Patent Application
20230351359
Transaction tax engines are used to calculate taxes applicable to transactions for goods and services at locations around the world. To determine tax for a product at a transaction device, the transaction device sends a request for a transaction tax calculation over a computer network such as the Internet to a transaction tax engine provisioned at a cloud server by a cloud service provider. The transaction tax engine returns a tax calculation response to the transaction device. While such an approach offers the convenience of point-of-sale tax calculations, some drawbacks exist. For example, it may require large amounts of time and money for a company to manually map its products to a tax category. With tax rates and rules in various jurisdictions being complicated and ever-changing, it can be challenging to efficiently and accurately map a product to the appropriate tax category. Additionally, inaccurately mapping a product to a tax category could result in an error in the tax calculation. This could lead to a loss of profit for the company if the customer is charged too low of a tax rate at the time of the transaction, as the company may be responsible for paying the outstanding tax amount and may be assessed a fine by the jurisdiction if the error is discovered. Further, when the error in tax calculation results in overcharging a customer, the company may face legal liability.
To address the above issues, a computer system for mapping products to taxability categories is provided. The computer system has server-side and client-side aspects. According to one aspect, the computer system includes one or more processors configured to execute, in a run-time inference phase, an artificial intelligence model, a taxability category mapping engine, and a taxability category driver record association engine. The artificial intelligence model is configured to receive as run-time input product text including a product name and product description associated with a product catalog, and output a run-time output including a predicted tax category for a product associated with the product catalog. The taxability category mapping engine is configured to link a taxability driver to the product. The taxability category driver record association engine is configured to create a taxability category mapping drivers record associating the taxability driver that is linked to the product to a taxability category. The predicted tax category output from the artificial intelligence model and the taxability category mapping drivers record are stored in a product taxability record. Other aspects of the system are described below.
This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter. Furthermore, the claimed subject matter is not limited to implementations that solve any or all disadvantages noted in any part of this disclosure.
As schematically illustrated in
As described in detail below with reference to
The computing device 10 further includes a taxability category driver record association engine 28. Once the taxability driver 26 is linked to the product, the taxability category driver record association engine 28 associates the taxability driver with a tax category and creates a taxability category mapping drivers record 30 for the product that is included in a product taxability record 32. The product taxability record 32 further includes a taxability category mapping record 34 which is associated with the predicted tax category 36 for the product output by the tax category AI model 16 and the taxability category mapping drivers record 30. The taxability categories 36 are predefined groups that help determine the selection of one or more tax rule based on the product and the geographic region in which the transaction occurs. Examples of the taxability category mapping drivers record 30 and the taxability category mapping record 34 are shown below with reference to
In some implementations, taxability driver records, taxability category mapping driver records, taxability category mapping records, and taxability category records may be uploaded by the client. These records provide information regarding existing taxability drivers and existing taxability mapping, thereby allowing the system to filter out products that are already mapped to a taxability category, filter out products that don't have a corresponding taxability driver, and determine a range of taxability categories most likely to be mapped to the client's products.
The product taxability record 32 is output to an administrative interface 38 and stored as product data 40. The client may utilize the client administrative device 20 to access the administrative interface 38 of the computing device 10 and upload client configuration settings 42 to the computing device 10. The client configuration settings 42 may include, for example, a client-specified subset of geographic regions 44 in which transactions take place or that govern the transactions due to the location at which they are legally deemed to take place (in the case of an online transaction, for example), which have been selected from among a global list of geographic regions stored at the computing device 10. The client configuration settings 42 may further include information on tax exemptions in rule and rate data that apply to transactions.
The product data 40, including the product taxability record 32, and the client configuration settings 42 are received by a transaction tax engine 46 included in the computing device 10. Transaction tax engine configuration settings 48, including the taxability driver 26 and predicted tax category 36, are extracted from the product taxability record 32 and the client configuration settings 42. The transaction tax engine 46 is in communication with an associated global tax rules database 50 included in the computing device 10. The global tax rules database 50 includes tax rate and rule data 52, such tax rates and rules to be applied to products in different tax categories and multiple geographic regions for a global list of geographic regions and a global list of tax categories, for example. The tax rates and rules can be indexed by geographic region, tax category, and other related indexes.
