage medium that stores a database 116. The database 116 includes a product database 118 and a vector database 120. The product database 118 includes product data such as part numbers, part revisions, synopsis, problem reports, problem report identifiers, etc., corresponding to a plurality of contexts associated with the product. The vector database 120 includes a plurality of context embeddings generated based on the plurality of contexts present in the product database 118. The process of generating the context embeddings is explained in detail later with reference to Figure 4. The input unit 110 may include input devices such as keypad, touch- sensitive display, camera (e.g., a camera receiving gesture-based inputs), etc. capable of receiving inputs indicative of the user query. The display unit 112 may be a device with a graphical user interface displaying options associated with editing the user query, and for viewing the natural language output generated by the Large Language Model. The bus 108 acts as interconnect between the processing unit 102, the memory unit 104, the storage unit 106, the input unit 110, and the display unit 112.

[0026] Those of ordinary skilled in the art will appreciate that the hardware components depicted in Figure 1 may vary for particular implementations. For example, other peripheral devices such as an optical disk drive and the like, Local Area Network (LAN) / Wide Area Network (WAN) / Wireless (e.g., Wi-Fi) adapter, graphics adapter, disk controller, input / output (I / O) adapter also may be used in addition to or in place of the hardware depicted. The depicted example is provided for the purpose of explanation only and is not meant to imply architectural limitations with respect to the present disclosure.

[0027] The system 100 in accordance with an embodiment of the present disclosure includes an operating system employing a graphical user interface. The operating system permits multiple display windows to be presented in the graphical user interface simultaneously with each display window providing an interface to a different application or to a different instance of the same application. A cursor in the graphical user interface may be manipulated by a user through the pointing device. The position of the cursor may be changed, and / or an event such as clicking a mouse button may be generated to actuate a response.

[0028] Figure 2 is a schematic representation of a system 200 for managing queries pertaining to the product, according to another embodiment. In one embodiment, the system 200 includes a cloud computing system 202 configured for providing cloud services for managing queries pertaining to a product.

[0029] The cloud computing system 202 includes a cloud communication interface 206, cloud computing hardware and OS 208, a cloud computing platform 210, the query management module 114, and the database 116. The cloud communication interface 206 enables communicationbetween the cloud computing platform 210, and user devices 212A-N such as smart phone, tablet, computer, etc. via a network 204.

[0030] The cloud computing hardware and OS 208 may include one or more servers on which an operating system (OS) is installed and includes one or more processing units, one or more storage devices for storing data, and other peripherals required for providing cloud computing functionality. The cloud computing platform 210 is a platform that implements functionalities such as data storage, data analysis, data visualization, data communication on the cloud hardware and OS 208 via APIs and algorithm, and delivers the aforementioned cloud services using cloud-based applications (e.g., Computer-Aided Engineering applications). The cloud computing platform 210 employs the query management module 114 for providing services related to managing queries pertaining to the product as described in Figure 1. The cloud computing platform 210 also includes the database 116 including a product database 118 for storing product data and a vector database 120 for storing context embeddings corresponding to different contexts associated with the product.

[0031] The user devices 212A-N include graphical user interfaces 214A-N for managing queries pertaining to the product. Each of the user devices 212A-N may be provided with a communication interface for interfacing with the cloud computing system 202. Users of the user devices 212A-N may access the cloud computing system 202 via the graphical user interfaces 214A-N. The graphical user interfaces 214A-N are specifically configured for accessing the query management module 1 14 in the cloud computing system 202.

[0032] Figure 3 illustrates a block diagram of a product management system 300 for managing manage queries pertaining to a product, according to yet another embodiment. Particularly, the product management system 300 includes a server 302 and a plurality of user devices 306A-N. Each user device of the plurality of user devices 306A-N is connected to the server 302 via a network 304 (e.g., Local Area Network (LAN), Wide Area Network (WAN), Wi-Fi, etc.). The system 300 is another implementation of the system 100 of Figure 1, in which the virtual query management module 114 resides in the server 302 and is accessed by user devices 306 A-N via the network 304.

[0033] The server 302 includes the virtual query management module 114 and the database 116. The server 302 may also include a processing unit (not shown), a memory unit (not shown), and a storage unit (not shown). The virtual query management module 114 may be stored on the memory unit in the form of machine-readable instructions and executable by the processing unit. The database 116 may be stored in the storage unit. The server 302 may also include a communication interface for enabling communication with user devices 306A-N via the network 304.

[0034] Figure 4 is a process flowchart 400 depicting an example method for generating the plurality of context embeddings, in accordance with an embodiment.

[0035] At act 405, a product database 118 is queried to generate a list of the plurality of contexts associated with the product. The list includes semi-structured data or structured data. For example, the list including semi- structured data may be generated as a JavaScript Object Notation (JSON file), Comma-Separated Values (CSV) file, or an Extensible Markup Language (XML) file. Similarly, the list including structured data may be generated as an Excel file.

