Detecting compatibility mismatches through generative artificial intelligence

Abstract

Detecting compatibility mismatches through generative artificial intelligence is described. Compatibility data is obtained (e.g., by accessing a database). The compatibility data is associated with compatibility between items (e.g., an item with a vehicle category, or an item with another item), and includes a recommended compatibility list between items and user reported compatibility for at least one item. A machine learning model is generated for detecting compatibility mismatches between the first item and the second item and / or between the item and the vehicle category. At least a portion of the compatibility data is provided as input to generative artificial intelligence to generate a machine learning model. An update to the recommended compatibility list is determined based on the detected compatibility mismatch.

Description

Background Art

[0001] The computer system can implement machine learning techniques or artificial intelligence to generate outputs given prompts as input. For example, the computer system can utilize a generative artificial intelligence model to generate content, data, or output that is not explicitly programmed in the training data or provided to the generative artificial intelligence model. The generative artificial intelligence model is trained using deep learning techniques (e.g., neural networks) to detect patterns and structures within the training data. In some examples, the generative artificial intelligence model may include one or more machine learning models for generating outputs (e.g., text) in response to prompts or queries. The machine learning model can capture patterns and relationships in the data, enabling the model to detect context, generate coherent text, and perform various natural language processing tasks. Summary of the Invention

[0002] The compatibility detection system obtains, analyzes, and maintains recommended compatibilities for items. In some examples, the compatibility detection system obtains compatibility data describing the compatibility between an item and a device or equipment (e.g., a vehicle). For example, the item may be a replacement part for a vehicle component. The compatibility data may describe the compatibility between the item and the type of device or equipment (e.g., if the equipment is a vehicle, the brand and / or model and other identifying characteristics). If the item meets the operating standards of the device or equipment, the measurement standards for integration with the device or equipment, the manufacturing standards for integration with the device or equipment, and other parameters and standards, then the item may be compatible with the device or equipment. The compatibility detection system may generate a machine learning model for detecting compatibility mismatches between categories or types of equipment and items, for example, by inputting the compatibility data into generative artificial intelligence. If a compatibility mismatch is detected, the compatibility detection system may update the list of recommended compatibilities and / or may generate an indication of a compatibility mismatch.

[0003] The Summary introduces some concepts in a simplified form that will be further described below in the Detailed Description. Therefore, the Summary is not intended to identify essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter. BRIEF DESCRIPTION OF THE DRAWINGS

[0004] Detailed description of the embodiments is given with reference to the accompanying drawings.

[0005] Figure 1 is an illustration of an environment in an example implementation that is operable to employ the techniques described herein.

[0006] Figures 2 to 4 Depicted is compatibility mismatch detection logic in an example implementation for detecting compatibility mismatches through generative artificial intelligence.

[0007] Figure 5 and Figure 6 Depicted is an example of a user interface for detecting compatibility mismatches through generative artificial intelligence.

[0008] Figure 7 and Figure 8 Depicts a process in an example implementation for detecting compatibility mismatches through generative artificial intelligence.

[0009] Figure 9 An example of a system is shown that includes an example computing device, which is representative of one or more computing systems and / or devices that may implement the various techniques described herein. DETAILED DESCRIPTION

[0010] Overview

[0011] Compatibility mismatch detection for items using generative artificial intelligence is described. According to the described technology, a compatibility detection system inputs data into a generative artificial intelligence (e.g., one or more machine learning models) and receives an output from the generative artificial intelligence, the output indicating one or more items that are incompatible with a device or equipment, referred to as a compatibility mismatch. The machine learning model is trained to detect a compatibility mismatch between the equipment and the item indicated by the input data, and outputs information indicating the detected compatibility mismatch. For example, if the item is a part of a vehicle, the compatibility detection system may input data including a review of the item into the machine learning model. The machine learning model may process the data and output an indication of one or more vehicles (e.g., types of vehicles) that are incompatible with the item.

[0012] An online marketplace application can facilitate the exchange of items between users of the online marketplace application. For example, a seller can input data into the online marketplace application to be included in a detail page (listing) of an item for sale on the online marketplace application, and a buyer can provide user input to initiate a transaction to purchase the listed item for sale. In some examples, the online marketplace application can include a relatively large number of detail pages for items for sale (e.g., greater than a threshold, millions of items per year). One or more items can be docked with one or more other items. For example, an item can include parts or accessories for a vehicle, household appliance, electronic device, mechanical device, and / or other equipment. In some other examples, an item can include a vehicle, household appliance, electronic device, mechanical device, and / or other equipment.

[0013] In some cases, such as for items that interface with other items, compatibility criteria may exist between items based on the operation of the items when installed. The compatibility criteria for an item may include the capabilities of the item, the physical characteristics of the item, the mechanical characteristics of the item, and / or the software characteristics of the item, among other criteria. For example, a vehicle component may be compatible with one make and model of vehicle, but may not be compatible with another make and model of vehicle due to differences in the physical dimensions of the component's interface with the vehicle. In some other examples, a television may be compatible with the software system of one or more audio components of a first type, but may not be compatible with the software system of one or more audio components of a second type (e.g., from a different manufacturer) due to differences in operating system, system architecture, and / or security protocol differences.

[0014] An online marketplace application may maintain a list of items that meet compatibility criteria, referred to as a recommended compatibility list. For example, the list may include, for an item listed for sale in the online marketplace application, an indication of other items that are compatible with the corresponding item. The online marketplace application may display the recommended compatibility list, or may otherwise indicate the recommended compatibility list to a user who is purchasing an item, to reduce compatibility mismatches between items by informing the user of compatibility. However, due to the relatively large number of item detail pages on the online marketplace application, the recommended compatibility list may include errors and / or may not be maintained, resulting in outdated information. In some cases, a user (e.g., an item seller) may manually enter the item's compatibility information into the online marketplace application, which may introduce errors due to typos, incorrect data, etc. caused by manually entering the compatibility information. If a user obtains the item from the online marketplace application (e.g., a buyer purchases and receives the item), the user may not determine that the item is incompatible with another item (e.g., a device or appliance) until they try to use the item. If the item has a compatibility mismatch, the user can return the item and / or obtain a different item from the online marketplace application to replace it. Users returning items and / or obtaining replacement items may result in inefficient use of computing resources (e.g., processing and memory resources) due to increased search queries for the item and / or replacement item and increased exchange of transaction information when returning an item and / or ordering a replacement item. Furthermore, users may not be aware that the item is incompatible with another item and may use the item, resulting in premature deterioration of the item and / or malfunction of the item, among other drawbacks.

[0015] As described herein, in order to prevent inefficient use of computing resources and / or degradation and failure of items, a compatibility detection system may obtain information indicating a compatibility mismatch between one or more items, referred to as compatibility data. In some cases, the compatibility data may include an explicit indication of a compatibility mismatch, such as user input including a comment indicating that the item is incompatible with another item (e.g., a device or appliance). In some other cases, the compatibility data may include an implicit indication of a compatibility mismatch. For example, the compatibility data may include an indication that a user returned an item. The compatibility detection system may determine, based on the user data of the user, that the item is docked with one or more other items associated with the user, and may infer that the item was returned due to incompatibility with the other items (e.g., due to language in the item review, historical transaction data of purchases made by the user, and / or compatibility information of the item, among other factors).

[0016] In some examples, the compatibility detection system can train one or more machine learning models, such as one or more machine learning models implemented by generative artificial intelligence, to detect compatibility mismatches based on the compatibility data. For example, the compatibility detection system can use a recommended compatibility list between items and one or more compatibilities between items reported by users as training data to train and / or fine-tune the machine learning model. Figure 1 and Figure 2 Further detailed description. Once the machine learning model is trained, the compatibility detection system may provide compatibility data as an input to the machine learning model and may receive as an output from the machine learning model an indication of one or more compatibility mismatches between items. In some examples, the compatibility data may include audio, text, and / or visual data input by a user, such as one or more reviews and / or other comments associated with the items. In some cases, the compatibility detection system may instruct the computing device to display an indication of the compatibility mismatch (e.g., via a graphical user interface (GUI)) so that the user can update the recommended compatibility list between the items. In some other cases, the compatibility detection system may automatically (e.g., without user input or independently of user input) update the recommended compatibility list between the items.

[0017] In some aspects, the technology described herein relates to a computer-implemented method comprising: obtaining compatibility data between a plurality of items and a plurality of vehicle categories, the compatibility data comprising a recommended compatibility list between the plurality of items and the plurality of vehicle categories, and user-reported compatibility between at least one of the plurality of items and at least one of the plurality of vehicle categories; generating a machine learning model for detecting compatibility mismatches between vehicle categories and items based on providing at least a portion of the compatibility data as input to generative artificial intelligence; and determining an update to the recommended compatibility list based on detecting the compatibility mismatch using the machine learning model.

[0018] In some aspects, the techniques described herein relate to a computer-implemented method wherein generating a machine learning model further comprises modifying one or more parameters of the machine learning model until a performance criterion associated with the machine learning model is met.

[0019] In some aspects, the techniques described herein relate to a computer-implemented method in which modification of one or more parameters of a machine learning model is based on providing additional compatibility data as input to a generative artificial intelligence.

[0020] In some aspects, the technology described herein relates to a computer-implemented method wherein detecting a compatibility mismatch further comprises providing additional compatibility data as input to the machine learning model based on performance criteria associated with the machine learning model being satisfied, the additional compatibility data comprising at least one user-reported compatibility corresponding to a vehicle class and an item.

[0021] In some aspects, the techniques described herein relate to computer-implemented methods wherein the performance criteria include one or more of a threshold accuracy metric associated with the machine learning model, a threshold recall metric associated with the machine learning model, a threshold F1 score associated with the machine learning model, or a threshold return rate associated with a plurality of items.

[0022] In some aspects, the technology described herein relates to a computer-implemented method wherein detecting a compatibility mismatch further comprises receiving an indication of a compatibility mismatch as an output from a machine learning model based on providing at least one user-reported compatibility corresponding to a vehicle class and an item as input to the machine learning model.

[0023] In some aspects, the techniques described herein relate to a computer-implemented method further comprising sending an indication of an update to the recommended compatibility list to a device for display to a user.

[0024] In some aspects, the techniques described herein relate to a computer-implemented method further comprising, in response to the indication, receiving a request to perform an update to the recommended compatibility list.