The transaction tax engine 46 is configured to receive a tax calculation request 54 from a transaction device 56 processing a transaction for the product. The transaction device 56 may be implemented as a customer computing device, a point-of-sale terminal, or another computing device suitable for executing a transaction application 58, such as a shopping cart application for an e-commerce site, a point-of-sale terminal application, or other transaction software, for example. Upon receiving the tax calculation request 54, the transaction tax engine 46 is configured to process the tax calculation request 54 according to the transaction tax engine configuration settings 48 and the tax rate and rule data 52, and transmit a tax calculation response 60 to the transaction device 56, such that the correct tax amount is included in the price of the product and the transaction for the product may be completed. This is accomplished using the driver architecture discussed above, as the transaction tax engine 46 filters the tax rate and rule data 52 according to the predicted tax category 36 associated to the taxability driver 26 that is linked to the product and included in the transaction tax engine configuration settings 48.
The global tax rules database 50 and transaction tax engine 46 require significant memory requirements due to their large size, require deployment on servers that are scalable to address spikes in demand, and require continuous updating as the tax rules of the various geographic regions around the world change. For these reasons, the computing device 10 is often centrally hosted in a data center of a cloud computing service that offers its platform as a service. However, in some implementations, the computing device 10 may utilize containerization, deployment, and management of a compact and client-localized version of the transaction tax engine 46 at the client administration device 20. It will be appreciated that the computing device 10, the client administrative device 20, and the transaction device 56 are configured to communicate over a network, with client administrative device 20 and the transaction device 56 behind a client firewall to control traffic between the internet and the client administrative device 20 and the transaction device 56.
The tax category AI model 16 may be configured as a sequence-to-category AI model, in which the input is a sequence of text or tokenized text, and the output is a single category. To this end, the tax category AI model 16 further includes a tokenizer 68. The tokenizer 68 is configured to tokenize the training product text 18AT to produce a set of tokenized text, which includes a sequence of tokens representing the words in the product name and product description for the product included in the training product text 18AT. Examples of suitable tokenizers include BlazingText, Word2Vec, WordPiece, etc.
The tax category AI model 16 further includes a neural network 70 configured to receive the tokenized text as an input vector 72. The neural network 70 may have an input layer, one or more convolutional layers, one or more fully connected hidden layers, and an output layer. It will be appreciated that the neural network 70 may be configured to include one or more convolutional layers, one or more fully connected hidden layers, or both one or more convolutional layers and one or more fully connected hidden layers. The input layer includes a plurality of nodes corresponding to the training phase product text 18AT. The output layer includes a plurality of output nodes corresponding to the ground truth classification 66, which indicates a training phase tax category 36T for the training data pair 64. Nodes in each layer are linked by weighted associations, and the tax category AI model 16 is configured to adjust internal weights to enhance one or more of the plurality of tokens in the sequence via a backpropagation algorithm according to a loss function during training to increase the accuracy of the output nodes during run-time.
In some implementations, the neural network 70 may be a transformer neural network (TNN). The TNN may be a sequence classifier that is configures to receive a sequence of tokens in the tokenized text and output a classification for the sequence of tokens. In this implementation, the neural network 70 has been pretrained to compute embeddings E for the input vector 72, i.e., tokenized text, via one or more embedding layers 74. The TNN is configured to compute attention weights for each of the embeddings E. In this way, it functions as an attention encoder. A classification layer is provided as an output layer for the TNN and includes a plurality of classifications, such as Tax Category 1, Tax Category 2, and so on. The final layer of the TNN is fully connected to the classification layer.