[0036] In an example, the product database 118 is queried to extract a JSON file including different contexts associated with the product. In an example, the JSON file may include contexts associated with different parts of the product. For example, if the product is a car, the JSON file may include a plurality of contexts, as shown below:[{"id": "e9c5e84a-b02c-446b-8674-0c5c6ec6d7ad","part_number": "Apollo Inc. - Eldur - VISIBILITY","part_revision": "D","synopsis": "Windshield Wiper Motor May Fail","description": "The front windshield wiper motor may become inoperative, causing the wipers to fail. ","associated_Requirements ": {"type": "Requirement","reqID": "REQ- 1022","name": "Apollo Inc. - Eldur - VISIBILITY","reqSeverity ": "Medium ","parameters": [{"key ": "failureRate ","value": "low" f{"key ": "operationalStatus ","value ": "inoperative "}1,"owningUser": "Carl","parameter _tests" : [I"type": "Test Case","testID": "TST-0001","name": "Verify failure rate is set to low","test case ": ["Set the failure rate of the part to low","Check if the failure rate is correctly set to low"1,"testCaseOwner" : "Anthony " f{"type": "Test Case","testID": "TST-0002","name": "Verify operational status is set to inoperative","test case ": ["Set the operational status of the part to inoperative","Check if the operational status is correctly set to inoperative”1,"testCaseOwner" : "Samaira "}1 f"associated_ECN" : {"type": "ECN","ECNNumber": "ECN-6715","synopsis": "Windshield Wiper Motor Failure ","description": "The front windshield wiper motor has a potential to become inoperative, resulting in wiper failure. To resolve this issue, a change is required in the impacted part. ","impactedltem" : "Windshield Wiper Motor","requestor": "Angela" f"associated_solution_part ": {"part_number": "Apollo Inc. - El dur - VISIBILITY"," part _r evision": "D","solID": "SOL-7809"},"problem_report_id": "PR-73888 "},{"id": "321cdl8c-0918-46c4-bb74-2bc0ea4b83el ","part number": "Apollo Inc. - Eldur - WHEELS","part_revision ": "A ","synopsis": "Damaged Wheel Hub Locking Clips","description": "The locking clips of the wheel hub retaining nut may have been damaged during installation, which may cause the wheel hub to loosen and separate from the axle housing. ",”associated_Requirements ": {"type": "Requirement","reqID": "REQ-9317","name": "Wheel Hub Locking Clips Requirement","reqSeverity ": "High ","parameters": [{"dimension ": "Clip Material ","estimated_range": "Metal (Steel or Aluminum)" fI"dimension ": "Clip Strength ","estimated_range ": "High "}1,"owningUser": "Carl","parameter_tests ": [!’type": "Test Case",'testID": "TST-0001","name": "Verify Clip Material is within the expected range","test case": ["Select a sample of Wheel Hub Locking Clips","Inspect the material of each clip","Ensure the material is either Steel or Aluminum "],"testCaseOwner" : "Anthony "},{"type": "Test Case","testID": "TST-0002","name ": "Verify Clip Strength is high ","test case ": ["Select a sample of Wheel Hub Locking Clips ","Apply force to each clip to test its strength","Ensure each clip can withstand high levels of force"],"testCaseOwner": "Samaira "}1},"associated_ECN" : {"type": "ECN","ECNN umber": "ECN -1085","synopsis": "Damaged Wheel Hub Locking Clips","description": "The locking clips of the wheel hub retaining nut may have been damaged during installation, leading to potential loosening and separation of the wheel hub from the axle housing. ","impactedltem": "Apollo Inc. - WHEELS","requestor": "Angela"7,"associated_solution_part ": {"id": "321cdl8c-0918-46c4-bb74-2bc0ea4b83el ","synopsis": "Damaged Wheel Hub Locking Clips","part_number": "Apollo Inc. - El dur - WHEELS","pa rt_r evision ": "A ","description ": "The locking clips of the wheel hub retaining nut may have been damaged during installation, which may cause the wheel hub to loosen and separate from the axle housing. ","@ search, score": 1.0,"@ search. reranker_score null," @ search, highlight ": null,"@ search. captions": null,"solID": "SOL-8953" f"problem_report_id" : "PR-55673 "},{"id": "8d772518-0501-4034-9806-5a208419a4ee ","part _number": "Apollo Inc. - Eldur - ELECTRICAL SYSTEM","part _revision": "D","synopsis": "HVAC Controls May Become Inoperative","description": "The heating, ventilation, and air-conditioning (HVAC) system's controls, including those for defog and defrost, may become inoperative. ","associated_Requirements ": {"type": "Requirement","reqID": "REQ-7170","name": "Apollo Inc. - Eldur - ELECTRICAL SYSTEM","reqSeverity ": "High ","parameters": [{"key ": "part -revision ","value": "4" f{"key ”: "defog_controls ","value ": "inoperative " f{"key "defrost -Controls ","value "inoperative "}1,"owningUser": "Carl","parameter_tests ": [{"type": "Test Case","testID": "TST-0001","name": "Verify part_revisio parameter","test case ": ["Get the pari revision of the product","Check if the part revision matches the expected value","If it does, the test passes, otherwise it fails"7."testCaseOwner": "Anthony "},I"type": "Test Case","testID": "TST-0002","name": "Verify defog_controls parameter","test case ": ["Activate the defog controls of the product","Check if the defog controls are working properly","If they are, the test passes, otherwise it fails"],"testCaseOwner": "Samaira " f{"type": "Test Case","testID": "TST-0003","name": "Verify’ defrost _controls parameter","test case ": ["Activate the defrost controls of the product","Check if the defrost controls are working properly","If they are, the test passes, otherwise it fails"],"testCaseOwner": "Jonathan"}] f"associated_ECN" {"type": "ECN","ECNN umber "ECN- 7592 ","synopsis": "HVAC Controls May Become Inoperative","description": "The heating, ventilation, and air-conditioning (HVAC) system's controls, including those for defog and defrost, may become inoperative. ","impactedltem": "Apollo Inc. - Eldur - ELECTRICAL SYSTEM","requestor": "Angela" f"associated_solution_part ": {"type ": "Solution Item ","solID": "SOL-8213","name": "Updated HVAC Controls","Description": "This solution part is an updated version of Apollo Inc. - Eldur - ELECTRICAL SYSTEM part. It fixes the inoperability issue of the HVAC system's controls, including defog and defrost functions. ","owningUser": "Jacob" f"problem_report_id" : "PR-64142 "},{"id": "53574722-5128-4aa5-b25e-10a625216a59","part number": "Apollo Inc. - Eldur - FUEL SYSTEM, GASOLINE","part_revision": "C","synopsis": "High Pressure Fuel Pipe May Leak","description": "The nuts for the high pressure fuel pipe may loosen, which can result in a fuel leak. ","associated_Requirements {"type": "Requirement","reqlD": "REQ- 1548","name": "Fuel System Requirement Document","reqSeverity ": "High ","parameters ": [{"dimension ": "Tightness of Nuts ","estimated_range ": "60-80 Nm "1,{"dimension": "Leakage threshold","estimated_range ": "1 -2 ml / min "}],"owningUser”: "Carl" f"associated_ECN": {"type": "ECN","ECNN umber": "ECN -3095","synopsis": "High Pressure Fuel Pipe May Leak","description": "The nuts for the high pressure fuel pipe may loosen, which can result in a fuel leak. ","impactedltem": "Apollo Inc. - FUEL SYSTEM, GASOLINE","requestor": "Angela"},"associated_solution_part ": {"problemID": "53574722-5128-4aa5-b25e-10a625216a59","synopsis": "High Pressure Fuel Pipe May Leak","part_number": "Apollo Inc. - FUEL SYSTEM, GASOLINE","part_revision": "C","description": "The nuts for the high pressure fuel pipe may loosen, which can result in a fuel leak. ","solID": "SOL-8723""problem_report_id" : "PR-88397"}7