[0025] In some aspects, the technology described herein relates to a computer-implemented method that further includes: performing an update to a recommended compatibility list, wherein the update to the recommended compatibility list includes one or more of: removing a vehicle category from the recommended compatibility list of an item, or adding a vehicle category to the recommended compatibility list of an item.

[0026] In some aspects, the technology described herein relates to a computer-implemented method further comprising: sending a request for compatibility data to a device for display to a user; receiving user input corresponding to one or more of the recommended compatibility list or user-reported compatibilities in response to the request; and storing one or more of the recommended compatibility list or user-reported compatibilities in a database based on processing the user input to determine the user-reported compatibilities.

[0027] In some aspects, the techniques described herein relate to a computer-implemented method wherein processing further comprises parsing user input to determine respective string values ​​and respective character values ​​corresponding to one or more of at least one item or at least one vehicle category.

[0028] In some aspects, the technology described herein relates to a system comprising: one or more processors; and a computer-readable storage medium storing instructions executable by the one or more processors to perform operations comprising: obtaining compatibility data between a plurality of items, the compatibility data comprising a recommended compatibility list between the plurality of items and a user-reported compatibility associated with at least one of the plurality of items; generating a machine learning model for detecting a compatibility mismatch between a first item in the plurality of items and a second item in the plurality of items based on providing at least a portion of the compatibility data as input to generative artificial intelligence; and determining an update to the recommended compatibility list based on detecting the compatibility mismatch using the machine learning model.

[0029] In some aspects, the techniques described herein relate to systems wherein, to generate a machine learning model, the operations further comprise modifying one or more parameters of the machine learning model until a performance criterion associated with the machine learning model is met.

[0030] In some aspects, the techniques described herein relate to systems in which modification of one or more parameters of a machine learning model is based on providing additional compatibility data as input to a generative artificial intelligence.

[0031] In some aspects, the technology described herein relates to a system wherein, to detect a compatibility mismatch, the operations further comprise providing additional compatibility data as input to the machine learning model based on performance criteria associated with the machine learning model being satisfied, the additional compatibility data comprising at least one user-reported compatibility corresponding to the first item and the second item.

[0032] In some aspects, the technology described herein relates to a system wherein, to detect a compatibility mismatch, the operations further comprise receiving an indication of a compatibility mismatch as an output from a machine learning model based on providing at least one user-reported compatibility corresponding to a first item and a second item as an input to the machine learning model.

[0033] In some aspects, the techniques described herein relate to a system wherein the operations further comprise sending an indication of an update to the recommended compatibility list to a device for display to a user.

[0034] In some aspects, the technology described herein relates to a system, wherein the operations further comprise: performing an update to a recommended compatibility list, wherein the update to the recommended compatibility list comprises one or more of: removing the first item from the recommended compatibility list of the second item, removing the second item from the recommended compatibility list of the first item, adding the first item to the recommended compatibility list of the second item, or adding the second item to the recommended compatibility list of the first item.

[0035] In some aspects, the technology described herein relates to a system wherein the operations further comprise: sending a request for compatibility data to a device for display to a user; receiving user input corresponding to one or more of a recommended compatibility list or user-reported compatibilities in response to the request; and storing one or more of the recommended compatibility list or user-reported compatibilities in a database based on processing the user input to determine the user-reported compatibilities.

[0036] In some aspects, the technology described herein relates to one or more computer-readable storage media, wherein the one or more computer-readable storage media include computer-executable instructions stored thereon, and the computer-executable instructions, in response to execution by one or more processors, perform operations including the following: obtaining compatibility data between multiple items and multiple vehicle categories, the compatibility data including a recommended compatibility list between the multiple items and the multiple vehicle categories, and user-reported compatibility between at least one of the multiple items and at least one of the multiple vehicle categories; generating a machine learning model for detecting compatibility mismatches between vehicle categories and items based on providing at least a portion of the compatibility data as input to generative artificial intelligence; and determining an update to the recommended compatibility list based on detecting the compatibility mismatch using the machine learning model.

[0037] In the following discussion, an exemplary environment in which the techniques described herein may be employed is first described. Next, examples of implementation details and procedures that may be performed in the exemplary environment and in other environments are described. Performance of the exemplary procedures is not limited to the exemplary environment, nor is the exemplary environment limited to performance of the exemplary procedures.

[0038] Environment Examples

[0039] Figure 1 is an illustration of an environment 100 in an example implementation operable to employ the techniques described herein. The environment 100 includes one or more computing devices 102 and a compatibility detection system 104. In some examples, the compatibility detection system 104 includes a generative artificial intelligence manager 106 to train, fine-tune, and / or implement generative artificial intelligence. In one or more implementations, the computing device 102 and the compatibility detection system 104 are communicatively coupled to each other via a network 108. One example of the network 108 is the Internet, although in various implementations, the computing device 102 and the compatibility detection system 104 can be communicatively coupled using one or more different connections or different networks (e.g., a wireless network).

[0040] Although the compatibility detection system 104 is depicted in the environment 100 as being separate from the computing device 102, in one or more implementations, the compatibility detection system 104, in its entirety or in its various portions, may be implemented at or by the computing device 102. In at least one implementation, for example, at least a portion of the compatibility detection system 104 is implemented by an application 110 of the computing device 102 and / or by using various resources of the computing device 102, such as hardware resources, an operating system, firmware, and the like. Alternatively or in addition, the compatibility detection system 104 is implemented by server-based storage resources, processing resources, and the like of a device other than the computing device 102. For example, at least a portion of the compatibility detection system 104 is implemented using a third-party service (e.g., a web service platform that provides one or more hardware and / or other computing resources to support the provision of services by a web service provider). In a variation, the compatibility detection system 104, in its entirety or in its various portions, is implemented at or by a user's device (e.g., a mobile device, a laptop computer, a wearable device, or any other device).

[0041] The computing device 102 that implements the environment 100 can be configured in a variety of ways. For example, the computing device 102 can be configured as a desktop computer, a laptop computer, a mobile device (e.g., assuming a handheld configuration such as a tablet computer or mobile phone), an Internet of Things (IoT) device, a wearable device (e.g., a smart watch, ring, or smart glasses), an augmented reality and / or virtual reality device (e.g., smart glasses), a server, etc. Thus, the computing devices 102 range from full-resource devices with large amounts of memory and processor resources to low-resource devices with limited memory and / or processing resources. Although the examples discussed below refer to the singular computing device 102, the computing device 102 can also represent a plurality of different devices, such as a plurality of servers of a server farm for performing operations "on the cloud," as described with respect to FIG. Figure 8 Further described.

[0042] In at least one implementation, application 110 supports data communication between computing device 102 and compatibility detection system 104 across network 108. By supporting such data communication, application 110 provides access to online marketplace 112 to the respective user of computing device 102 (as well as users of other computing devices 102). For example, computing device 102 receives data from compatibility detection system 104. Based on the received data, application 110 causes various systems of computing device 102 to output a user interface for online marketplace 112, such as by displaying the user interface via a display device or providing voice-based access to the user interface.

[0043] Through user interaction with computing device 102, application 110 receives user input via one or more user interfaces of online marketplace 112. Examples of such input include, but are not limited to, receiving touch input associated with portions of a displayed user interface, receiving one or more voice commands or other audio input, receiving typed input (e.g., via a physical or virtual ("soft") keyboard), receiving mouse or stylus input, and the like. One example of application 110 is a browser that is operable to navigate to a website for online marketplace 112, display pages of the website, and facilitate user interaction with the web pages of the website for online marketplace 112. Another example of application 110 is a web-based computer application of online marketplace 112, such as a mobile application or a desktop application. Without departing from the spirit or scope of the technology described herein, application 110 can be configured in different ways that enable users to interact with their computing devices 102 and, by extension, perform actions on online marketplace 112.

[0044] In one or more implementations, a user registers to obtain a corresponding user account with the online marketplace 112. Such registration may include, for example, providing an email address and establishing a username and password combination. After registration, the computing device 102 facilitates logging into the user account or otherwise authenticating the user account in various ways, such as by: receiving a username and a matching password; receiving biometric information (e.g., at least one captured image of a face or captured information of another body part (e.g., a thumb or finger)) that is appropriately matched to stored biometric information associated with the user account; and the like.

[0045] Broadly speaking, the online marketplace 112 is configured to generate detail pages for items and expose these detail pages (e.g., publish them) to one or more computing devices 102. For example, the online marketplace 112 may generate detail pages for items for sale and expose these detail pages to the computing devices 102 so that users of the computing devices 102 can interact with the detail pages via a user interface to initiate transactions (e.g., purchase, add to a wish list, share, etc.) related to one or more items corresponding to the detail pages. In accordance with the described technology, the online marketplace 112 is configured to generate detail pages for one or more items of various types of physical goods or property (e.g., parts of a device or appliance, accessories for a device or appliance, clothing and / or clothing accessories, collectibles, furniture, decorative items, textiles, luxury goods, electronics, real estate, a physical computer-readable storage device storing one or more video games, etc.), services (e.g., child care, dog walking, house cleaning, etc.), digital items that can be downloaded via the network 108 (e.g., digital images, digital music, digital videos), and blockchain-backed assets (e.g., non-fungible tokens (NFTs)), to name a few examples.

[0046] The online marketplace 112 is configured to generate detail pages for items that can interface with aspects of other items (e.g., devices and / or appliances). In some cases, the online marketplace 112 may include detail pages for a part of another item, an accessory for another item, and / or an item that can be configured to interact or interface with the systems of other items. For example, the online marketplace 112 may include detail pages for parts and / or accessories for electrical and / or mechanical devices or equipment (e.g., vehicles, cellular devices or other electronic devices, laptop computers, or appliances), to name a few examples.

[0047] In some cases, vehicle components may interface with other components of the vehicle, including components that are part of the vehicle's electrical and / or mechanical systems. For example, the online marketplace 112 may include one or more detail pages for brake pads for a vehicle. The vehicle's braking system may include one or more components including brake pads, brake rotors or discs, calipers, brake fluid, a backing plate that provides structural support for the brake pads, and other components. During installation of the brake pads and during operation of the vehicle, the brake pads may interface with other components of the braking system. For example, the brake pads may be physically coupled to the caliper and backing plate to ensure that the brake pads are positioned in physical contact with the brake rotor or disc, thereby generating friction for stopping the vehicle during operation.