Alternatively, the neural network may be a long short-term memory (LSTM) recurrent neural network (RNN), which is also configured as a sequence classifier that receives a sequence of tokens in the tokenized text and outputs a classification for the sequence of tokens. In this implementation, an embedding technique such as Word2Vec can be used to create embeddings E in the embedding layer 74 for each token in the sequence of tokens provided as the input vector 72. The LSTM RNN has input, convolutional, and fully connected layers, the last of which is fully connected to each output node in the classification layer. To train the LSTM RNN, a suitable training algorithm such as backpropagation may be used to adjust the weights of the LSTM RNN for each ground truth classification and training product text 18AT that are input to the tax category AI model 16.
A schematic diagram of the tax category AI model 16 during a run-time inference phase is illustrated in
In some implementations, the predicted tax category 36 may be displayed by an evaluation module 78 that is executed by the processor 12. The evaluation module 78 may take the form of a graphical user interface (GUI) that is displayed on the client administrative device 20 when the predicted tax category 36 is output, for example. The GUI may be configured to receive a user input indicating that the predicted tax category 36 is correct prior to adopting the predicted tax category 36 for usage in the transaction tax engine 46. When the user confirms that the predicted tax category 36 is correct, the predicted tax category 36 is included in the taxability category mapping record 34.
The computing device 10 may be configured to collect the user input indicating whether the predicted tax category 36 is correct as user feedback 80. The feedback 80 may be collected via an implicit user feedback interface or an explicit user interface, such as the evaluation module 78, for example, and perform feedback training of the tax category AI model 16 based at least in part on the user feedback 80. As an example, explicit feedback may be collected via a selector in taxability category mappings editor, as discussed below with reference to the example explicit feedback shown in
Concurrent with the run-time input 76 received by the tax category AI model 16, the taxability category mapping engine 22 is configured to receive the product code 18B and taxpayer data 18C as supplemental run-time input 82 when the product catalog 18 is uploaded from the client administrative device 20. The supplemental run-time input 82 may include the taxability driver 16. As described above, the taxability category mapping engine 22 links the taxability driver 26 to the product. The taxability category mapping drivers record 30 is created for the product by the taxability category driver record association engine 28.
The taxability mapping record 34, including the predicted tax category 36, derived from the run-time input 76 and the taxability category mapping drivers record 30 derived from the supplemental run-time input 82 are included in the product taxability record 32 and output. As described above, the product taxability record 32 is output to the administrative interface 38 as product data 40, and the product data 40 and the client configuration settings 42 are communicated to a transaction tax engine 46 where they are stored as transaction tax engine configuration settings 48. The transaction tax engine 46 is in communication with an associated global tax rules database 50 that includes tax rate and rule data 52. Upon receiving a tax calculation request 54 from the transaction device 56, the transaction tax engine 46 is configured to process the tax calculation request 54 according to the transaction tax engine configuration settings 48 and the tax rate and rule data 52, and return a tax calculation response 60 to the transaction device 56.
As described above and shown in
The evaluation module 78 may be configured as an explicit user interface by which the client may provide user feedback 80 regarding the accuracy of the tax category mapping. As discussed above and shown in
As described above, the taxability driver 26 is received by the transaction tax engine 46, and used to filter the tax rate and rule data 52 based on the tax category 36 to determine an applicable tax rate and/or rule for the transaction. Example tax rates and rule data 52 applicable to the products in
At step 202, the method 200 may include, at one or more processors of a computing system in a run-time inference phase, receiving, as input for an artificial intelligence model, run-time input product text including a product name and product description from a product catalog associated with a product. The method 200 may further include tokenizing, by a tokenizer included in the artificial intelligence model, the product text to thereby produce tokenized text, and receiving, by a neural network included in the artificial intelligence model, the tokenized text. When the neural network is configured as a transformer neural network or a long short-term memory recurrent neural network, the method 200 may further include configuring the transformer neural network as a sequence classifier, receiving a sequence of tokens in the tokenized text, and outputting a classification for the sequence of tokens.
In an initial training phase, the method 200 may include training the artificial intelligence model with a training data set including a plurality of training data pairs, each training data pair including training product text associated with a training product catalog and a ground truth classification indicating a tax category for the product associated with the training product catalog.