[0037] At act 410, the list of the plurality of contexts is parsed to demarcate each of the contexts. In the present example, the JSON file is further parsed to generate the context embeddings corresponding to each of the contexts. The individual contexts may be identified, for generating the context embeddings, based on presence of the demarcators ‘ { ‘ and ‘ } ’ in the JSON file. Further, each of the contexts include the same attributes: ‘part_number,’ ‘part_revision,’ ‘synopsis,’ ‘description,’ and ‘problem_report_id.’ Here, the attribute ‘problem_report_id’ in each of the contexts indicates a report identifier associated with the product. The attributes associated with each of the contexts (e.g., parts in the present example) are characteristic to the report identifier. Therefore, in another context including a different report identifier, a different set of attributes may be present. Further, each report identifier may be associated with a report type. For example, if the attributes in a context include ‘part_number,’ ‘part_revision,’ ‘synopsis,’ and ‘description,’ the report type may be predefined as ‘assembly _line.’ In another example, a different context may include additional attributes such as ‘part_id,’ ‘change_id,’ and ‘date_modified,’ and the report type for such parts may be predefined as ‘quality_rec.’

[0038] At act 415, a vectorization algorithm is used to generate the context embeddings corresponding to each of the contexts in the list of the plurality of contexts. Each of the context embeddings includes a numerical representation of the corresponding context. Non-limiting examples of vectorization algorithms include Word2Vec, GloVe, FastText, Bag of Words, and Continuous Bag of Words. The context embeddings thus generated are stored in the vector database 120.

[0039] Figure 5 is a process flowchart 500 depicting an example method for managing queries pertaining to a product, in accordance with an embodiment. It is to be understood that the system and methods described herein may be implemented in various forms of hardware, software, firmware, special purpose processing units, or a combination thereof. One or more of the present embodiments may take a form of a computer program product including program modules accessible from computer-usable or computer-readable medium storing program code for use by or in connection with one or more computers, processing units, or instruction execution system.

[0040] For the purpose of this description, a computer-usable or computer-readable medium may be any apparatus that may contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. The medium may be an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system (or apparatus or device), or propagation mediums in and of themselves as signal carriers are not included in the definition of physical computer-readable medium including a semiconductor or solid state memory, magnetic tape, a removable computer diskette, random access memory (RAM), a read only memory (ROM), a rigid magnetic disk, optical disk such as compact disk readonly memory (CD-ROM), compact disk read / write, and digital versatile disc (DVD), or any combination thereof. Both processing units and program code for implementing each aspect of the technology may be centralized or distributed (or a combination thereof) as known to those skilled in the art.

[0041] At act 505, a user query is received from a user device, by the processing unit 102. In the present example, the user query is received as a natural language text comprising text, numerals, symbols, or a combination thereof. For example, the user query may be ‘Show me the parts affected by Poor Material Quality in Apollo Eldur? ’

[0042] At act 510, one or more contexts from a plurality of contexts are identified by comparing the user query with each context among the plurality of contexts. Each of the contexts includes values of one or more attributes characteristic to at least a part of the product. The one or more attributes in a specific context form a schema.

[0043] In an embodiment, identifying the one or more contexts include generating a query embedding corresponding to the user query using a vectorization algorithm. The query embedding includes a numerical representation of the user query. For example, the numerical representation is indicative of meaning of words, phrases, and other data types, and relationships thereof. The query embedding is generated by processing the user query using the vectorization algorithm. Nonlimiting examples of vectorization algorithms include Word2Vec, GloVe, FastText, Bag of Words, and Continuous Bag of Words.

[0044] Further, the query embedding is compared with a plurality of context embeddings stored in a vector database 120. Each context embedding of the plurality of context embeddings corresponds to a context among the plurality of contexts. The process of generating the context embeddings is explained in detail with reference to Figure 5. From the plurality of context embeddings, one or more closest matching context embeddings, in relation to the query embedding, are identified. In an embodiment, the one or more closest matching context embeddings, in relation to the query embedding, is identified based on a value of cosine similaritycomputed for the query embedding and each of the context embeddings in the vector database 120. In an implementation, the context embedding for which the value of cosine similarity is highest is selected as the closest matching context embedding. In another implementation, one or more context embeddings with the value of cosine similarity greater than a predefined threshold (e.g., 0.8) are selected as closest matching context embeddings. The closest matching context embedding is indicative of one or more contexts associated with the user query. More specifically, each of the context embeddings corresponds to a specific context.