[0048] While brake pads are provided as an example, there are any number of items listed on the online marketplace 112 that interface with other items. For example, a television listed on the online marketplace 112 may interface with one or more input / output (I / O) devices, including an audio output system, a gaming control system, a home automation device, etc. Similarly, a home appliance may have one or more components that interface with each other, including a water filter for a water system in a refrigerator device. Thus, an item interfacing with another item, system, device, or apparatus may include physical coupling, electrical or electronic coupling, and / or mechanical coupling.

[0049] In some examples, the compatibility of an item can depend on an interface, such as the item's mechanical operability, the item's electrical operability, or both. An item may be incompatible with another item if it negatively impacts the operation of the other item (e.g., by reducing the item's operational efficiency or by causing the item to become inoperable). In some examples, the threshold impact on the operation of an item for compatibility can be defined by the manufacturer of the item. For example, one or more vehicle manufacturers may specify a set of tolerances or thresholds for the operation of various components of a vehicle. The incompatibility of an item with another item can be referred to as a compatibility mismatch.

[0050] In some examples, the compatibility mismatch can be attributed to one or more capabilities of the item (e.g., mechanical and / or electrical capabilities), one or more physical characteristics of the item, one or more mechanical characteristics of the item, and / or one or more software characteristics of the item, as well as other characteristics of the item. The capabilities of the item may include the ability to wirelessly connect to a network or another item (e.g., a device or appliance), the processing capabilities of the item, the storage capabilities of the item, or the display capabilities of the item, as well as other capabilities. The physical characteristics of the item may include one or more of the following: an I / O interface of the item, one or more physical dimensions that define the size and / or shape of the item, one or more physical coupling points of the item (e.g., coupling locations, including the locations of screws, bolts, rivets, clamps, ties, etc.), or one or more physical features of the item, as well as other physical characteristics. The one or more software characteristics of the item may include the software system of the item (e.g., an operating system or other system), the software architecture implemented at the item, or the secure software interface implemented at the item, as well as other software characteristics.

[0051] For example, a television may not have wireless connectivity (e.g., Wi-Fi and / or Bluetooth) for interfacing with one or more other systems (including an audio output system). Consequently, an audio output system that operates using a wireless connection may be incompatible with the television, while an audio output system with a wired connection may be compatible with the television. In some other examples, brake pads may be compatible with a vehicle's braking system if their physical dimensions meet one or more threshold physical dimensions for interfacing with the brake rotors or discs and / or calipers and other components of the vehicle's braking system. In another example, a thermostat may operate using a software system that is compatible with one set of home automation devices but incompatible with another set of home automation devices. An item may meet one or more compatibility criteria or conditions to be compatible with another item. In some cases, an item may have capabilities, physical characteristics, mechanical characteristics, and / or software characteristics that meet the compatibility criteria or conditions. For example, an item may have wireless connectivity, meet one or more thresholds related to physical dimensions and / or threshold tolerances for coupling point locations, have compatible software systems, and so on. In some other cases, an item may not have one or more of the capabilities, physical properties, mechanical properties, and / or software properties that meet the compatibility criteria or conditions and therefore may not be compatible or may have a compatibility mismatch.

[0052] In the illustrated environment 100, the online marketplace 112 includes a storage device 114, which is depicted as storing compatibility information. Compatibility information includes, but is not limited to, recommended compatibilities 116, user-reported compatibilities 118, and, in some cases, a recommended compatibility list 120. Additionally or alternatively, the storage device 114 of the online marketplace 112 stores information related to one or more detail pages of items for sale, including real-time detail page data for detail pages on the online marketplace 112, purchase history and / or transaction data related to purchases by users (e.g., based on user identifiers from user accounts, including account usernames or other identifiers), information entered by users, information about users inferred based on transaction data, and the like. The storage device 114 may represent one or more databases and / or other types of storage devices capable of storing compatibility information. Examples of the storage device 114 include, but are not limited to, mass storage devices and virtual storage devices. For example, in one or more implementations, the storage device 114 may be virtualized across multiple data centers and / or be cloud-based storage. The compatibility detection system 104 can implement the online marketplace 112 by using a server that executes stored instructions to deploy various services of the compatibility detection system 104 so that these services perform a large amount of computation to effectively provide the functionality described above and below. It should be understood that the online marketplace 112 can include more, fewer, or different components without departing from the spirit or scope described herein.

[0053] In variations, computing device 102 can use I / O manager 122 to collect user input and provide information to the user. The I / O manager can configure computing device 102 to display or otherwise present controls that can be selected by the user to provide user input 124 and / or prompts for requesting user input 124. In some examples, I / O manager 122 displays the controls and / or prompts to the user via a GUI of computing device 102. In some other examples, the I / O manager displays the requests to the user via a GUI of another device communicatively coupled to computing device 102 (e.g., another computing device 102 coupled to computing device 102 via network 108). The I / O manager can visually display the controls and / or prompts, can emit audio versions of the controls and / or prompts via an audio output component, and the like.

[0054] In some examples, I / O manager 122 receives user input 124 via one or more input components of a user interface. User input 124 can be in response to a request for user input 124 from a computing device and / or can be initiated by a user of computing device 102. Examples of such user input 124 include, but are not limited to, receiving touch input associated with a displayed portion of a user interface, receiving one or more voice commands, receiving typed input (e.g., via a physical or virtual ("soft") keyboard), receiving mouse or stylus input, and the like.

[0055] The user may enter information via the interactive elements of the GUI (e.g., fill in text elements, select selectable elements, etc.). The information may include details about items associated with the user (e.g., owned by the user and / or otherwise interacted with by the user). For example, the user may enter information identifying the type of vehicle associated with the user, including the make of the vehicle, the year of manufacture of the vehicle, the model of the vehicle, the trim level of the vehicle, or the engine of the vehicle, among other information, which may be related to the vehicle. Figure 5 Detailed description is provided below. The I / O manager 122 may store the information entered by the user (e.g., in a local database) for access upon request by the compatibility detection system 104. Additionally or alternatively, the computing device 102 may send the user-entered information to the compatibility detection system 104 for storage at the storage device 114. In some examples, the compatibility detection system 104 may store the user-entered information with an associated user identifier (e.g., a user account of the user who entered the information).

[0056] In one or more implementations, the online marketplace 112 may be accessed by decentralized computing devices corresponding to “clients” of the online marketplace 112 (e.g., users with accounts on the online marketplace 112). In some cases, there may be different types of user accounts registered with the online marketplace 112. For example, the different types of user accounts may include seller user accounts and buyer user accounts. The seller user account may be used by a user to enter information in order to list an item for sale via the online marketplace 112. The buyer user account may be used by a user to enter transaction information in order to purchase an item listed for sale via the online marketplace 112. The computing device 102 may display a different user interface of the online marketplace 112 for the buyer account than for the seller account. In variations, the user account may be both a buyer account and a seller account, and the computing device 102 may include a control that can be selected to switch between different user interfaces of the online marketplace 112. The user interface may include different GUI displays, referred to as a seller display and a buyer display. For the seller display, computing device 102 may display: one or more controls selectable by the user to provide information related to the detail page of the item for sale; and / or prompts to the user to provide information related to the detail page of the item for sale. The controls may include options for selecting a type and / or category of the item, selecting from a list of item characteristics based on input information about the item, and the like. The prompts may include prompts for providing information about the item for sale, including a description of the item, the category of the item, the price of the item, and / or the condition of the item (e.g., new, used, etc.), as well as other information.

[0057] For buyer display, computing device 102 may display a search function for searching for items from an item detail page. Once a user enters a search query for an item, computing device 102 may display query results that may be selected by the user. The query results may include items for sale whose titles and / or information are within a threshold match of the search query (e.g., a threshold number of characters and / or string values ​​of the search query that match the item title and / or other information about the item). If computing device 102 receives a selection of a query result, computing device 102 may display additional information about the item corresponding to the query result and one or more controls and / or prompts for the user to select an item to purchase, which will be related to Figure 6 Described in further detail.

[0058] In some cases, the user account of the user, labeled "A," interacting with the online marketplace 112 via the computing device 102 may be a seller user account. The application 110 may collect data related to the online marketplace 112 from the user. For example, the application 110 may collect information about one or more items to be listed for sale at the online marketplace 112. In some examples, the information about the one or more items may include a recommended compatibility list 120 for the one or more items. The recommended compatibility list 120 may include a list of one or more items (e.g., devices and / or appliances) that are compatible with the items in the respective items listed by the seller user account. In variations, the seller user account maintains the recommended compatibility list 120. In some other variations, another system operating the online marketplace 112 and / or the compatibility detection system 104 maintains the recommended compatibility list 120.

[0059] In some other cases, the user account of the user, labeled "B," who interacts with the online marketplace 112 via the computing device 102 may be a buyer user account. The application 110 may collect data related to the online marketplace 112 from the user. For example, the computing device 102 may collect user input 124 via an interactive GUI of the computing device 102. In some examples, the user input 124 may explicitly indicate the compatibility of items for sale via the online marketplace 112, which is referred to as user-reported compatibility 126. That is, the user input 124 may include textual, audio, and / or visual user input 124 indicating that the item is incompatible with another item. In some other examples, the user input 124 may implicitly indicate user-reported compatibility 126 of items for sale via the online marketplace 112. That is, the user input 124 may include textual, audio, and / or visual user input 124 related to the item, and the compatibility detection system 104 may determine, based on the user input 124, that the item is incompatible with the other item.

[0060] The online marketplace 112 may store transaction history data (e.g., at storage 114) indicating previous purchases by one or more users of the online marketplace 112. In some examples, the online marketplace 112 may store transaction data related to item purchases. For example, a user may purchase a television, and the online marketplace 112 may store transaction data indicating that the user (e.g., with an associated user account and / or user identifier) ​​purchased the television. The compatibility detection system 104 may access the transaction data (including historical transaction data) to determine whether the item is compatible with the device using implicit indications of item compatibility. In some cases, the user-reported compatibility 126 may indicate that the user returned an item that was capable of docking with another item previously purchased by the user. Additionally or alternatively, the user-reported compatibility 126 may indicate that the user returned the item and ordered a different item related to the item (e.g., an item of a different type, a similar item from a different manufacturer, a similar item with different characteristics, etc.). The compatibility detection system 104 may analyze the transaction history data and the user-reported compatibility 126 to determine that the item has a compatibility mismatch with the item.