Continuing from step 202 to step 204, the method 200 may include linking, by a taxability category mapping engine, a taxability to the product. The method 200 may further include receiving, by the taxability category mapping engine, a product code and taxpayer data from the product catalog associated with the product, the taxpayer data including a taxpayer code and taxpayer partition, and linking the taxability driver to the product based on the product code and taxpayer data.
Proceeding from step 204 to step 206, the method 200 may include creating, by a taxability category driver record association engine, a taxability category mapping drivers record including the taxability driver that is linked to the product.
Advancing from step 206 to step 208, the method 200 may include outputting, by the artificial intelligence model, a predicted tax category for the product. The method 200 may further include displaying, by an evaluation module, the predicted tax category, and receiving a user input indicating that the predicted tax category is correct prior to adopting the predicted tax category for usage in the tax transaction engine.
Continuing from step 208 to step 210, the method 200 may include storing the predicted tax category output from the artificial intelligence model and the taxability category mapping drivers record in a product taxability record. The method 200 may further include, at 212, extracting transaction tax engine configuration settings for a transaction tax engine from the product taxability record, and, at 214, provisioning a tax calculation engine of a tax server according to the configuration settings. The method 200 may further include, at 216, receiving, by the transaction tax engine, a tax calculation request from a transaction device, at 218, processing the tax calculation request according to the configuration settings, and at 220, transmitting a tax calculation response to the transaction device. The tax calculation response may be displayed and/or stored on the transaction device for further processing.
Using the above-described systems and methods, an artificial intelligence model can be trained and executed to predict a tax category for a product from a product catalog uploaded by a client, thereby enabling efficient and accurate tax calculations during a transaction. This model reduces the amount of time and money required for clients to manually map their products to a tax category. Clients are additionally spared the burden of deciphering and updating tax rates and rules in various jurisdictions. Such an architecture also has the advantage that fewer transactions may have an incorrect tax applied, thereby resulting in less tax burden that must be borne by the client due to failure to properly charge tax at the time of the transaction.
In some embodiments, the methods and processes described herein may be tied to a computing system of one or more computing devices. In particular, such methods and processes may be implemented as a computer-application program or service, an application-programming interface (API), a library, and/or other computer-program product.
Computing system 900 includes a logic processor 902 volatile memory 904, and a non-volatile storage device 906. Computing system 900 may optionally include a display subsystem 908, input subsystem 910, communication subsystem 912, and/or other components not shown in
Logic processor 902 includes one or more physical devices configured to execute instructions. For example, the logic processor may be configured to execute instructions that are part of one or more applications, programs, routines, libraries, objects, components, data structures, or other logical constructs. Such instructions may be implemented to perform a task, implement a data type, transform the state of one or more components, achieve a technical effect, or otherwise arrive at a desired result.
The logic processor may include one or more physical processors (hardware) configured to execute software instructions. Additionally or alternatively, the logic processor may include one or more hardware logic circuits or firmware devices configured to execute hardware-implemented logic or firmware instructions. Processors of the logic processor 902 may be single-core or multi-core, and the instructions executed thereon may be configured for sequential, parallel, and/or distributed processing. Individual components of the logic processor optionally may be distributed among two or more separate devices, which may be remotely located and/or configured for coordinated processing. Aspects of the logic processor may be virtualized and executed by remotely accessible, networked computing devices configured in a cloud-computing configuration. In such a case, these virtualized aspects are run on different physical logic processors of various different machines, it will be understood.
Non-volatile storage device 906 includes one or more physical devices configured to hold instructions executable by the logic processors to implement the methods and processes described herein. When such methods and processes are implemented, the state of non-volatile storage device 906 may be transformed, e.g., to hold different data.
Non-volatile storage device 906 may include physical devices that are removable and/or built-in. Non-volatile storage device 906 may include optical memory (e.g., CD, DVD, HD-DVD, Blu-Ray Disc, etc.), semiconductor memory (e.g., ROM, EPROM, EEPROM, FLASH memory, etc.), and/or magnetic memory (e.g., hard-disk drive, floppy-disk drive, tape drive, MRAM, etc.), or other mass storage device technology. Non-volatile storage device 906 may include nonvolatile, dynamic, static, read/write, read-only, sequential-access, location-addressable, file-addressable, and/or content-addressable devices. It will be appreciated that non-volatile storage device 906 is configured to hold instructions even when power is cut to the non-volatile storage device 906.