[0045] In the current example, the query embedding corresponding to the user query ‘Show me the parts affected by Poor Material Quality in Apollo Eldur?’, is compared with each of the context embeddings in the vector database 120. Further, the context embeddings corresponding to the following five contexts may be identified from the vector database 120, as the closest matching context embeddings:[{"problem_report_id": "PR-51355 ","part_number": "Apollo Inc. - Eldur - STEERING", "synopsis": "Steering Gear Assembled Incorrectly", "description "The steering gears may have been assembled incorrectly, which can cause the gears to fracture. ","part_revision "A ","associated_ECN": "{\"type\": \"ECN\", VECNNumberX": \"ECN-7081\", \"synopsis\": \"Propose correct assembly process for Steering Gear\", \"description\" : \"The steering gears of the Apollo Inc. - Eldur - STEERING part may have been assembled incorrectly, causing the gears to fracture. A new assembly process is being proposed to rectify the issue.X", \"impactedltem\": X'Apollo Inc. - Eldur - STEERINGX", X" requestors.": X'AngelaX"}",”associated_Requirements": ”{\"type\": \"Requirement\", \"reqID\": \"REQ-7583X", \"name\": \" Apollo Inc. - Eldur - STEERINGX", X"reqSeverity\": \"High\", VparametersX" : [{\"dimension\": X'AssemblyX", \"range\": \"Check for correct assembly of steering gearsX"}, {\"dimension\": \"Fracture\", \"range\": \"Ensure gears do not fractureX"}], X"owningUser\": \"Carl\", \"parameter_tests\": [{\"type\": \"Test CaseX", X"testID\": \"TST- 0001X", \"name\": \"Verify correct assembly of steering gearsX", \"test caseX": [X'Tnspect the steering gears for proper alignmentX", \"Ensure all bolts and screws are tight andsecureX", \"Check for any missing or loose partsX", \"Verify that all components are properly lubricatedX"], X"testCaseOwner\": X'AnthonyX"}, {X"type\": \"Test CaseX", X"testID\": \"TST-0002\", \"name\": \"Check for fractures in steering gearsX", \"test caseX": fVisually inspect all steering gears for any signs of fractures or cracksX", \"Use nondestructive testing methods like ultrasonic or magnetic particle inspection to detect hidden fracturesX", \"Perform stress tests on steering gears to ensure they can withstand the required loadX"], X"testCaseOwner\": \"Samaira\"}]}","associated_solution_part": "{\"id\": \"19752990-7f97-4fcc-ae74-661af86b3772\",\"synopsis\": \"Steering Gear Assembled Incorr ectlyX", \"part_number\": X'Apollo Inc. - Eldur - STEERINGX", \"part_revisiori\" : \"A\", \"description\" : \"The steering gears may have been assembled incorrectly, which can cause the gears to fracture \", \"@ search.scoreX" : 1.0, \"@ search. reranker_score\": null, \"@search.highlights\": null, \"@search.caplions\": null, X"solID\": X"SOL-4803\"}"},{"problem_report_id": "PR-51875 ","part_number": "Apollo Inc. - Eldur - STRUCTURE","synopsis ": "Cross Member May Come Loose Or Detach ","description": "The cross member in the slide pan frame may develop cracks, which over time could cause the cross member to become loose or detach from the frame. ","part _r evision": "D","associated_ECN": "f\"type\": \"ECNX", \"ECNNumber\": \"ECN-3321\", \"synopsis\": \"Cross Member May Come Loose Or DetachX", \"description\": \"The cross member in the slide pan frame may develop cracks, which over time could cause the cross member to become loose or detach from the frameX", \"impactedltem\" : X'Apollo Inc. - Eldur - STRUCTURES", X" requestors": X'AngelaX"}","associated_Requirements": "{\"type\': \"Requirement\", \'reqID\': \"REQ-31 ] 5\' \"name\": X'Apollo Inc. - Eldur - STRUCTURE - Revision 4\", X"reqSeverity\": \"High\", \"parameters\": [ f'dimensionX": X" Crack Lengths", \"range\": \"Omm - 10mni\"j, f\" dimensions": 'Cross Member Thickness ", \"range\": \"5mm - 15mm\"}], X" owningUserS": X'AnamX", X" parameter _tests\": [{X"type\": \"Test CaseX", X"testID\": \"TST-0001\", \"name\": \"Check if the crack length is within the specified rangeX", \"test caseX": f\"Create a new product partX", \"Set the crack length to 0.5mm\", \"Check if the crack length is set correctlyX", \"Set the crack length to 10.5mm\", \"Check if the cracklength is set correctlyX" ] , \"testCaseOwner\" : SAnthonyX"], {StypeS: S’Test CaseX", X"testID\": STST-0002X", S'nameX": \"Check if the cross member thickness is within the specified rangeX", S'test caseX": jX" Create a new product partS', \"Set the cross member thickness to 4.5mm\", X" Check if the cross member thickness is set correctlyX", S’Set the cross member thickness to 15.5mm\", \"Check if the cross member thickness is set correctlyX"], X"testCaseOwner\" : \"Samaira\"]]]","associated_solution_part": "{StypeX": \"Solution Iteni\", X"solID\": X"SOL-1945\", S’nameX": \"Reinforced Cross Member for Apollo Inc. Eldur Slide Pan FrameX", X"Descriptiori\": \"This solution part is a reinforced cross member specifically designed for Apollo Inc. Eldur slide pan frame. It addresses the issue of developing cracks in the original cross member, ensuring that it remains secure and prevents any potential detachment from theframeX", S’owningUserX": S'VikasX"}" }, {"probl em_report_id" "PR-26663 ","part_number": "Apollo Inc. - Eldur - SUSPENSION","synopsis ": "Incorrect Components in Rear Suspension Module ","description": "Incorrect rear suspension module components may have been installed due to a labeling error. ","part_revision "B ","associated_ECN": "{X"type\": S'ECNX", VECNNumberX": \"ECN-5337\", SsynopsisX": \"Incorrect Components in Rear Suspension ModuleX", X" descriptions": \"Incorrect rear suspension module components may have been installed due to a labeling error.X", \"impactedltem\" : X" Apollo Inc. - Eldur - SUSPENSIONS', X" requestor ': S'AngelaX"]","associated_Requirements": "(S'typeX": S'RequirementX", X"reqID\": \"REQ-3897X", S'nameX": \”Rear Suspension Module ComponentsX", 'reqSeverityX": X"Critical\", \"parameters\" : ]{\"key\": S'partjnumberX", S'valueX": X" Apollo Inc. - Eldur - SUSPENSIONS'], ]\"key\": X"part_revisionX", S'valueX": \"2\"}], S’owningUserX": 'CarlX", \"parameter_tests\": ] / \"type\": S’Test CaseX", S'testlDX": S'TST-OOOIX", S’nameX": S’Verify that the part number is corrects', S'test caseX": SRetrieve the part number from the rear suspension moduleX", VCompare the retrieved part number with the expected value Apollo Inc. - Eldur - SUSPENSIONS', S'Pass the test if the part numbers match, otherwise faiX"], S'testCaseOwnerX": X'AnthonyS], fS'typeX": S’Test CaseX", S'testlDX": STST-0002X", S'nameX": S’Verify the part revision is incremented:", S'test caseX": ^"Retrieve the partrevision from the rear suspension module\", VConvert the revision value to an integers", \"Increment the value by IS", \"Compare the incremented revision value with the expected value 3\", S"Pass the test if the values match, otherwise fails"], S" testCaseOwnerS": \"Samaira\"}]}","associated_solution_part": ”{S"typeS": VSolulion ItemS", \"solID\": S"SOL-3481S", S"nameS": VCorrected Rear Suspension ModuleS", \"Description\": \"The solution part replaces the incorrect rear suspension module components with the correct ones, resolving the labeling error issue.S", \"owningUser\": \"Vikas\'j" }, I"problem _report Jd" : "PR-48930","part _number": "Apollo Inc. - Eldur - SERVICE BRAKES, HYDRAULIC","synopsis": "Loss ofABS / ESC","description": "An improperly welded rotor and shaft may separate within the integrated brake system. ","part_revision "A ","associated_ECN": "{S"typeS": S'ECNS", VECNNumberS": \"ECN-6702\", \"synopsis\": S'Loss of ABS / ESC\", \"description\": ^Improperly welded rotor and shaft causing separation within the integrated brake systemS", \"impactedltem\" : \" Apollo Inc. - Eldur - SERVICE BRAKES, HYDRA ULICS", S" requestors" : S'AngelaS"}","associated_Requirements": "]S" typeS": \"Requirement\", \"reqID\": \"REQ-8664\", \"name\": VApollo Inc. - Eldur - SERVICE BRAKES, HYDRAULICS", \"reqSeverity\": S'HighS”, \”parameters\": ]{S" dimensions": VWelding QualityS", \"range\": VProperly Weldeds"}, {\"dimension\": \"Rotor-Shaft Integrations", S"rangeS": S"Secure Integrations"}], S" owningUserS": S"Carl\", S"parameter_testsS" : ]{S"typeS": S"Test CaseS", S"testIDS": S"TST-0001S", S"nameS": S"Test the welding quality of the partS", S"test caseS": S"l. Inspect the welded areas of the partS\n2. Check for any signs of improper welding sn3. Ensure all welds are of high qualityS", S"testCaseOwnerS" : S'AnthonyS"}, {S"typeS": \"Test CaseS", S"testIDS": S"TST-0002S", S"nameS": S"Test the integration of the rotor and shafts", S"test caseS": S"l. Verify the secure integration of the rotor and shaffSn2. Apply force to check for any movement or loosenessSn3. Ensure the integration is strong and stableS", S"testCaseOwnerS" : S"Samaird\"} ]} ","associated_solution_part": "{S"typeS": S"Solution ItemS", S"solIDS": S"SOL-1254S", S"nameS": S'lmproved Integrated Brake SystemS", S'DescriptioriS": S"This solution partaddresses the loss of ABS / ESC by introducing an improved integrated brake system. The improperly welded rotor and shaft are replaced with a more reliable and durable design, ensuring that they do not separate and cause the loss of ABS / ESC functionalityX", \"owningUser\": \"Jacob\"} "}, {"problem_report_id" : "PR-98321 ","part_number": "Apollo Inc. - Ultra) - POWER TRAIN","synopsis": "Unintentional Rollaway Due to Transmission Washer","description": "A washer within the vehicle's transmission can disintegrate, causing metallic debris to become lodged in the 'Park' mechanism. ","part _r evision ": "B ","associated_ECN": "{\"type\": \"ECN\", \"ECNNumber\": \"ECN-931&s", \"synopsis\": \" Unintentional Rollaway Due to Transmission Washers", \" description's": \"A washer within the vehicle's transmission can disintegrate, causing metallic debris to become lodged in the 'Park' mechanism.\", \"impactedltem\" ': \"Apollo Inc. - Ultra) - POWER TRAINS", \" requestors": \" Angela's"}","associated_Requirements": "{\"type\': \"Requiremenfs", \'reqID\': \"REQ-8095\", \"name\": \"Transmission Washer Failure's", \"reqSeverity ": \"High\”, \"parameters\": [ {S" parameter's" : \"Washer Materials", \" range's": \"Metallic\" } ], \"owningUser\": \"Carl ", \"parameter_tests\": [ / \"type\": \"Test Case's", \"testID\": \"TST-1234s", \"name\": \"Test the washer material to ensure it is metallicS", \”test case's": [\"Check the material of the washer's", VVerify that the material is metallicS"], \"testCaseOwner\": \"Anthony\"}]}","ass()ciated_solulion_part": "{S"type\": \"Solution Item\", \"solJD\": \"SOL-7424\", \"name\": \"Revised Transmission Washer's", \"Description\" : \"This solution part is a revised transmission washer that has been redesigned to prevent disintegration. It is made of a more durable material to ensure that no metallic debris is produced, thus eliminating the risk of it becoming lodged in the ’Park' mechanismX", \" ownin gU s er\" : \"Jacob\"}" } 1