[0061] For example, the compatibility detection system 104 may determine based on transaction history that a user purchased a television and an audio system that is configured to be docked with the television. The compatibility detection system 104 may determine that the television and the audio system are configured to be docked based on descriptions in the item detail pages of the television or the audio system, based on recommended compatibility lists 120, and other data sources. The compatibility detection system 104 may receive user input 124 indicating that the audio system is incompatible with the television (e.g., an event that triggers the return of the audio system may be detected, an event that triggers the purchase of a different audio system may be detected, a review of the audio system that indicates a compatibility mismatch may be received, and / or another implicit or explicit indication of a compatibility mismatch). The compatibility detection system 104 may analyze the user input 124, transaction history, and / or information from the item detail page and other data to detect a compatibility mismatch between the audio system and the television, and may additionally or alternatively detect that an alternative audio system (e.g., an audio system that the user ordered as a replacement) is compatible with the television.

[0062] In some examples, user input 124 may include one or more reviews and / or comments related to an item listed via online marketplace 112. For example, computing device 102 may display a request via a GUI of computing device 102 for a user to provide user input 124. Computing device 102 may display the request based on a trigger condition (e.g., a user-initiated return of an item, a user-initiated trade of an item, and / or a threshold duration from a user-initiated return of an item or a user-initiated trade of an item). The request may be in the form of a message, for example, via a communication platform (e.g., an email messaging platform, a text messaging platform, etc.) and / or via a user interface of online marketplace 112 generated by application 110. The request may include an indication that, if the user initiates a return of the item, the user provides a return reason and / or may provide a selectable list of return reasons. The list of return reasons may include a compatibility mismatch and other return reasons.

[0063] In some examples, the online marketplace 112 may have a relatively large number of items listed for sale (e.g., greater than a threshold, millions of item detail pages for sale). The computing device 102 may display a recommended compatibility list 120 to indicate whether an item is compatible with one or more devices. Additionally or alternatively, the online marketplace 112 may include a filtering feature, such as a reference to a list of items that are compatible with the device. Figure 5 As described, the filtering feature allows a user to filter search results for items that are compatible with a defined or specified item. However, due to the relatively large number of items listed for sale, the recommended compatibility list 120 for each item may be outdated, inaccurate (e.g., may include errors), and / or may be incomplete. For example, the recommended compatibility list 120 for a corresponding item may erroneously indicate that one or more other items are compatible with the item. Additionally or alternatively, the recommended compatibility list 120 for a corresponding item may not include one or more items that are compatible with another item. In some examples, if a seller user account is responsible for maintaining the recommended compatibility list 120, the user of the seller user account may manually enter compatible items for the corresponding item. Due to the relatively large number of items listed for sale at the online marketplace 112, the user's manual entry of compatible items may result in errors in the recommended compatibility list 120.

[0064] In some other examples, computing device 102 and / or compatibility detection system 104 may generate recommended compatibility list 120 based on data collected by computing device 102 and / or compatibility detection system 104 (e.g., information from an item's user manual, manufacturer recommendations for the item, and / or other available information related to the item). However, computing device 102 and / or compatibility detection system 104 may not be able to verify that recommended compatibility list 120 (e.g., input by a user and / or generated based on data collected by computing device 102 and / or compatibility detection system 104) is correct due to a lack of available data and / or the relatively large number of items listed for sale at online marketplace 112. Consequently, a user may purchase an item that interfaces with another item and determine that the items are incompatible. Due to the compatibility mismatch, the user may return the item and / or exchange it for an alternative item. User returns or exchanges of items may increase computing resource usage due to increased signaling, processing, and memory usage for additional searches and transactions, thereby resulting in computational inefficiencies for computing device 102 and / or online marketplace 112. For example, the user may search for additional items, resulting in additional use of processing and memory resources at computing device 102, as well as high signaling overhead due to communications between computing device 102 and online marketplace 112 to populate the search results for display at computing device 102. In some other examples, to return an item, the user may enter information into computing device 102 to initiate the return, resulting in use of processing and memory resources of computing device 102 and / or online marketplace 112 (e.g., due to relisting the item for sale once the return is complete), as well as high signaling overhead due to communications between computing device 102 and online marketplace 112 to initiate and complete the return process.

[0065] Furthermore, in some examples, a compatibility mismatch may not prevent a user from using the item. For example, one or more physical dimensions of the item may be within a tolerance or threshold for mounting the item on a device, but may not be within a tolerance or threshold for device operation. Consequently, a user may attempt to operate the device after installing an item with a compatibility mismatch, resulting in degradation of the device and / or item and / or premature failure of the device and / or item.

[0066] To prevent or reduce compatibility mismatches between an item listed at the online marketplace 112 and one or more other items, the compatibility detection system 104 can implement generative artificial intelligence to detect compatibility mismatches and can use the detected compatibility mismatches to update the recommended compatibility list 120 at the online marketplace 112. In one or more implementations, the compatibility detection system 104 implements a generative artificial intelligence manager 106 that trains, fine-tunes, and / or implements a machine learning model 128. In one or more implementations, the compatibility detection system 104 can implement the generative artificial intelligence manager 106 by using a server that executes stored instructions to deploy various services of the compatibility detection system 104, so that these services perform a large amount of computation to effectively provide the functionality described above and below. It should be understood that the generative artificial intelligence manager 106 can include more, fewer, or different components without departing from the spirit or scope described herein.

[0067] In this example, the generative AI manager 106 includes or otherwise has access to model training logic 130. Generative AI manager 106 can utilize model training logic 130 to train or fine-tune one or more machine learning models 128. Example machine learning models include, but are not limited to, large language models and / or conditional generative models. Large language models are AI models designed to generate natural language. In some examples, large language models are pre-trained on diverse text datasets to learn the structure, grammar, and semantics of a language. Conditional generative models are AI models designed to generate output based on one or more input conditions or labels. Machine learning models 128 used for generative AI can be built using deep learning techniques and can have a greater number of parameters than other AI models. Unlike traditional AI systems that rely on rule-based or deterministic approaches, generative AI employs algorithms and models that can autonomously produce output that closely resembles human-generated content. These algorithms are designed to learn patterns and structure from existing data and then use this learned information to generate new content that is coherent, relevant, and contextually appropriate.

[0068] Although techniques utilizing generative artificial intelligence are described, in variations, different types of artificial intelligence can be utilized without departing from the spirit or scope of the described techniques. For example, the generative artificial intelligence manager 106 can train and / or fine-tune any number of machine learning models 128, for example, to generate a distributed network of machine learning models 128. The distributed network can include a large language model that prompts other machine learning models 128 that generate results. The machine learning models 128 that generate results can be trained or fine-tuned to detect compatibility mismatches between items and devices.

[0069] In some examples, a machine learning model 128 designed for generative artificial intelligence can be fine-tuned or trained for a specific application using data from a specific application. Fine-tuning the machine learning model 128 can include updating an existing or pre-trained machine learning model 128 by training the machine learning model 128 with a more specific data set to adapt the machine learning model 128 to the task or context. The model training logic 130 is configured to access a storage device 132, which is depicted as maintaining the compatibility data 134, by executing a retrieval command for obtaining the compatibility data 134. The storage device 132 can represent one or more databases and / or other types of storage devices capable of storing the compatibility data 134. Examples of the storage device 132 include, but are not limited to, mass storage devices and virtual storage devices. For example, in one or more implementations, the storage device 132 can be virtualized across multiple data centers and / or be a cloud-based storage device.

[0070] In some examples, the compatibility data 134 includes user-reported compatibilities 126. For example, the computing device 102 may send data including the user-reported compatibilities 126 to the compatibility detection system 104. The compatibility detection system 104 may filter the data to determine the user-reported compatibilities 126 and may store the user-reported compatibilities 126 at the storage device 132. The data may include one or more comments associated with the item, one or more reviews associated with the item, and a reason why the item was returned. The compatibility detection system 104 may pre-process the data by filtering the data to remove one or more values ​​(e.g., character and / or string values) from the data that do not indicate compatibility information related to whether the item is compatible with another item and / or device (or vice versa), which will be related to Figure 3 Additionally or alternatively, preprocessing can include updating the format of the data to a defined format for input to the machine learning model 128. The compatibility data 134 can include a recommended compatibility list 120, which the generative artificial intelligence manager 106 can obtain from the online marketplace 112.

[0071] Once the compatibility data 134 is obtained by the generative artificial intelligence manager 106 of the compatibility detection system 104, the compatibility detection system 104 may instruct the generative artificial intelligence manager 106 to generate one or more trained machine learning models for detecting compatibility mismatches between items and devices. For example, the model training logic 130 may include instructions for inputting the compatibility data 134 into one or more machine learning models 128 during a fine-tuning process to generate trained machine learning models that will be used to determine compatibility mismatches between items and devices. Figure 3Detailed description is provided below. The machine learning model 128 is trained to output an indication of one or more compatibility mismatches between an item and another item or device based on input compatibility data 134 (e.g., some or all of the user-reported compatibilities 126). A compatibility mismatch can include an item that is listed as compatible with another item or device but is not compatible with the other item or device (or vice versa), and / or an item that is not listed as compatible with another item or device but is compatible with the other item or device (or vice versa). The machine learning model 128 can include any number of models, including generative artificial intelligence models.

[0072] In some examples, the model training logic 130 may include instructions for continuing to train the machine learning model 128 until a threshold amount or number of data is received indicating user-reported compatibilities 126 corresponding to compatibility mismatches. For example, the generative artificial intelligence manager 106 may continue to input a training data set into the machine learning model 128 to train and / or fine-tune the machine learning model 128 until the number of detected compatibility mismatches meets (e.g., is less than) a threshold value. That is, the generative artificial intelligence manager 106 may continue to train the machine learning model 128 until the number of compatibility mismatches detected from reviews, comments, and / or other user-reported data falls below a threshold value. The compatibility detection system 104 may define the threshold value using one or more factors, including the number of listings in the online marketplace 112 that have compatibility information, the total amount or number of data entries from user input, or any other factor.

