System and method of producing a structured assessment of an object using contextual factors relating to the object
The system addresses the lack of systematic community impact assessment by using AI and a taxonomy framework to provide accurate, relevant, and adaptive community insights.
Patent Information
- Application Number
- PCT/US2026/028470
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-05-15
- Filing Date
- 2026-05-15
- Publication Date
- 2026-11-19
AI Technical Summary
Current methods lack a systematic, scalable way to understand the impact of residential communities on human health and wellbeing, integrating environmental, social, and economic factors, and adapt to real-time resident experiences.
A system using service-oriented architecture (SOA) components with artificial intelligence (AI) processes contextual data through a taxonomy framework to generate structured assessments, including large language models (LLMs) for actionable insights and continuous adaptation.
Enhances accuracy, relevance, and trustworthiness of community assessments, enabling personalized guidance and strategic recommendations, while allowing for continuous feedback and model evolution.
Smart Images

Figure US2026028470_19112026_PF_FP_ABST
Abstract
Description
SYSTEM AND METHOD OF PRODUCING A STRUCTURED ASSESSMENT OF AN OBJECT USING CONTEXTUAL FACTORS RELATING TO THE OBJECTCross-Reference to Related Applications
[0001] This application claims the benefit of U.S. Provisional Patent Application No.63 / 806,482, filed May 15, 2025, the entirety of which is incorporated by reference herein.Technical Field
[0002] This disclosure relates generally to the field of efficient building and analysis of data, and relates more specifically to the field of building and analysis of data relating to an object, a context in which the object exists, and potential users of the object.Background
[0003] Planners of community projects, e.g. real estate projects, have historically had no systematic way to understand how the project, and its location, relate to a robust and articulated set of human health and wellbeing factors that will affect users of the project, e.g. actual or potential residents thereof. Using current technological solutions, market analyses focus on comparable sales and demographics. Environmental reports focus on physical conditions. Neither integrates the policy landscape, the available community programs, the geographic assets, or the documented health and social needs of the surrounding area into a unified, actionable picture.
[0004] Human beings face a range of health, wellbeing, social, and economic challenges that are increasingly understood to be deeply influenced by the environments in which they live. Research has shown that factors such as physical health, mental wellbeing, child development, aging, social connection, safety, and economic stability are all impacted by the quality and design of the home and broader community environment.
[0005] However, despite this recognized importance, there is currently a lack of systematic, scalable methods for capturing the “impact potential” of residential communities, translating it into actionable insights, and continuously adapting it in response to real resident experience.
[0006] Traditional methods of resident engagement (such as surveys, focus groups, or onetime assessments) are often static, limited in scope, and disconnected from dynamic intervention opportunities.
[0007] Advances in artificial intelligence, particularly large language models, offer a powerful opportunity to bridge this gap. However, general-purpose Al models are not optimized for the nuanced, place-specific, human-centered knowledge needed to support residential impact strategies.
[0008] Accordingly, there is a need to gather data from multiple sources, organize it across a taxonomy of contextual categories, and using artificial intelligence agents, such as large language models, to synthesize findings into structured assessments, strategic recommendations, and a vision for the development.Summary
[0009] A system of one or more computers can be configured to perform particular operations or actions by virtue of having software, firmware, hardware, or a combination of them installed on the system that in operation causes or cause the system to perform the actions. One or more computer programs can be configured to perform particular operations or actions by virtue of including instructions that, when executed by data processing apparatus, cause the apparatus to perform the actions.
[0010] In some embodiments of the present disclosure, a system is disclosed. The system includes a networked computing environment including one or more servers configured to implement service-oriented architecture (“SOA”) components for producing an artificial intelligence based structured assessment of an object using contextual factors relating to the object. The SOA components include a first SOA back-end component of the one or more servers configured to receive an input comprising contextual input data indicating a contextual environment in which the object exists, query a first proximal asset database, using the contextual input data and a proximity distance to identify a plurality of proximal assets representing environmental aspects present within the proximity distance of the object, the first proximal asset database returning an asset type for each of the plurality of proximal assets, for each of the plurality of proximal assets, infer a contextual category, from a taxonomy of contextual categories, fromthe asset type, and store each result from the first proximal asset database, and the inferred contextual category, as a proximal asset finding in a storage medium.