[0046] At act 515, a prompt is generated by updating the user query based on the one or more contexts identified. In an embodiment, generating the prompt by updating the user query based on the one or more contexts identified includes transforming the one or more contexts into one or more contextual descriptions using a natural language processing technique. Each of the contextual descriptions is a string including a natural language description of at least one of the contexts. In a further embodiment, the natural language processing technique transforms the one or more contexts into one or more contextual descriptions by selecting a description template from a plurality of description templates based on the attributes present in each of the one or more contexts identified. In an implementation, the attributes associated with the closest matching context embedding is identified. Further, the description template is selected from the plurality of description templates based on the attributes identified. In another implementation, the description template is predefined for each report identifier. In yet another implementation, description template is broadly predefined for each report type. For example, if there are two description templates, one corresponding to each report type (denoted as prType): ■assembly line’ and ‘ quality _rec’, the description templates may be selected as shown using the pseudocode below: if (prType == 'assembly_line'): template_string = 'The ’ + geiRevString(obj['pari_revision']) + ' revision of ’ + obj ['part _number'] + ’ has a problem that the ' + obj['synopsis'] + ' and with the description that ' + obj['descriptior] else if (prType == ’ quality _rec’): template_string = 'The ' + getRevString(obj['part_revision']) + ' revision of ' + objlp' artjd'] + ' found issues with quality in ' + obj['synopsis'] + ' in respect to the change ' + obj[‘ ’change _id'] + ' that had a modification date of ' + obj['date_modified'] + ' and a description that ' + obj ['description']

[0047] The one or more contextual descriptions are further generated by populating one or more dynamic fields in the selected description template based on the attributes in each of the one or more contexts identified. In other words, the contextual description is generated by replacing the attributes in the contextual description with specific values corresponding to the context associated with each of the closest matching context embedding(s). In the present example, the description template: 'The ' + getRevString(obj['part_revision']) + ' revision of ' + obj ['part_number'] + ' has a problem that the ’ + obj['synopsis'] + ' and with the description that ' + obj ['description'] isselected, as the report type is identified as ‘assembly_line.’ Further, the description template is populated with values of the attributes p ' art_revision', 'part_number', 'synopsis' and 'description' in the dynamic fields getRevString(obj['part_revision']), obj['part_number'], objl'synopsis'] and objl'description'] respectively, from each of the contexts identified.