[0073] In some cases, the generative artificial intelligence manager 106 includes compatibility detection logic 136 for detecting a compatibility mismatch between an item and another item and / or device using the machine learning model 128 trained by the model training logic 130. The generative artificial intelligence manager 106 may input a prompt 138 to the machine learning model 128, and the machine learning model 128 may generate a compatibility mismatch indication 140 as an output. In some examples, the prompt 138 includes data, such as user-reported compatibilities 126 for one or more items and / or devices. The user-reported compatibilities 126 may include a collection of textual, audio, and / or visual data that explicitly or implicitly indicates that an item listed for sale on the online marketplace 112 has a compatibility mismatch. Due to the relatively large number of items listed for sale on the online marketplace 112, it may be difficult for a user to manually sort through the data comprising the user-reported compatibilities 126. Thus, by implementing the machine learning model 128, the compatibility detection system 104 is able to identify and / or detect compatibility mismatches and generate a compatibility mismatch indication 140 for reporting to a user and / or for updating the recommended compatibility list 120 (e.g., without requiring the user to manually update the recommended compatibility list 120). For example, the compatibility detection system 104 can report the compatibility mismatch indication 140 to a user (e.g., a seller) via the computing device 102. If the user maintains a recommended compatibility list 120, the user can update the recommended compatibility list 120 based on the compatibility mismatch indication 140. If the compatibility detection system 104 maintains a recommended compatibility list 120, the compatibility detection system can automatically update the recommended compatibility list 120 (e.g., without requiring or independent of user input) and can indicate the update to the list to the user (e.g., a seller).

[0074] Although not depicted, in some examples, the compatibility detection system 104 and the computing device 102 implement a communication manager to support data communication between the computing device 102 and the compatibility detection system 104 across the network 108. By supporting such data communication, the communication manager provides the computing device 102 with access to compatibility mismatch detection technologies that would otherwise be inaccessible to the compatible computing device 102. In one or more implementations, the compatibility detection system 104 and the computing device 102 use the communication manager to exchange data (e.g., data including user-reported compatibilities 126, recommended compatibility lists 120, and / or compatibility mismatch indications 140). The I / O manager 122 of the computing device 102 can display an indication of one or more updates to the compatibility mismatch indication 140 and / or the recommended compatibility list 120 to the user via a GUI of the computing device 102, via an audio output component of the computing device 102, via a tactile component of the computing device 102 (e.g., through vibration of the computing device), or any combination thereof.

[0075] The compatibility detection system 104 can use a machine learning model 128 to build a self-learning system that analyzes user input 124 (e.g., return reviews, comments, feedback, etc.) to improve the accuracy of the recommended compatibility list 120 for items listed for sale at the online marketplace. The self-learning system can analyze the user input 124, detect compatibility mismatches, and adjust the recommended compatibility list 120 accordingly (e.g., by adding or removing compatibilities between items and / or vehicle categories from the list). Due to the relatively large amount of user input 124 to the online marketplace (e.g., greater than a threshold, millions of different user inputs 124) and the complexity and diversity of the content of the user input 124, one or more users may be unable to manually process the user input 124.

[0076] With examples of environments in mind, consider now a discussion of some example details of techniques for detecting compatibility mismatches through generative artificial intelligence in accordance with one or more implementations.

[0077] Detecting compatibility mismatches through generative AI

[0078] Figure 2 Depicted is compatibility mismatch detection logic 200 in an example implementation for detecting compatibility mismatches through generative artificial intelligence.

[0079] The compatibility mismatch detection logic 200 can implement Figure 1 aspects of, or by Figure 1 For example, the compatibility mismatch detection logic 200 may be implemented by a compatibility detection system (e.g., referring to Figure 1 In one or more implementations, the compatibility mismatch detection logic 200 may include user input 124, one or more machine learning models 128, and a compatibility mismatch indication 140, which may be referenced. Figure 1 Examples of the corresponding features described.

[0080] In some examples, the compatibility mismatch detection logic 200 includes obtaining user input 124 and inputting the user input 124 into one or more machine learning models 128. The user input 124 may be filtered and / or otherwise pre-processed before being input into the machine learning model 128. The filtering and / or pre-processing of the data may include removing aspects of the data that are irrelevant or unrelated to the compatibility of an item with another item and / or the compatibility of an item with a device, which may be related to the compatibility of the item with another item and / or the compatibility of the item with a device. Figure 3In some cases, user input 124 may include one or more user-indicated or user-reported compatibilities between the item and another item and / or device (e.g., user-reported compatibility 126, as described in reference to FIG. Figure 1 For example, user input 124 may include an explicit indication that the item is compatible or incompatible with another item and / or device. In some other examples, user input 124 may include an implicit indication, such as an indication to return the item. If user input 124 includes an implicit indication, the compatibility of the item with the other item and / or device may be determined based on prior user purchases and / or details associated with the item and / or device, as well as other data.

[0081] In some examples, the machine learning model 128 may output a compatibility mismatch indication 140 indicating detected compatibility mismatch data 202. The detected compatibility mismatch data 202 may include data entries detected from the user input 124 that indicate an incompatibility but are on a recommended compatibility list (e.g., a reference to a compatible device). Figure 1 The compatibility mismatch indication 140 may be a combination of one or more items or devices listed as compatible in a recommended compatibility list 120 described above. For example, the detected compatibility mismatch data 202 may include one or more vehicle components that are incompatible with a corresponding vehicle type and / or vehicle class, wherein the vehicle component is listed as compatible with the corresponding vehicle type and / or vehicle class in the recommended compatibility list for the vehicle component. The compatibility information associated with the vehicle may be referred to as fitment data, wherein fitment refers to the compatibility of an automotive part (e.g., a vehicle component) with a specified vehicle class or vehicle type. In some examples, the compatibility mismatch indication 140 may trigger the sending of a message to a user account (e.g., a seller feedback file feed), the display of a message to a user account (e.g., via a GUI), and / or may trigger an update to the recommended compatibility list 120 (e.g., an update to a registry recording the compatibility mismatch).

[0082] In a variation, the system can perform machine learning model training 204. For example, the system can use a portion of the user input 124 as test data for training the machine learning model 128 and evaluating the performance of the machine learning model 128. The system can receive detected compatibility mismatch data 202 as output of the machine learning model 128 in response to the input test data, and can compare the output with a set of known compatibility mismatches from the test data. The system can update one or more parameters (e.g., weights) of the machine learning model 128 based on the evaluation, for example, based on an accuracy metric, a recall metric, and / or an F1 score that defines the performance of the machine learning model 128. The F1 score is the harmonic mean of the precision and recall of the machine learning model 128 and has a value in the range between 0 and 1.

[0083] The system can improve the machine learning model 128 by updating hyperparameters, fine-tuning the machine learning model 128, and / or changing the training data set until the performance criteria of the machine learning model 128 are met. The performance criteria may include meeting corresponding thresholds for a precision metric, a recall metric, and / or an F1 score. Additionally or alternatively, the performance criteria may include meeting a threshold return rate for one or more items listed for sale in the online marketplace. The hyperparameters of the machine learning model 128 include one or more settings of the machine learning model 128 defined for the machine learning model 128, such as a learning rate, a batch size, a number of hidden layers (e.g., for a neural network), one or more regularization parameters, a kernel size, and other parameters. Figure 4 The training of the machine learning model 128 is further described.

[0084] Figure 3 Depicted is compatibility mismatch detection logic 300 in an example implementation for detecting compatibility mismatches through generative artificial intelligence.

[0085] The compatibility mismatch detection logic 300 can implement Figure 1 aspects of, or by Figure 1 For example, the compatibility mismatch detection logic 300 can be implemented by a compatibility detection system (e.g., referring to Figure 1 In one or more implementations, the compatibility mismatch detection logic 300 may include user input 124, one or more machine learning models 128, and a compatibility mismatch indication 140, which may be referenced. Figure 1 and Figure 2 Examples of the corresponding features described.

[0086] In some examples, the compatibility detection system can collect data, including user input 124. For example, the compatibility mismatch detection logic 300 includes obtaining user input 124, such as from a database or other data storage device. The user input 124 may include one or more reviews, comments, and / or return reasons for items listed for sale on the online marketplace. A computing device (e.g., computing device 102, as shown in FIG. 1 ) may be used to determine the compatibility of an item. Figure 1 The compatibility detection system may receive a command from a user with a corresponding user account to initiate a purchase of an item, such as via a user interface displaying items for sale at an online marketplace. The compatibility detection system may prompt the user to provide user input 124. For example, the compatibility detection system may instruct the computing device to display a feedback request for the user account that initiated the purchase of the item. The computing device may receive user input 124 in response to the feedback request.

[0087] In some cases, user input 124 may include text data, audio data, and / or video data. User input 124 may include an identifier of the corresponding user providing user input 124, such as a username. The compatibility detection system may infer which item the user input 124 corresponds to from the identifier and / or other application data provided from the computing device. For example, user input 124 may include an identifier "User A" and corresponding text data "It does not fit the headlight housing." The compatibility detection system may determine that user input 124 corresponds to a headlight item (e.g., based on purchase history transaction data of the corresponding user account and / or based on information provided by the computing device). Additionally or alternatively, user input 124 may include an identifier "User B" and corresponding text data "This does not fit my YYYY, make, model vehicle," an identifier "User C" and corresponding text data "This item does not fit my car," and an identifier "User D" and corresponding text data "Great! 5 / 5." Although four user identifier and text data examples are provided, user input 124 may include any number of data entries. For example, user input 124 may include a relatively large number of data entries (eg, greater than a threshold, millions of data entries) that a user may not be able to manually process.

[0088] In some examples, the compatibility mismatch detection logic 300 includes inputting user input 124 into one or more machine learning models 128, wherein the one or more machine learning models 128 are trained to detect compatibility mismatches between an item listed for sale on an online marketplace and another item. At 302, the compatibility detection system can extract items for detecting compatibility mismatches from the user input 124. If the user input 124 is textual data, the compatibility detection system can parse the user input 124 to determine one or more items that are compatible with the item associated with the user input 124 (e.g., the item being reviewed). If the user input is audio data, the compatibility detection system can convert the audio data into textual data for parsing. If the user input is visual data (e.g., an image and / or video), the compatibility detection system can use image processing techniques (including feature and object detection techniques) to determine one or more items from the visual data. In some cases, for example, if the user input 124 is for one or more vehicle parts, extracting items for detecting compatibility mismatches may include extracting one or more vehicles mentioned in the user input 124 and / or indicated by the user providing the user input 124. The machine learning model 128 may process the user input 124 to detect one or more compatibility mismatches between the extracted items and items corresponding to the user input 124 (e.g., items purchased by the user associated with the user account, items reviewed, items linked to reviews and / or returned, etc.).