[0011] The SOA components include a second SOA back-end component of the one or more servers configured to query a second contextual database, using the contextual input data, to return a summary environmental data structure relating to the contextual environment within the proximity distance, use an artificial intelligence large language model (“LLM”) to create one or more summary environmental findings data structures relating the summary environmental data structure to one or more of the taxonomy of contextual categories, using a deterministic, category-routed architecture, and store the one or more summary findings relating to the summary environmental data structures in the storage medium.
[0012] The SOA components include a third SOA back-end component of the one or more servers configured to select at least one environmental scope, each environmental scope indicating a boundary for environmental context research, for each pair of a selected environmental scope and an item on the taxonomy of contextual categories, build a natural language query for determining an extent to which assets matching the item are present within the selected environmental scope, process each natural language query using an artificial intelligence answer engine, to create an answer engine output signal comprising the extent to which assets matching the item are present within the selected environmental scope, process each answer engine output with the LLM, to create a contextual finding data structure relating to the extent to which assets matching the item are present within the selected environmental scope, and store the one or more structured contextual finding data structures to the storage medium.
[0013] The SOA components include a fourth SOA back-end component of the one or more servers configured to, for each proximal asset finding, each summary finding, and each structured contextual finding: sort the findings by environmental scope, construct a prompt including all findings, and at least one objective of the object, transmit the prompt to the LLM, generating a final structured output, the final structured output including a summary of the object and one or more listings of strengths, weaknesses, and opportunities relating to one or more items on the taxonomy of contextual categories, and store the final structured output in the storage medium.
[0014] In some embodiments, the third SOA back-end component may be further configured to, prior to processing each natural language query using the artificial intelligence answer engine, process the natural language query on a context cache to determine if the natural language query had already been processed by the artificial intelligence answer engine, and store the one or more structured contextual findings to the context cache.1. In some embodiments, the fourth SOA back-end component may be further configured to identify multi-category findings in the storage medium that apply to multiple items on the taxonomy of contextual categories, and synthesizing an executive-level strategic priority based on the multi -category findings. In some embodiments, the fourth SOA back-end component may be further configured to prompt the LLM to generate a vision statement for the object. In some embodiments, the fourth SOA back-end component may be further configured to provide an editing interface for the vision statement. In some embodiments, the fourth SOA back-end component may be further configured to detect contradictions between a first of the findings having a first environmental scope and a second of the findings having a second environmental scope. In some embodiments, the fourth SOA back-end component may be further configured to evaluate a user-submitted text for alignment with the vision statement.
[0015] Some embodiments may further comprising a fifth SOA back-end component configured to determine compatibility between the object and a potential user of the object, the determining compatibility comprising receiving a user consumption input relating to potential consumption of the object by the potential user, the user consumption inputs comprising user consumption information and user priority information, using the LLM to generate user preference data relating to at least some of the taxonomy of contextual categories, and prompting the LLM to create a user fit report, wherein the prompting comprises pointing the LLM to the user preference data and the findings and structing the LLM to create a comparison report comparing the user consumption information, the user priority information, and at least one of the proximal asset findings, the summary finding, and the structured contextual findings.
[0016] In some embodiments, the fourth SO A back-end component may be further configured to prompt the LLM to generate a structured set of actional strategies organized by contextual category and by scope.
[0017] Other embodiments may include corresponding computer systems, apparatus, and computer programs recorded on one or more computer storage devices, each configured to perform the actions of the methods.Brief Description of the Drawings
[0018] The foregoing summary, as well as the following detailed description of the invention, will be better understood when read in conjunction with the appended drawings. For the purpose of illustrating the invention, there are shown in the drawings embodiments which are presently preferred. It should be understood, however, that the invention can be embodied in different forms and thus should not be construed as being limited to the embodiments set forth herein. The appended drawings may not be drawn to scale.