[0048] The contextual descriptions may be of the following format:[{"id": " 19752990-7f97-4fcc-ae74-661af86b3772 ","thread_summary" : "This problem report (PR-51355) is about the Apollo Inc. - Eldur - STEERING part. The problem is that the steering gears may have been assembled incorrectly, leading to gear fracture. The associated requirements (REQ-7583) state the need to check for correct assembly of steering gears and ensure that they do not fracture. The owning user responsible for this issue is Carl rignTo address the problem, an Engineering Change Notice (ECN-7081 ) has been proposed. The ECN suggests a correct assembly process for the steering gear of the Apollo Inc. - Eldur - STEERING part to prevent gear fracture. The requestor of the ECN is Angela. The proposal aims to rectify the issue by implementing a revised assembly process.Additionally, there is an associated solution part (SOL-4803) with the same synopsis and description as the problem. The solution part details the steps to fix the incorrectly assembled steering gears and prevent gear fracture. ","problem_report_id": "PR-51355 ","part_number" : "Apollo Inc. - Eldur - STEERING"},{"id": "bal5J7db-bl 79-41f5-8f23-c73a3a21ec5f","thread_summary" : "This Problem contains details about PR-51875. The problem is related to the part number \' Apollo Inc. - Eldur - STRUCTURE^' and its revision 4. The synopsis of the problem is that the cross member in the slide pan frame may come loose or detach. This issue is caused by cracks developing in the cross member over time. The associated requirements, owned by Anam, include parameters such as crack length ( 0mm - 10mm) and cross member thickness (5mm - 15mm). The associated ECN (Engineering Change Notice) numbered ECN-3321, requested by Angela, addresses the same issue. Thesolution part SOL- 1945, owned by Vikas, is a reinforced cross member designed specifically for Apollo Inc. Eld r's slide pan frame to address the issue of crack development and prevent detachment from the frame. ","problem_report_id": "PR-51875 ","part_number": "Apollo Inc. - Eldur - STRUCTURE" f {"id": "d908becl-d97c-474f-841f-080d063belb6","thread_summary" : "This problem report (PR-26663) relates to a situation where incorrect components were installed in the rear suspension module due to a labeling error. The part number for this module is \" Apollo Inc. - Eldur - SUSPENSIONS' and the current revision is B. \n\nThe associated requirement (REQ-3897) for the rear suspension module components is critical. The owning user for this requirement is Carl. \n\nAn engineering change notice (ECN -5337) has been raised for this problem, highlighting the incorrect components in the rear suspension module. The requestor for this ECN is Angela. \rt\nTo resolve this issue, a solution part (SOL-3481), called the Corrected Rear Suspension Module, has been identified. This solution part replaces the incorrect components with the correct ones, thereby rectifying the labeling error. The owning user for this solution part is Vikas. ","problem_report_id" : "PR-26663 ","part _number": "Apollo Inc. - Eldur - SUSPENSION"}, {"id": "7ef600e2-76ff-4dfa-bl8l-0cb9lbehdac7","thread summary" : "This Problem contains details about PR-48930. The problem is related to the part number\"Apollo Inc. - Eldur - SERVICE BRAKES, HYDRAULICS" with part revision A. The problem is described as a \"Loss of ABS / ESCs” due to an improperly welded rotor and shaft that may separate within the integrated brake system. \n\nThe associated requirement with ID REQ-8664 is owned by Carl and has a severity level of High. The requirement addresses the dimensions of\"Welding Quality's" and\"Rotor-Shaft Integrations' with the ranges of S'Properly Welded:" and S'Secure Integrations' respectively. \rtsnThe associated engineering change notice (ECN) with number ECN -6702 was requested by Angela. It addresses the same issue of loss of ABS / ESC caused by the improperly welded rotor and shaft. SisnThe associated solution item with ID SOL- 1254 iscalled VImproved Integrated Brake System\" and is owned by Jacob. This solution introduces an improved integrated brake system design that replaces the problematic rotor and shaft, ensuring they do not separate and cause the loss of ABS / ESC functionality. ", "problem_report_id" : "PR-48930","part_number": "Apollo Inc. - Eldur - SERVICE BRAKES, HYDRAULIC"}, {"id": "7cc78030-caae-4975-8bec-ebf784245714 ","thread_summary" ': "This problem report (PR-B) is about an issue with the transmission washer in the Apollo Inc. - Ultra) - POWER TRAIN. The problem is that the washer can disintegrate, causing metallic debris to become stuck in the Park' mechanism, leading to unintentional rollaways. The associated requirement (REQ-8095) states that the washer material is metallic and the severity is high. Carl is responsible for this requirement. \ri\nAngela has raised an Engineering Change Notice (ECN-9316)for this problem, which has the same synopsis and description. The impacted item is also the Apollo Inc. - Ultra ) - POWER TRAIN. \n\nThe solution to this problem is a revised transmission washer (SOL- 7424 ), redesigned to prevent disintegration. It is made of a more durable material to eliminate the production of metallic debris, thus eliminating the risk of it becoming lodged in the 'Park' mechanism. Jacob is responsible for this solution part. ","problem _report Jd" : "PR-98321 ","part _number": "Apollo Inc. - Ultra) - POWER TRAIN"} ]

[0049] The user query is further augmented with the one or more contextual descriptions, using a text data augmentation technique, to generate an updated prompt. In an implementation, augmenting the user query with the contextual description includes appending the user query with the contextual description. In another embodiment, augmenting the user query with the contextual description includes random insertion of the one or more contextual descriptions into the user query. Non-limiting examples of other text data augmentation techniques usable for augmenting the user query with the one or more user descriptions include synonym replacement, random swap, and random deletion.