[0089] If one or more compatibility mismatches are detected, the compatibility detection system may generate a compatibility mismatch indication 140. For example, the compatibility detection system may receive one or more indications of compatibility mismatches, including data indicating that an item and one or more corresponding items are incompatible. Additionally or alternatively, the compatibility mismatch indication 140 may include text and / or audio output generated by one or more large language models. In variations, for example, if a user maintains a recommended compatibility list, the compatibility mismatch indication 140 may include text indicating one or more suggestions or recommended updates to the recommended compatibility list for one or more items listed for sale at the online marketplace. In some other variations, for example, if the compatibility detection system maintains a recommended compatibility list, the compatibility detection system may use data output from a machine learning model to update the recommended compatibility list. The compatibility mismatch indication 140 may include text indicating that one or more compatibilities on the recommended compatibility list for items listed for sale at the online marketplace are updated. The compatibility detection system may send or otherwise trigger output of a compatibility mismatch indication 140 for display at one or more computing devices of a user (e.g., a seller of one or more items with recommended updates or automatic updates to a recommended compatibility list).

[0090] Figure 4 Depicted is compatibility mismatch detection logic 400 in an example implementation for detecting compatibility mismatches through generative artificial intelligence.

[0091] The compatibility mismatch detection logic 400 can implement Figure 1 aspects of, or by Figure 1 For example, the compatibility mismatch detection logic 400 can be implemented by a compatibility detection system (e.g., referring to Figure 1 In one or more implementations, the compatibility mismatch detection logic 400 may include a recommended compatibility list 120, user input 124, one or more machine learning models 128, and a compatibility mismatch indication 140, which may be referenced. Figures 1 to 3 Examples of the corresponding features described.

[0092] The compatibility mismatch detection logic 400 may include the compatibility detection system obtaining a training data set that includes the recommended compatibility list 120 and user input 124. The compatibility detection system may obtain the training data set from a database or other data storage device of the online marketplace. The training data set may include existing data that may be provided by a user (e.g., the recommended compatibility list 120) and the user input 124, which may include feedback, comments, and / or other reviews related to one or more items for sale at the online marketplace, such as reference. Figure 3 The compatibility detection system may clean and pre-process the collected or obtained data. For example, the compatibility detection system may remove one or more characters and / or string values ​​from the user input 124 that are not relevant to the compatibility detection. Example characters and / or string values ​​that may be removed include, but are not limited to, characters and / or string values ​​that do not include information describing item compatibility of one or more items (e.g., the purchased item is not compatible with another item already owned by the user). Additionally or alternatively, cleaning and pre-processing the data may include formatting the data (e.g., into separate data entries) for input and / or training of the machine learning model 128.

[0093] At 402, the compatibility mismatch detection logic 400 may include training the machine learning model 128 using one or more training data sets. In some examples, the machine learning model 128 is pre-trained on a relatively large amount of data (e.g., greater than a threshold) to learn patterns in the data. The pre-training phase may include unsupervised training techniques in which the data is not labeled. The compatibility detection system may use supervised learning techniques to further train the pre-trained generative artificial intelligence model (e.g., including one or more machine learning models 128). The machine learning model 128 is trained on the training data set at a lower learning rate than during the pre-training phase (e.g., by adjusting the parameters of the model with a step size less than a threshold). The supervised learning technique may include inputting pre-processed data as a training data set to the machine learning model 128 and receiving an output (e.g., a compatibility mismatch indication 140) indicating one or more compatibility mismatches detected from the pre-processed data. The machine learning model 128 may be an example of a distributed neural network including multiple machine learning models 128, and / or may include a single machine learning model 128.

[0094] Once the machine learning model 128 is trained on the initial training data set, the compatibility detection system can evaluate the machine learning model 128 to determine whether the machine learning model 128 meets performance criteria. In some cases, the compatibility detection system can evaluate the performance of the machine learning model 128 by inputting additional test data and comparing the output of the machine learning model 128 to known results. For example, at 404, the compatibility detection system can receive compatibility mismatch feedback and use the compatibility mismatch feedback as additional test data. The compatibility detection system can determine one or more metrics based on the evaluation, including precision, recall, and F1 score of the machine learning model 128. Additionally or alternatively, the compatibility detection system can monitor the return rate of items purchased from the online marketplace and can continue to train, fine-tune, and / or modify the parameters of the machine learning model 128 until the return rate of the items meets a threshold. In some examples, the compatibility detection system can monitor the return rate of items returned due to compatibility mismatches (e.g., based on an indication from a user that the return was due to a compatibility mismatch and / or based on evaluating user input, including reviews, comments, purchase history transaction data, or other input).

[0095] If the metric and / or the bounce rate satisfies one or more corresponding thresholds (e.g., performance criteria), the compatibility detection system can implement the machine learning model 128 to generate a compatibility mismatch indication 140 using the user input 124 collected from the online marketplace. If the metric fails to satisfy one or more corresponding thresholds, the compatibility detection system can continue to fine-tune and / or train the machine learning model 128 using additional datasets. For example, the compatibility detection system can improve the machine learning model 128 by modifying (e.g., updating) one or more hyperparameters of the corresponding machine learning model 128, fine-tuning the machine learning model 128, and / or changing the training dataset. Once the machine learning model 128 is trained and improved, the compatibility detection system can use the machine learning model 128 to detect compatibility mismatches for items listed for sale at the online marketplace.

[0096] Figure 5 An example 500 of a user interface for detecting compatibility mismatches through generative artificial intelligence is depicted.

[0097] The illustrated example 500 includes a display via a user interface of a computing device that is referenced Figure 1 5. At a first time, a computing device generates a display 502. Display 502 may include an instance of an online marketplace application 504. The instance of online marketplace application 504 may be associated with a user account (e.g., a buyer user account) having a corresponding identifier 506. Identifier 506 may be a username of the user account (e.g., "User A").

[0098] In some examples, display 502 may include one or more features with which a user can interact, referred to as interactive features of display 502 and / or user interface. For example, display 502 may include a search feature 508 for a user to input data. Search feature 508 may prompt the user to enter a search term or search query, such as a string and / or character value. The user may enter the search term and / or search query as text input via a component coupled to display 502 (e.g., a keyboard). Online marketplace application 504 may use the search query to disseminate one or more results. For example, online marketplace application 504 may access a database storing detail pages of items for sale and may compare the search query with information stored for corresponding items for sale. Online marketplace application 504 may select one or more corresponding items based on whether the similarity between the search query and the information meets a threshold (e.g., having a threshold number of matching strings or character values).

[0099] Display 502 may include a control panel 510 having one or more input features 512 for collecting information about a user's items and / or devices (e.g., owned by or otherwise in the possession of the user). Input features 512 may include one or more drop-down menus for collecting one or more features of textual, audio, and / or visual input, or any other options selectable by the user. For example, a user may manually enter information about an item (e.g., a vehicle). In some other examples, a user may upload an image of an item, and online marketplace application 504 may use image processing techniques (e.g., feature detection, object detection, etc.) to identify one or more items and / or devices from the image. In a variation, the item and / or device is a vehicle, and input features 512 include year, make, model, trim, and / or engine, as well as other vehicle identifiers. Control panel 510 may include prompts such as "Find a suitable part" and "We need more information about your vehicle to confirm fit" to instruct the user to provide vehicle information. Input features 512 may include a drop-down menu listing options for different vehicle features. For example, a drop-down menu for specifying a vehicle trim may include options A through Z. A user may use navigation tools 514 (e.g., a mouse, keyboard, etc.) to navigate control panel 510. Although a vehicle is shown in example 500, display 502 may include control panel 510 for collecting information about any item that has compatibility with another item (e.g., an appliance, an electronic device, an electromechanical device, etc.).

[0100] Online marketplace application 504 can filter search results for items and / or devices compatible with information provided by a user (e.g., via control panel 510). For example, an item listed for sale at online marketplace application 504 can be compatible with one or more other items (e.g., vehicles). Online marketplace application 504 can propagate search results for display 502, corresponding to items compatible with the item indicated by the information. For example, online marketplace application 504 can propagate search results to display 502 items (e.g., parts and / or accessories) that are compatible with a specified vehicle. In some examples, a user can provide additional input to instruct online marketplace application 504 to filter one or more search results, including brand selection, brand type selection, manufacturer warranty, and other item characteristics that can be used to filter the search results.

[0101] In some examples, at a second time, the computing device may generate a display 516 to confirm the item indicated by the information from the control panel 510. The computing device may receive user input that triggers the display 516. The user input may indicate that the user has provided sufficient information for the item and / or device. The user input may include activating a "Done" button via the navigation tool 514 (e.g., by clicking on the navigation tool 514). The display 516 may include one or more buttons 518 that can be selected by the user to display search results filtered based on information (e.g., year, make, model, trim, etc.), and / or to change or otherwise update the information. Once the computing device receives user input instructing the computing device to display search results, the computing device may generate a user interface that displays the search results, which will be related to Figure 6 Described in further detail.

[0102] Figure 6 An example 600 of a user interface for detecting compatibility mismatches through generative artificial intelligence is depicted.

[0103] The illustrated example 600 includes a display via a user interface of a computing device that is referenced Figure 1 At a first time, a computing device generates a display 602. The display 602 may include an instance of an online marketplace application 604.

[0104] In some examples, display 602 may include one or more features with which a user can interact. For example, display 602 may include one or more posted detail pages 606 of items for sale at an online marketplace application 604. Posted detail pages 606 may include items that meet compatibility with another item 608 and meet a threshold similarity value when compared to a search query. Figure 5 As described, a user can specify information related to another item 608. In some examples, the other item 608 can be a vehicle having a year, make, model, and trim. In some other examples, the other item 608 can be any item that is compatible with the other item. The posted detail page 606 can include an image of the item for sale at the online marketplace application 604 and a description of the item. The description can include information such as the title of the item (e.g., "282mm front and rear drilled rotor ceramic brake pads"), the price of the item (e.g., $XYZ.AB), and additional information related to the item and / or the delivery of the item.