[0019] FIG. 1 illustrates a network environment configured to provide a system and method of producing a structured assessment of an object using contextual factors relating to the object, in accordance with some embodiments;
[0020] FIG. 2 illustrates a computer system configured to implement one or more processes, in accordance with some embodiments;
[0021] FIG. 3 is a flow diagram of three independent data gathering pipelines in accordance with some embodiments;
[0022] FIG. 4 is a flow diagram of a synthesis pipeline in accordance with some embodiments;
[0023] FIG. 5 is a flow diagram of resident intake and personalized report generation in accordance with some embodiments;
[0024] FIG. 6 is a structural diagram of an LLM formed in a transformer architecture, which may be utilized by at least some systems and methods described herein, in accordance with some embodiments.Detailed Description
[0025] The present disclosure provides a system and method of producing an artificial intelligence based structured assessment of an object using contextual factors relating to the object. In some implementations, the object may be a location in geographic space, such as a home, an apartment complex, or other real estate development. A taxonomic framework of contextual factors may be used, which may be grouped into families. In a real estate contexts, contextual factor families may include environmental, medical, social, and developmental. Data may be tagged against the taxonomy after it is gathered. Summaries may be generated in accordance with the taxonomy, data retrieval may be routed through the taxonomy, and user-facing, e.g. residentfacing and developer-facing, interfaces may be organized by it. The consistent application of such a shared taxonomy across ingestion, storage, retrieval, generation, and display drives the system’s coherence and auditability, and avoids hallucinations that may plague other artificial intelligence based systems.
[0026] The system serves two primary user classes, developers and residents, with architecturally distinct interfaces built atop a shared impact representation of one or more objects, e.g., property addresses.
[0027] In one aspect, the present disclosure comprises the collection, structuring, and formatting of universal impact-related content and data according to a predefined impact framework, which organizes human impact drivers into categories such as Home Environment, Built Environment, Transportation and Mobility, Nutrition, Physical Activity, Sleep, Stress Reduction, Preventive Healthcare, Human Connection, Faith and Spirituality, and other related domains, collectively drawn from fields such as real estate, population health, and social sciences, among others.
[0028] In a further aspect, the disclosure comprises the collection, structuring, and formatting of community and residential unit-specific information related to a particular home, residential building, neighborhood, or larger residential community. Such information includes, but is not limited to, architectural features, landscape elements, amenities, programming, service partners, resident demographics, and impact-relevant outcomes or usage patterns. In an alternative or complementary embodiment, the system organizes and assesses the residentialunit or community based on a structured taxonomy of physical spaces, structural features, and environmental elements, collectively referred to as contextual categories. Contextual categories may also in some references herein be referred to as “pathways.”
[0029] In the context of a geographic object, contextual categories may be associated with specific resident behaviors and choices, potential impact measurements, success indicators, recommended programming, partner types, incentive strategies, narrative storytelling prompts, supporting literature references, and the like.
[0030] Structuring contextual category information allows the intelligent agent to evaluate the presence, quality, and usage of specific contextual categories within a given object, identify opportunities for activation or enhancement, generate personalized resident guidance, impact reports, storytelling content, fundraising materials, and programming recommendations accordingly. Contextual categories may, in some embodiments, be grouped or otherwise organized into contextual domains.
[0031] In some embodiments, the present disclosure may include combining universal impact framework data with the residential unit-specific data to train, fine-tune, or augment a large language model (LLM), to produce a customized intelligent agent representing that specific residential environment and capable of interacting with residents, leadership, partners, and other stakeholders. In a further aspect, the present disclosure may enable the intelligent agent to be configured in a first-person, personified mode, wherein the agent adopts the persona and narrative voice of the residential unit or community it represents. In this mode, the agent may be configured to communicate as though it were the community itself, using first-person language (e.g., “I am a neighborhood designed for connection,” “I offer trails for your morning runs,” etc.). This capability differs materially from traditional agents that describe communities from an external or third-party perspective. The personification approach enhances emotional connection, storytelling richness, community pride, and resident engagement.