Volatile memory 904 may include physical devices that include random access memory. Volatile memory 904 is typically utilized by logic processor 902 to temporarily store information during processing of software instructions. It will be appreciated that volatile memory 904 typically does not continue to store instructions when power is cut to the volatile memory 904.
Aspects of logic processor 902, volatile memory 904, and non-volatile storage device 906 may be integrated together into one or more hardware-logic components. Such hardware-logic components may include field-programmable gate arrays (FPGAs), program- and application-specific integrated circuits (PASIC/ASICs), program- and application-specific standard products (PSSP/ASSPs), system-on-a-chip (SOC), and complex programmable logic devices (CPLDs), for example.
The terms “module,” “program,” and “engine” may be used to describe an aspect of computing system 900 typically implemented in software by a processor to perform a particular function using portions of volatile memory, which function involves transformative processing that specially configures the processor to perform the function. Thus, a module, program, or engine may be instantiated via logic processor 902 executing instructions held by non-volatile storage device 906, using portions of volatile memory 904. It will be understood that different modules, programs, and/or engines may be instantiated from the same application, service, code block, object, library, routine, API, function, etc. Likewise, the same module, program, and/or engine may be instantiated by different applications, services, code blocks, objects, routines, APIs, functions, etc. The terms “module,” “program,” and “engine” may encompass individual or groups of executable files, data files, libraries, drivers, scripts, database records, etc.
When included, display subsystem 908 may be used to present a visual representation of data held by non-volatile storage device 906. The visual representation may take the form of a graphical user interface (GUI). As the herein described methods and processes change the data held by the non-volatile storage device, and thus transform the state of the non-volatile storage device, the state of display subsystem 908 may likewise be transformed to visually represent changes in the underlying data. Display subsystem 908 may include one or more display devices utilizing virtually any type of technology. Such display devices may be combined with logic processor 902, volatile memory 904, and/or non-volatile storage device 906 in a shared enclosure, or such display devices may be peripheral display devices.
When included, input subsystem 910 may comprise or interface with one or more user-input devices such as a keyboard, mouse, touch screen, or game controller. In some embodiments, the input subsystem may comprise or interface with selected natural user input (NUI) componentry. Such componentry may be integrated or peripheral, and the transduction and/or processing of input actions may be handled on- or off-board. Example NUI componentry may include a microphone for speech and/or voice recognition; an infrared, color, stereoscopic, and/or depth camera for machine vision and/or gesture recognition; a head tracker, eye tracker, accelerometer, and/or gyroscope for motion detection and/or intent recognition; as well as electric-field sensing componentry for assessing brain activity; and/or any other suitable sensor.
When included, communication subsystem 912 may be configured to communicatively couple various computing devices described herein with each other, and with other devices. Communication subsystem 912 may include wired and/or wireless communication devices compatible with one or more different communication protocols. As non-limiting examples, the communication subsystem may be configured for communication via a wireless telephone network, or a wired or wireless local- or wide-area network, such as a HDMI over Wi-Fi connection. In some embodiments, the communication subsystem may allow computing system 900 to send and/or receive messages to and/or from other devices via a network such as the Internet.
It will be understood that the configurations and/or approaches described herein are exemplary in nature, and that these specific embodiments or examples are not to be considered in a limiting sense, because numerous variations are possible. The specific routines or methods described herein may represent one or more of any number of processing strategies. As such, various acts illustrated and/or described may be performed in the sequence illustrated and/or described, in other sequences, in parallel, or omitted. Likewise, the order of the above-described processes may be changed.
The subject matter of the present disclosure includes all novel and non-obvious combinations and sub-combinations of the various processes, systems and configurations, and other features, functions, acts, and/or properties disclosed herein, as well as any and all equivalents thereof.