[0105] In some examples, the computing device outputting display 602 may receive user input indicating a selection of a posted detail page 606. In response to the user input, the computing device may output display 610. Display 610 may include additional details describing the selected posted detail page 606. For example, display 610 may include a compatibility message 612 indicating whether the item in posted detail page 606 is compatible with the item indicated by the user (e.g., a vehicle). In some examples, display 610 may include a compatibility display 614 including a recommended compatibility list 120. In some examples, if the item in posted detail page 606 is a part, accessory, or otherwise vehicle-related, the recommended compatibility list 120 may include an indication of compatible vehicle types. For example, the recommended compatibility list 120 may include the year, make, model, trim, and engine type of one or more vehicles that are compatible with the item.

[0106] although Figure 5 and Figure 6 Vehicle parts are shown as an example of an item listed for sale via the online marketplace, but the item may be any item (eg, including parts, accessories, etc.) that has compatibility with another item.

[0107] Having discussed exemplary details of detecting compatibility mismatches through generative artificial intelligence, we now consider some examples of the process to illustrate additional aspects of the technology.

[0108] Example Process

[0109] This section describes examples of processes for detecting compatibility mismatches using generative artificial intelligence. Various aspects of these processes can be implemented in hardware, firmware, or software, or a combination thereof. These processes are shown as a set of blocks that specify operations performed by one or more devices and are not necessarily limited to the order shown for performing the operations by the respective blocks.

[0110] Figure 7 Depicted is a procedure 700 in an example implementation for detecting compatibility mismatches through generative artificial intelligence.

[0111] At 702, compatibility data associated with compatibility between a set of items and a set of vehicle categories is obtained (e.g., by accessing a database). By way of example, the compatibility detection system accesses a data storage device to retrieve the compatibility data (e.g., the compatibility detection system 104, the storage device 114, and the compatibility data 134, as described with reference to FIG. Figure 1The compatibility data may include a recommended compatibility list between the set of items and the set of vehicle categories, and a user-reported compatibility between at least one item in the set of items and at least one vehicle category in the set of vehicle categories.

[0112] At 704, a machine learning model for detecting compatibility mismatches between vehicle classes and items is generated based on providing at least a portion of the compatibility data as input to the generative artificial intelligence. By way of example, the generative artificial intelligence manager may generate the machine learning model using model training logic (e.g., the generative artificial intelligence manager 106 and the model training logic 130, as described with reference to FIG. Figure 1 In some embodiments, the machine learning model comprises a plurality of machine learning models (e.g., in a distributed network of machine learning models).

[0113] In some cases, a request for compatibility data is sent to a device for display to a user (e.g., via a user interface of the device). In response to the request, user input corresponding to one or more of a recommended compatibility list or user-reported compatibilities is received. Based on processing the user input to determine the user-reported compatibilities, one or more of the recommended compatibility list or user-reported compatibilities is stored at a database. In variations, for example, for text data and / or audio data converted into text data, the user input is parsed to determine individual string values ​​and individual character values ​​corresponding to one or more of at least one item or at least one vehicle category. In some other variations, for example, for visual data (e.g., image and / or video data), the user input is processed using image processing techniques to determine at least one item or at least one vehicle category. The image processing techniques may include feature recognition and / or object recognition to detect one or more features of an item, and / or detect one or more features of a vehicle to classify the vehicle.

[0114] In some examples, one or more parameters (e.g., hyperparameters) of a machine learning model are modified to generate the machine learning model. By way of example, one or more hyperparameters of the machine learning model are modified until a performance criterion is met. The performance criterion may include one or more of a threshold accuracy metric for the machine learning model, a threshold recall metric for the machine learning model, a threshold F1 score for the machine learning model, or a threshold return rate for the set of items (e.g., after implementing the machine learning model at an online marketplace). Modifying the parameters may include providing additional training data sets to the machine learning model to further train and / or fine-tune the machine learning model.

[0115] In some cases, based on the performance criteria being met, additional compatibility data is provided as input to a machine learning model to detect a compatibility mismatch. The additional compatibility data includes at least one user-reported compatibility indicating a vehicle class and an item having a compatibility mismatch. In some examples, an indication of a compatibility mismatch is received as an output from the machine learning model based on providing at least one user-reported compatibility corresponding to the vehicle class and the item as input to the machine learning model.

[0116] At 706, an update to the recommended compatibility list is determined based on detecting the compatibility mismatch using the machine learning model. In some examples, an indication of the update to the recommended compatibility list is sent to the device for display to a user (e.g., via a user interface of the device). In response to the indication, a request to perform an update to the recommended compatibility list can be received.

[0117] In some examples, an update to the recommended compatibility list is performed. The update to the recommended compatibility list may include one or more of: removing a vehicle category from the recommended compatibility list for the item, or adding a vehicle category to the recommended compatibility list for the item.

[0118] Figure 8 Depicted is a procedure 800 in an example implementation for detecting compatibility mismatches through generative artificial intelligence.

[0119] At 802, compatibility data between a set of items is obtained (e.g., by accessing a database). By way of example, the compatibility detection system accesses a data storage device to retrieve the compatibility data (e.g., the compatibility detection system 104, the storage device 114, and the compatibility data 134, as described with reference to FIG. Figure 1 The compatibility data may include a recommended compatibility list between the set of items and a user-reported compatibility associated with at least one item in the set of items.

[0120] At 804, a machine learning model for detecting a compatibility mismatch between a first item in the group of items and a second item in the group of items is generated based on providing at least a portion of the compatibility data as an input to the generative artificial intelligence. By way of example, the generative artificial intelligence manager may generate the machine learning model using model training logic (e.g., generative artificial intelligence manager 106 and model training logic 130, as described with reference to FIG. Figure 1 In some embodiments, the machine learning model comprises a plurality of machine learning models (e.g., in a distributed network of machine learning models).

[0121] In some cases, a request for compatibility data is sent to a device for display to a user (e.g., via a user interface of the device). In response to the request, user input corresponding to one or more of a recommended compatibility list or user-reported compatibilities is received. Based on processing the user input to determine the user-reported compatibilities, one or more of the recommended compatibility list or user-reported compatibilities is stored at a database. In variations, for example, for text data and / or audio data converted into text data, the user input is parsed to determine individual string values ​​and individual character values ​​corresponding to one or more of at least one item or at least one vehicle category. In some other variations, for example, for visual data (e.g., image and / or video data), the user input is processed using image processing techniques to determine at least one item or at least one vehicle category. The image processing techniques may include feature recognition and / or object recognition to detect one or more features of a first item and / or detect one or more features of a second item.

[0122] In some examples, one or more parameters (e.g., hyperparameters) of a machine learning model are modified to generate the machine learning model. By way of example, one or more hyperparameters of the machine learning model are modified until a performance criterion is met. The performance criterion may include one or more of a threshold accuracy metric for the machine learning model, a threshold recall metric for the machine learning model, a threshold F1 score for the machine learning model, or a threshold return rate for the set of items (e.g., after implementing the machine learning model at an online marketplace). Modifying the parameters may include providing additional training data sets to the machine learning model to further train and / or fine-tune the machine learning model.

[0123] In some cases, additional compatibility data is provided as input to a machine learning model to detect a compatibility mismatch based on the performance criteria being met. The additional compatibility data includes at least one user-reported compatibility indicating an item having a compatibility mismatch. In some examples, an indication of a compatibility mismatch is received as an output from the machine learning model based on providing at least one user-reported compatibility corresponding to a first item and a second item as input to the machine learning model.

[0124] At 806, an update to the recommended compatibility list is determined based on detecting the compatibility mismatch using the machine learning model. In some examples, an indication of the update to the recommended compatibility list is sent to the device for display to a user (e.g., via a user interface of the device). In response to the indication, a request to perform an update to the recommended compatibility list can be received.

[0125] In some examples, an update to the recommended compatibility list is performed. The update to the recommended compatibility list can include one or more of: removing the first item from the recommended compatibility list of the second item, removing the second item from the recommended compatibility list of the first item, adding the first item to the recommended compatibility list of the second item, or adding the second item to the recommended compatibility list of the first item.

[0126] Having described examples of processes according to one or more implementations, consider now examples of systems and devices that can be used to implement the various techniques described herein.

[0127] Example systems and devices

[0128] Figure 9 An example of a system is shown generally at 900, including an example of a computing device 902, which represents one or more computing systems and / or devices that may implement the various techniques described herein. This is illustrated by the inclusion of application 110 and compatibility detection system 104. Computing device 902 may be, for example, a server of a service provider, a device associated with a client (e.g., a client device), a system on a chip, and / or any other suitable computing device or computing system.

[0129] As shown, the example computing device 902 includes a processing system 904, one or more computer-readable media 906, and one or more I / O interfaces 908 that are communicatively coupled to each other. Although not shown, the computing device 902 may also include a system bus or other data and command transmission system that couples the various components to each other. The system bus may include any one or a combination of different bus structures, such as a memory bus or memory controller, a peripheral bus, a universal serial bus, and / or a processor or local bus utilizing any one of a variety of bus architectures. Various other examples are also contemplated, such as control lines and data lines.

[0130] Processing system 904 represents a function for performing one or more operations using hardware. Therefore, processing system 904 is shown as including hardware elements 910 that can be configured as processors, functional blocks, etc. This can include implementation in the form of hardware, as a dedicated integrated circuit or other logic device formed using one or more semiconductors. Hardware elements 910 are not limited by the materials from which they are formed or the processing mechanisms employed therein. For example, a processor can be composed of semiconductors and / or transistors (e.g., electronic integrated circuits (ICs)). In such a case, the processor executable instructions can be electronic executable instructions.

[0131] Computer-readable media 906 is shown as including memory / storage 912. Memory / storage 912 represents memory / storage capacity associated with one or more computer-readable media. Memory / storage 912 may include volatile media (e.g., random access memory (RAM)) and / or non-volatile media (e.g., read-only memory (ROM), flash memory, optical disks, magnetic disks, etc.). Memory / storage 912 may include fixed media (e.g., RAM, ROM, fixed hard drives, etc.) and removable media (e.g., flash memory, removable hard drives, optical disks, etc.). As further described below, computer-readable media 906 may be configured in various other ways.

[0132] The input / output interface 908 represents functionality that enables a user to input commands and information to the computing device 902, and also enables information to be presented to the user and / or other components or devices using various input / output devices. Examples of input devices include a keyboard, a cursor control device (e.g., a mouse), a microphone, a scanner, touch functionality (e.g., a capacitive sensor or other sensor configured to detect physical touch), a camera (e.g., which may employ visible or invisible wavelengths such as infrared frequencies to identify movement as gestures that do not involve touch), and the like. Examples of output devices include a display device (e.g., a monitor or projector), speakers, a printer, a network card, a tactile response device, and the like. Thus, as further described below, the computing device 902 may be configured in various ways to support user interaction.