[0032] In contrast to general-purpose language models, which are susceptible to factual inaccuracies and hallucinations, the present disclosure systematically anchors the model's outputs in a corpus of structured, pl ace- specific information. The system of the present disclosure greatly enhances the accuracy, relevance, and trustworthiness of outputs produced by the model, enablingsafer and more actionable community use. Moreover, the system of the present disclosure creates an opportunity for scale-based information advantages. As more communities participate and contribute structured data, aggregate insights, cross-community benchmarking, model enrichment, and longitudinal trend analyses may result. This level of insight and coordination would be prohibitively difficult for individual communities to achieve independently through isolated or ad hoc efforts.
[0033] The system of the present disclosure may be configured to produce a variety of outputs, including but not limited to Resident survey analyses and insight reports, impact assessments and benchmarking dashboards, personalized resident guidance, personal narrative storytelling and testimonial generation, marketing and public relations materials, fundraising and philanthropic case materials, resident-facing informational content, recommendations for community design and feature enhancements, programming and partnership recommendations, and educational and behavior change content.
[0034] In a further aspect, the present disclosure may also include a continuous feedback mechanism whereby new resident feedback, updated environmental conditions, programmatic changes, or outcome measurements are systematically captured, structured, and reintegrated into the model. This enables the intelligent agent to evolve and remain reflective of the living conditions, needs, and aspirations of the residential community overtime.
[0035] In various embodiments, a community-specific intelligent agent may be implemented in multiple forms, including but not limited to: a custom-trained large language model interface (such as a Custom ChatGPT agent), a chatbot, a standalone software application, a mobile application, an interactive web-based platform, or an embedded module within smart home or community infrastructure. The agent may be accessed through a variety of devices, including personal computers, laptops, smartphones, tablets, smart home devices, dedicated community kiosks, wearable technology, or other digital interfaces. In other embodiments, existing agents or models may be used with specified prompting strategies and data inputs to produce the desired results including improved comprehensive outputs, efficient operation, and reduced hallucination.
[0036] For any given object, the system begins with collecting contextual input data indicating a contextual environment in which the object exists. Contextual input data may be collected from a user via a user interface wizard. Other data collection methods may be used in lieu of or in addition to a user interface wizard. In some embodiments the contextual input data may only collect information sufficient to identify the contextual environment. E.g., if the object is a building or a real estate development, the wizard may only collect an address or other locating information such as latitude and longitude coordinates.
[0037] FIG. 1 illustrates a network environment 2 configured to provide a system and method of producing a structured assessment of an object using contextual factors relating to the object, in accordance with some embodiments. A network environment 2 may include one or more servers configured to implement SOA (Service-Oriented Architecture) components. The network environment 2 includes a plurality of devices or systems configured to communicate over one or more network channels, illustrated as a network cloud 22. For example, in various embodiments, the network environment 2 may include, but is not limited to, a gathering and synthesis computing device 4, a web server 6, a cloud-based engine 8 including one or more processing devices 10, a database 14, and / or one or more user computing devices 16, 18, 20 operatively coupled over the network 22. The gathering and synthesis computing device 4, the web server 6, the processing device(s) 10, and / or the user computing devices 16, 18, 20 may each be a suitable computing device that includes any hardware or hardware and software combination for processing and handling information. For example, each computing device may include, but is not limited to, one or more processors, one or more field-programmable gate arrays (FPGAs), one or more application-specific integrated circuits (ASICs), one or more state machines, digital circuitry, and / or any other suitable circuitry. In addition, each computing device may transmit and receive data over the communication network 22.
[0038] In some embodiments, each of the gathering and synthesis computing device 4 and the processing device(s) 10 may be a computer, a workstation, a laptop, a server such as a cloudbased server, or any other suitable device. In some embodiments, each of the processing devices 10 is a server that includes one or more processing units, such as one or more graphical processing units (GPUs), one or more central processing units (CPUs), and / or one or more processing cores. Each processing device 10 may, in some embodiments, execute one or more virtual machines. Insome embodiments, processing resources (e.g, capabilities) of the one or more processing devices 10 are offered as a cloud-based service (e.g., cloud computing). For example, the cloud-based engine 8 may offer computing and storage resources of the one or more processing devices 10 to the gathering and synthesis computing device 4.