[0050] At act 520, the updated prompt is provided as input to a Large Language Model to generate a natural language output, on an output device. In an example, the Large Language Model may respond to the updated prompt with the following response:Based on the data provided, the components affected by Poor Material Quality’ in the Apollo Eldur are as follows:STEERING - Erroneous assembly may lead to fractures in the steering gears. [PR-51355] STRUCTURE - Cracks in the slide pan frame may result in the loosening or detachment of the cross member from the slide pan frame. [PR-51875]SUSPENSION - Labeling errors may lead to the incorrect installation of the rear suspension module. [PR-26663]HYDRAULIC SERVICE BRAKES - Improper welding of the rotor and shaft within the integrated braking system may cause separation within the braking system. [PR-48930]POWERTRAIN - A washer within the vehicle's transmission may disintegrate, resulting in metallic debris becoming lodged in the 'Park' mechanism. [PR-98321 ]

[0051] In an embodiment, the Large Language Model (LLM) is a Generative Pre-Trained Transformer (GPT) such as the one in Figure 6 or its variant such as GPT-2, GPT-3, GPT-5 etc., trained based on corpus based on generic or proprietary information such as facts, documentations, and rules from various sources. Non-limiting examples of other Large Language Models include Gemini, BERT, Claudia, and Llama. However, it is likely that the LLM cannot be trained for information related to specific products. It may be appreciated that the present embodiments facilitate the use of such an LLM for providing product-specific queries even if the LLM is not pretrained based on information related to the product. The output generated by the Large Language Model is further provided to the output device for viewing by the user.

[0052] Figure 6 shows structure of a GPT model 600. The GPT model 600 includes an embedding block 605, one or more transformer encoder blocks 610, and a linear activation layer 615. The embedding block 605 encodes text-based inputs (e.g., tokens) to dense vector representations. The transformer encoder block 610 enables capturing of contextual information and learning hierarchical representations of the input based on the dense vector representations. In particular, the transformer encoder block 610 includes self-attention mechanisms 620, feedforward neural networks 625, and normalization layers 630 and 635.

[0053] The GPT model 600 is pretrained using unsupervised learning techniques using corpus including data from various sources. The unsupervised learning techniques may include, but are not limited to, autoregressive language modeling, masked language modeling, and next-sentence prediction. In autoregressive language modeling, the model predicts the next word / token in asequence given the preceding context. This involves training the model to generate coherent and contextually relevant text by maximizing the likelihood of predicting the next token in a sequence based on the preceding tokens. In masked language modeling, certain tokens in the input sequence are masked, and the model is trained to predict the masked tokens based on the context provided by the unmasked tokens. This encourages the model to learn bidirectional representations of the input text. In next-sentence prediction, the model is trained to predict whether a given pair of sentences follow each other in the original text, in order to help the model learn semantic relationships and coherence between sentences.

[0054] Upon pretraining, the GPT model 600 is fine-tuned using supervised learning methods to perform specific downstream tasks such as text classification, sentiment analysis, named entity recognition, or machine translation. Each task is associated with a task-specific dataset containing labeled examples. Consequently, model parameters associated with the GPT model 600 are finetuned during the supervised learning to adapt the GPT model 600 for the specific downstream tasks. The model parameters are further iteratively updated through backpropagation and optimization algorithms, such as stochastic gradient descent, Adam, or RMSprop, to minimize loss functions and enhance model performance across various natural language processing tasks. As the training of a GPT model 600 is apparent to a person skilled in the art, it is not further detailed herein.

[0055] However, it may be understood that it is not possible to train a Large Language Model with every possible information related to specific products. This may require regular training or updating of the Large Language Model with updated information related to the products. Such training or updating of the Large Language Model may be hectic and ineffective as such training involves manual preparation of corpus for the training. Advantageously, the present invention enables use of existing Large Language Models to provide context-based information related to specific products of interest for a user without the need for training the Large Language Model based on corpus related to the specific product. In particular, query from a user is preprocessed to add context related to the specific context and fed to the Large Language Model in order to enable the Large Language Model to generate a context-relevant output. Further, the present invention eases retrieval of product-related information using queries in natural language, from product databases, thereby making discovery of digital threads, co-relations, and useful insights from the product databases easier. Further, the use of domain-specific semantics facilitates generation of context-relevant responses, as opposed to generic responses, using the Large Language Model.

[0056] While the present disclosure has been described in detail with reference to certain embodiments, the present disclosure is not limited to those embodiments. In view of the presentdisclosure, many modifications and variations would present themselves to those skilled in the art without departing from the scope of the various embodiments of the present disclosure, as described herein.List of References100 system for managing queries pertaining to a product, according to an embodiment102 processing unit104 memory unit106 storage unit108 bus110 input unit112 display unit114 query management module116 database118 product database120 vector database200 system for managing queries pertaining to a product, according to another embodiment202 cloud computing system206 cloud communication interface208 cloud computing hardware and OS210 cloud computing platform212A-N user devices214A-N graphical user interfaces300 system for managing real-time multi-user collaboration, according to yet another embodiment302 server304 network306A-N plurality of user devices600 GPT model605 embedding block610 one or more transformer encoder blocks615 linear activation layer620 self- attention mechanisms625 feedforward neural networks630 and 635 normalization layers

Claims

CLAIMS1. A method for managing queries pertaining to a product, the method comprising: receiving, by a processing unit, a user query from a user device; identifying one or more contexts from a plurality of contexts, the identifying comprising comparing the user query with each context among the plurality of contexts, wherein each context of the plurality of contexts comprises values of one or more attributes characteristic to at least a part of the product; generating a prompt, the generating of the prompt comprising updating the user query based on the one or more contexts identified; and generating a natural language output on an output device, the generating of the natural language output comprising providing the prompt as input to a Large Language Model.

2. The method of claim 1, wherein identifying the one or more contexts from the plurality of contexts further comprises: generating a query embedding corresponding to the user query using a vectorization algorithm; comparing the query embedding with each context embedding of a plurality of context embeddings stored in a vector database, wherein each context embedding of the plurality of context embeddings corresponds to a context among the plurality of contexts; and identifying one or more closest matching context embeddings from the plurality of context embeddings, in relation to the query embedding, wherein the one or more closest matching context embeddings are indicative of one or more contexts associated with the user query.

3. The method of claim 2, further comprising generating the plurality of context embeddings, the generating of the plurality of context embeddings comprising: generating a list of the plurality of contexts associated with the product, the generating of the list of the plurality of contexts associated with the product comprising querying a product database; parsing the list of the plurality of contexts to demarcate each context of the plurality of contexts; and generating the context embeddings corresponding to the contexts in the structured list of the plurality of contexts using a vectorization algorithm.