[0133] Various techniques may be described herein in the general context of software, hardware elements, or program modules. Typically, such modules include routines, programs, objects, elements, components, data structures, etc. that perform specific tasks or implement specific abstract data types. As used herein, the terms "module," "function," and "component" generally refer to software, firmware, hardware, or a combination thereof. A feature of the techniques described herein is that they are platform-independent, meaning that they can be implemented on a variety of commercial computing platforms with a variety of processors.

[0134] An implementation of the described modules and techniques may be stored on or transmitted across some form of computer-readable media. Computer-readable media may include various media that can be accessed by computing device 902. By way of example, and not limitation, computer-readable media may include "computer-readable storage media" and "computer-readable signal media."

[0135] "Computer-readable storage media" may refer to media and / or devices that enable persistent and / or non-transitory storage of information, as compared to mere signal transmission, carrier waves, or the signals themselves. Thus, computer-readable storage media refers to non-signal-bearing media. Computer-readable storage media include hardware such as volatile and non-volatile, removable and non-removable media, and / or storage devices implemented in a method or technology suitable for storing information such as computer-readable instructions, data structures, program modules, logic elements / circuits, or other data. Examples of computer-readable storage media may include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVDs) or other optical storage devices, hard disks, magnetic cassettes, magnetic tape, magnetic disk storage devices or other magnetic storage devices, or other storage devices, tangible media, or articles of manufacture suitable for storing desired information and that can be accessed by a computer.

[0136] "Computer-readable signal media" may refer to signal-bearing media that is configured to send instructions to the hardware of the computing device 902, for example, via a network. Signal media may typically embody computer-readable instructions, data structures, program modules, or other data in a modulated data signal (such as a carrier wave, a data signal, or other transport mechanism). Signal media also includes any information delivery media. The term "modulated data signal" refers to a signal that has one or more characteristics set or changed in such a manner as to encode information in the signal. By way of example, and not limitation, communication media include wired media such as a wired network or a direct wired connection, and wireless media such as acoustic, RF, infrared, and other wireless media.

[0137] As previously described, hardware elements 910 and computer-readable medium 906 represent modules, programmable device logic, and / or fixed device logic that can be employed in some embodiments to implement at least some aspects of the technology described herein (e.g., executing one or more instructions) implemented in hardware. The hardware can include components of an integrated circuit or components of a system on a chip, an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a complex programmable logic device (CPLD), and other implementations in silicon or other hardware. In this case, the hardware can operate as follows: a processing device that performs program tasks defined by the instructions and / or logic embodied by the hardware, and hardware for storing execution instructions, such as the computer-readable storage medium described previously.

[0138] The aforementioned combination can also be adopted to implement the various technologies described herein. Therefore, software, hardware or executable modules can be implemented as one or more instructions and / or logic embodied in some form of computer-readable storage medium, and / or implemented by one or more hardware elements 910. Computing device 902 can be configured to implement specific instructions and / or functions corresponding to software modules and / or hardware modules. Therefore, the implementation of the module that can be executed as software by computing device 902 can be implemented at least in part with hardware, for example, by using computer-readable storage medium and / or the hardware elements 910 of processing system 904. Instructions and / or functions can be performed / operated by one or more products (e.g., one or more computing devices 902 and / or processing system 904) to implement the technology, modules and examples described herein.

[0139] The technology described herein can be supported by various configurations of computing device 902 and is not limited to the specific examples of the technology described herein. The functionality can also be implemented in whole or in part using a distributed system, such as on a "cloud" 914 via a platform 916, as described below.

[0140] Cloud 914 includes and / or represents a platform 916 for resources 918. Platform 916 abstracts the underlying functionality of the hardware (e.g., servers) and software resources of cloud 914. Resources 918 may include applications and / or data that can be utilized when performing computer processing on a server remote from computing device 902. Resources 918 may also include services provided over the Internet and / or over a subscriber network (e.g., a cellular or Wi-Fi network).

[0141] The platform 916 can abstract resources and functionality to connect the computing device 902 with other computing devices. The platform 916 can also be used to abstract the scale of resources to provide a corresponding level of scale to the demand encountered for resources 918 (which is implemented via the platform 916). Therefore, in an interconnected device embodiment, the implementation of the functionality described herein can be distributed throughout the system 900. For example, the functionality can be implemented partially on the computing device 902 and can be implemented via the platform 916 that abstracts the functionality of the cloud 914.

[0142] in conclusion

[0143] Although the systems and techniques have been described in language specific to structural features and / or methodological acts, it should be understood that the systems and techniques defined in the appended claims are not necessarily limited to the specific features or acts described. Rather, these specific features and acts are disclosed as example forms of implementing the claimed subject matter.

Claims

1. A computer-implemented method comprising: obtaining compatibility data between a plurality of items and a plurality of vehicle classes, the compatibility data comprising a recommended compatibility list between the plurality of items and the plurality of vehicle classes and a user-reported compatibility between at least one of the plurality of items and at least one of the plurality of vehicle classes; generating a machine learning model for detecting compatibility mismatches between vehicle classes and items based on providing at least a portion of the compatibility data as input to a generative artificial intelligence; as well as An update to the recommended compatibility list is determined based on detecting the compatibility mismatch using the machine learning model.

2. The computer-implemented method of claim 1 , wherein: Generating the machine learning model also includes: modifying one or more parameters of the machine learning model until a performance criterion associated with the machine learning model is met.

3. The computer-implemented method of claim 2, wherein: Modifying one or more parameters of the machine learning model is based on providing additional compatibility data as input to the generative artificial intelligence.

4. The computer-implemented method of claim 2, wherein: Detecting the compatibility mismatch also includes providing additional compatibility data as input to the machine learning model based on performance criteria associated with the machine learning model being met, wherein the additional compatibility data includes at least one user-reported compatibility corresponding to the vehicle category and the item.

5. The computer-implemented method of claim 2, wherein: The performance criteria include one or more of a threshold accuracy metric associated with the machine learning model, a threshold recall metric associated with the machine learning model, a threshold F1 score associated with the machine learning model, or a threshold return rate associated with the plurality of items.

6. The computer-implemented method of claim 1 , wherein: Detecting the compatibility mismatch also includes receiving an indication of the compatibility mismatch as an output from the machine learning model based on providing at least one user-reported compatibility corresponding to the vehicle category and the item as input to the machine learning model.

7. The computer-implemented method of claim 1 , further comprising: An indication of an update to the recommended compatibility list is sent to a device for display to a user.

8. The computer-implemented method of claim 7, further comprising: In response to the indication, a request is received to perform an update to the recommended compatibility list.

9. The computer-implemented method of claim 1 , further comprising: An update to the recommended compatibility list is performed, wherein the update to the recommended compatibility list comprises one or more of: removing the vehicle category from the recommended compatibility list for the item, or adding the vehicle category to the recommended compatibility list for the item.

10. The computer-implemented method of claim 1 , further comprising: sending a request for the compatibility data to a device for display to a user; receiving user input corresponding to one or more of the recommended compatibility list or the user-reported compatibilities in response to the request; as well as Based on processing the user input to determine the user-reported compatibilities, the one or more of the recommended compatibility list or the user-reported compatibilities are stored in a database.

11. The computer-implemented method of claim 10, wherein: The processing also includes parsing the user input to determine respective string values ​​and respective character values ​​corresponding to one or more of the at least one item or the at least one vehicle category.

12. A system comprising: one or more processors; as well as A computer-readable storage medium storing instructions executable by the one or more processors to perform operations comprising: obtaining compatibility data between a plurality of items and a plurality of vehicle classes, the compatibility data comprising a recommended compatibility list between the plurality of items and the plurality of vehicle classes and a user-reported compatibility between at least one of the plurality of items and at least one of the plurality of vehicle classes; generating a machine learning model for detecting compatibility mismatches between vehicle classes and items based on providing at least a portion of the compatibility data as input to a generative artificial intelligence; and An update to the recommended compatibility list is determined based on detecting the compatibility mismatch using the machine learning model.

13. A computer-implemented method comprising: obtaining compatibility data between a plurality of items, the compatibility data comprising a recommended compatibility list between the plurality of items and a user-reported compatibility associated with at least one item in the plurality of items; generating a machine learning model for detecting a compatibility mismatch between a first item in the plurality of items and a second item in the plurality of items based on providing at least a portion of the compatibility data as input to a generative artificial intelligence; as well as An update to the recommended compatibility list is determined based on detecting the compatibility mismatch using the machine learning model.

14. The computer-implemented method of claim 13, wherein: Generating the machine learning model also includes: modifying one or more parameters of the machine learning model until a performance criterion associated with the machine learning model is met.

15. The computer-implemented method of claim 14, wherein: Modifying one or more parameters of the machine learning model is based on providing additional compatibility data as input to the generative artificial intelligence.

16. The computer-implemented method of claim 14, wherein: Detecting the compatibility mismatch also includes providing additional compatibility data as input to the machine learning model based on performance criteria associated with the machine learning model being met, wherein the additional compatibility data includes at least one user-reported compatibility corresponding to the first item and the second item.

17. The computer-implemented method of claim 13, wherein: Detecting the compatibility mismatch also includes receiving an indication of the compatibility mismatch as an output from the machine learning model based on providing at least one user-reported compatibility corresponding to the first item and the second item as input to the machine learning model.

18. The computer-implemented method of claim 13, further comprising: An indication of an update to the recommended compatibility list is sent to a device for display to a user.

19. The computer-implemented method of claim 13, further comprising: Performing an update to the recommended compatibility list, wherein the updating to the recommended compatibility list comprises one or more of: removing the first item from the recommended compatibility list of the second item, removing the second item from the recommended compatibility list of the first item, adding the first item to the recommended compatibility list of the second item, or adding the second item to the recommended compatibility list of the first item.

20. The computer-implemented method of claim 13, further comprising: sending a request for the compatibility data to a device for display to a user; receiving user input corresponding to one or more of the recommended compatibility list or the user-reported compatibilities in response to the request; as well as Based on processing the user input to determine the user-reported compatibilities, the one or more of the recommended compatibility list or the user-reported compatibilities are stored in a database.

Images