[0039] In some embodiments, each of the user computing devices 16, 18, 20 may be a cellular phone, a smart phone, a tablet, a personal assistant device, a voice assistant device, a digital assistant, a laptop, a computer, or any other suitable device. In some embodiments, the web server 6 hosts one or more network environments, such as an e-commerce network environment. In some embodiments, the gathering and synthesis computing device 4, the processing devices 10, and / or the web server 6 are operated by the network environment provider, and the user computing devices 16, 18, 20 are operated by users of the network environment. In some embodiments, the processing devices 10 are operated by a third party (e.g., a cloud-computing provider).
[0040] Although FIG. 1 illustrates three user computing devices 16, 18, 20, the network environment 2 may include any number of user computing devices 16, 18, 20. Similarly, the network environment 2 may include any number of the gathering and synthesis computing device 4, the web server 6, the processing devices 10, and / or the databases 14. It will further be appreciated that additional systems, servers, storage mechanism, etc. may be included within the network environment 2. In addition, although embodiments are illustrated herein having individual, discrete systems, it will be appreciated that, in some embodiments, one or more systems may be combined into a single logical and / or physical system. For example, in various embodiments, one or more of the gathering and synthesis computing device 4, the web server 6, the database 14, the user computing devices 16, 18, 20, and / or the router 24 may be combined into a single logical and / or physical system. Similarly, although embodiments are illustrated having a single instance of each device or system, it will be appreciated that additional instances of a device may be implemented within the network environment 2. In some embodiments, two or more systems may be operated on shared hardware in which each system operates as a separate, discrete system utilizing the shared hardware, for example, according to one or more virtualization schemes.
[0041] The communication network 22 may be a WiFi® network, a cellular network such as a 3GPP® network, a Bluetooth® network, a satellite network, a wireless local area network (LAN), a network utilizing radio-frequency (RF) communication protocols, a Near Field Communication (NFC) network, a wireless Metropolitan Area Network (MAN) connecting multiple wireless LANs, a wide area network (WAN), or any other suitable network. The communication network 22 may provide access to, for example, the Internet.
[0042] Each of the user computing devices 16, 18, 20 may communicate with the web server 6 over the communication network 22. For example, each of the user computing devices 16, 18, 20 may be operable to view, access, and interact with a website, such as an e-commerce website, hosted by the web server 6. The web server 6 may transmit user session data related to a user’s activity (e.g., interactions) on the website. For example, a user may operate one of the user computing devices 16, 18, 20 to initiate a web browser that is directed to the website hosted by the web server 6. The user may, via the web browser, perform various operations such as searching one or more databases or catalogs associated with the displayed website, view data for elements associated with and displayed on the website, and click on interface elements presented via the website. The website may capture these activities as user session data, and transmit the user session data to the gathering and synthesis computing device 4 over the communication network 22. The website may also allow the user to interact with one or more of interface elements to perform specific operations, such as selecting one or more elements for further processing. In some embodiments, the web server 6 transmits user interaction data identifying interactions between the user and the website to the gathering and synthesis computing device 4.
[0043] In some embodiments, the gathering and synthesis computing device 4 may execute one or more models, processes, or algorithms, such as a machine learning model, deep learning model, statistical model, etc., to gather data from disparate sources and synthesize it in a consumable form, as described in more detail below.
[0044] The gathering and synthesis computing device 4 is further operable to communicate with the database 14 over the communication network 22. For example, the gathering and synthesis computing device 4 may store data to, and read data from, the database 14. The database 14 may be a remote storage device, such as a cloud-based server, a disk (e.g., a hard disk), a memorydevice on another application server, a networked computer, or any other suitable remote storage. Although shown remote to the gathering and synthesis computing device 4, in some embodiments, the database 14 may be a local storage device, such as a hard drive, a non-volatile memory, or a USB stick. The gathering and synthesis computing device 4 may store interaction data received from the web server 6 in the database 14. The gathering and synthesis computing device 4 may also receive from the web server 6 user session data identifying events associated with browsing sessions, and may store the user session data in the database 14.
[0045] In some embodiments, gathering and synthesis computing device 4 may interact with one or more artificial intelligence models via cloud based engine 8, which may include one or more processing devices 10 which may provide, run, and train the models. The models may accept queries and prompts to perform discrete tasks as described below. A system 2 implementing the present disclosure may interact with one or more instances of cloud based engine 8, each of which may be different from the others in various ways as a person having skill in the art would understand.