4. The method of claim 2, wherein the one or more closest matching context embeddings in relation to the query embedding is identified based on a value of cosine similarity computed for the query embedding and each context embedding of the plurality of context embeddings in the vector database.

5. The method of claim 1 , wherein generating the prompt further comprises: transforming the one or more contexts into one or more contextual descriptions using a natural language processing technique, wherein each of the one or more contextual descriptions is a string comprising a natural language description of at least one context of the plurality of contexts; and augmenting the user query with the one or more contextual descriptions using a text data augmentation technique.

6. The method of claim 5, wherein transforming the one or more contexts into one or more contextual descriptions using the natural language processing technique comprises: selecting a description template from a plurality of description templates based on the attributes present in each of the one or more contexts identified; and generating the one or more contextual descriptions, the generating of the one or more contextual descriptions comprising populating one or more dynamic fields in the selected description template based on the attributes in each of the one or more contexts identified.

7. A system comprising: a processing unit; and a memory unit communicatively coupled to the processing unit, wherein the memory unit comprises a query management module that, when executed by the processing unit, is configured to: receive a user query from a user device; identify one or more contexts from a plurality of contexts, the identification of the one or more contexts from the plurality of contexts comprising comparison of the user query with each context among the plurality of contexts, wherein each context of the plurality of contexts comprises values of one or more attributes characteristic to at least a part of the product; generate a prompt, the generation of the prompt comprising update of the user query based on the one or more contexts identified; andprovide the prompt as input to a Large Language Model, such that a natural language output is generated, on an output device.

8. The system of claim 7, wherein the identification of the one or more contexts from the plurality of contexts comprises comparison of the user query with each context among the plurality of contexts, wherein the query management module, when executed by the processing unit, is further configured to: generate a query embedding corresponding to the user query using a vectorization algorithm; compare the query embedding with each of a plurality of context embeddings stored in a vector database, wherein each context embedding of the plurality of context embeddings corresponds to a context among the plurality of contexts; and identify one or more closest matching context embeddings from the plurality of context embeddings, in relation to the query embedding, and wherein the one or more closest matching context embeddings is indicative of one or more contexts associated with the user query.

9. The system of claim 8, wherein the query management module, when executed by the processing unit, is further configured to generate the plurality of context embeddings, the generation of the plurality of context embeddings comprising: generation of a list of the plurality of contexts associated with the product, the generation of the list of the plurality of contexts associated with the product comprising query of a product database; parse the list of the plurality of contexts, such that each of the contexts is demarcated; and generation of the plurality of context embeddings corresponding to each of the contexts in the structured list of the plurality of contexts using a vectorization algorithm.10 The system of claim 8, wherein the one or more closest matching context embeddings in relation to the query embedding is identified based on a value of cosine similarity computed for the query embedding and each of the context embeddings in the vector database.

11. The system of claim 8, wherein the generation of the prompt further comprises: transformation of the one or more contexts into one or more contextual descriptions using a natural language processing technique, wherein each of the one or more contextual descriptions is a string comprising a natural language description of at least one of the one or more contexts; and augmentation of the user query with the one or more contextual descriptions using a text data augmentation technique.

12. The system of claim 11, wherein the transformation of the one or more contexts into the one or more contextual descriptions using the natural language processing technique comprises: selection of a description template from a plurality of description templates based on the attributes present in each of the one or more contexts identified; and generation of the one or more contextual descriptions, the generation of the one or more contextual descriptions comprising population of one or more dynamic fields in the selected description template based on the attributes in each of the one or more contexts identified.

13. A non- transitory computer-readable storage medium that stores machine-readable instructions executable by a system, the machine-readable instructions comprising: receiving a user query from a user device; identifying one or more contexts from a plurality of contexts, the identifying of the one or more contexts from the plurality of contexts comprising comparing the user query with each context among the plurality of contexts, wherein each context of the plurality of contexts comprises values of one or more attributes characteristic to at least a part of the product; generating a prompt, the generating of the prompt comprising updating the user query based on the one or more contexts identified; and generating a natural language output on an output device, the generating of the natural language output comprising providing the prompt as input to a Large Language Model.

14. The non-transitory computer-readable storage medium of claim 13, wherein identifying the one or more contexts from the plurality of contexts comprises further comprises: generating a query embedding corresponding to the user query using a vectorization algorithm;comparing the query embedding with each of a plurality of context embeddings stored in a vector database, wherein each context embedding of the plurality of context embeddings corresponds to a context among the plurality of contexts; and identifying one or more closest matching context embeddings from the plurality of context embeddings, in relation to the query embedding, wherein the one or more closest matching context embeddings are indicative of one or more contexts associated with the user query.

15. The non-transitory computer-readable storage medium of claim 14, wherein the machine- readable instructions further comprise generating the plurality of context embeddings, the generating of the plurality of context embeddings comprising: generating a list of the plurality of contexts associated with the product, the generating of the list of the plurality of contexts associated with the product comprising querying a product database; parsing the list of the plurality of contexts, such that each of the contexts is demarcated; and generating the context embeddings corresponding to each of the contexts in the structured list of the plurality of contexts using a vectorization algorithm.

16. The non-transitory computer-readable storage medium of claim 14, wherein the one or more closest matching context embeddings in relation to the query embedding is identified based on a value of cosine similarity computed for the query embedding and each of the context embeddings in the vector database.

17. The non-transitory computer-readable storage medium of claim 13, wherein generating the prompt further comprises: transforming the one or more contexts into one or more contextual descriptions using a natural language processing technique, wherein each of the one or more contextual descriptions is a string comprising a natural language description of at least one of the contexts; and augmenting the user query with the one or more contextual descriptions using a text data augmentation technique.

18. The non-transitory computer-readable storage medium of claim 17, wherein transforming the one or more contexts into one or more contextual descriptions using the natural language processing technique comprises:selecting a description template from a plurality of description templates based on the attributes present in each of the one or more contexts identified; and generating the one or more contextual descriptions, the generating of the one or more contextual descriptions comprising populating one or more dynamic fields in the selected description template based on the attributes in each of the one or more contexts identified.

Images