[0046] In some embodiments, the gathering and synthesis computing device 4 assigns the models (or parts thereof) for execution to one or more processing devices 10. For example, each model may be assigned to a virtual machine hosted by a processing device 10. The virtual machine may cause the models or parts thereof to execute on one or more processing units such as GPUs. In some embodiments, the virtual machines assign each model (or part thereof) among a plurality of processing units. Based on the output of the models, gathering and synthesis computing device 4 may generate proximal asset findings, summary findings, structured contextual finding, or the like as described below.
[0047] In some embodiments, web server 6 may provide a browser-based interface at a client level. In some embodiments, gathering and synthesis computing device 4 may provide an API layer, business logic, and artificial intelligence orchestration. In some embodiments, database 14 may provide persistent storage for application state, and my be implemented with database environments such as postgre SQL or Drizzle ORM. In some embodiments, cloud based engine 8 may include one or more research Al models such as Perplexity, and may also include one or more synthesis Al models such as OpenAI GPT. Gathering and synthesis computing device 4 may alsointeract with other web server 6 which may include data sources such as geographic and demographic data sources, e.g. Google Maps Places API or the US Census API.
[0048] FIG. 2 illustrates a block diagram of a computing device 50, in accordance with some embodiments. In some embodiments, each of the gathering and synthesis computing device 4, the web server 6, the one or more processing devices 10, the workstation(s) 12, and / or the user computing devices 16, 18, 20 in FIG. 1 may include the features shown in FIG. 2. Although FIG.2 is described with respect to certain components shown therein, it will be appreciated that the elements of the computing device 50 may be combined, omitted, and / or replicated. In addition, it will be appreciated that additional elements other than those illustrated in FIG. 2 may be added to the computing device.
[0049] As shown in FIG. 2, the computing device 50 may include one or more processors 52, an instruction memory 54, a working memory 56, one or more input / output devices 58, a transceiver 60, one or more communication ports 62, a display 64 with a user interface 66, and an optional location device 68, all operatively coupled to one or more data buses 70. The data buses 70 allow for communication among the various components. The data buses 70 may include wired, or wireless, communication channels.
[0050] The one or more processors 52 may include any processing circuitry operable to control operations of the computing device 50. In some embodiments, the one or more processors 52 include one or more distinct processors, each having one or more cores (e.g., processing circuits). Each of the distinct processors may have the same or different structure. The one or more processors 52 may include one or more central processing units (CPUs), one or more graphics processing units (GPUs), application specific integrated circuits (ASICs), digital signal processors (DSPs), a chip multiprocessor (CMP), a network processor, an input / output (VO) processor, a media access control (MAC) processor, a radio baseband processor, a co-processor, a microprocessor such as a complex instruction set computer (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, and / or a very long instruction word (VLIW) microprocessor, or other processing device. The one or more processors 52 may also be implemented by a controller, a microcontroller, an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a programmable logic device (PLD), etc.
[0051] In some embodiments, the one or more processors 52 are configured to implement an operating system (OS) and / or various applications. Examples of an OS include, for example, operating systems generally known under various trade names such as Apple macOS™, Microsoft Windows™, Android™, Linux™, and / or any other proprietary or open-source OS. Examples of applications include, for example, network applications, local applications, data input / output applications, user interaction applications, etc.
[0052] The instruction memory 54 may store instructions that are accessed ( .g., read) and executed by at least one of the one or more processors 52. For example, the instruction memory 54 may be a non-transitory, computer-readable storage medium such as a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), flash memory (e.g. NOR and / or NAND flash memory), content addressable memory (CAM), polymer memory (e.g., ferroelectric polymer memory), phase-change memory (e.g., ovonic memory), ferroelectric memory, silicon-oxide-nitride-oxide-silicon (SONOS) memory, a removable disk, CD-ROM, any non-volatile memory, or any other suitable memory. The one or more processors 52 may be configured to perform a certain function or operation by executing code, stored on the instruction memory 54, embodying the function or operation. For example, the one or more processors 52 may be configured to execute code stored in the instruction memory 54 to perform one or more of any function, method, or operation disclosed herein.
[0053] Additionally, the one or more processors 52 may store data to, and read data from, the working memory 56. For example, the one or more processors 52 may store a working set of instructions to the working memory 56, such as instructions loaded from the instruction memory 54. The one or more processors 52 may also use the working memory 56 to store dynamic data created during one or more operations. The working memory 56 may include, for example, random access memory (RAM) such as a static random access memory (SRAM) or dynamic random access memory (DRAM), Double-Data-Rate DRAM (DDR-RAM), synchronous DRAM (SDRAM), an EEPROM, flash memory (e.g. NOR and / or NAND flash memory), content addressable memory (CAM), polymer memory (e.g., ferroelectric polymer memory), phasechange memory (e.g., ovonic memory), ferroelectric memory, silicon-oxide-nitride-oxide-silicon (SONOS) memory, a removable disk, CD-ROM, any non-volatile memory, or any other suitable memory. Although embodiments are illustrated herein including separate instruction memory 54and working memory 56, it will be appreciated that the computing device 50 may include a single memory unit configured to operate as both instruction memory and working memory. Further, although embodiments are discussed herein including non-volatile memory, it will be appreciated that computing device 50 may include volatile memory components in addition to at least one nonvolatile memory component.
[0054] In some embodiments, the instruction memory 54 and / or the working memory 56 includes an instruction set, in the form of a file for executing various methods, such as methods for producing a artificial intelligence based structured assessment of an object using contextual factors relating to the object, as described herein. The instruction set may be stored in any acceptable form of machine-readable instructions, including source code or various appropriate programming languages. Some examples of programming languages that may be used to store the instruction set include, but are not limited to: Java, JavaScript, C, C++, C#, Python, Objective-C, Visual Basic, .NET, HTML, CSS, SQL, NoSQL, Rust, Perl, etc. In some embodiments a compiler or interpreter is configured to convert the instruction set into machine executable code for execution by the one or more processors 52.
[0055] The input-output devices 58 may include any suitable device that allows for data input or output. For example, the input-output devices 58 may include one or more of a keyboard, a touchpad, a mouse, a stylus, a touchscreen, a physical button, a speaker, a microphone, a keypad, a click wheel, a motion sensor, a camera, and / or any other suitable input or output device.
[0056] The transceiver 60 and / or the communication port(s) 62 allow for communication with a network, such as the communication network 22 of FIG. 1. For example, if the communication network 22 of FIG. 1 is a cellular network, the transceiver 60 is configured to allow communications with the cellular network. In some embodiments, the transceiver 60 is selected based on the type of the communication network 22 the computing device 50 will be operating in. The one or more processors 52 are operable to receive data from, or send data to, a network, such as the communication network 22 of FIG. 1, via the transceiver 60.
[0057] The communication port(s) 62 may include any suitable hardware, software, and / or combination of hardware and software that is capable of coupling the computing device 50 to one or more networks and / or additional devices. The communication port(s) 62 may be arranged tooperate with any suitable technique for controlling information signals using a desired set of communications protocols, services, or operating procedures. The communication port(s) 62 may include the appropriate physical connectors to connect with a corresponding communications medium, whether wired or wireless, for example, a serial port such as a universal asynchronous receiver / transmitter (UART) connection, a Universal Serial Bus (USB) connection, or any other suitable communication port or connection. In some embodiments, the communication port(s) 62 allows for the programming of executable instructions in the instruction memory 54. In some embodiments, the communication port(s) 62 allow for the transfer (e. ., uploading or downloading) of data, such as machine learning model training data.
[0058] In some embodiments, the communication port(s) 62 are configured to couple the computing device 50 to a network. The network may include local area networks (LAN) as well as wide area networks (WAN) including without limitation Internet, wired channels, wireless channels, communication devices including telephones, computers, wire, radio, optical and / or other electromagnetic channels, and combinations thereof, including other devices and / or components capable of / associated with communicating data. For example, the communication environments may include in-body communications, various devices, and various modes of communications such as wireless communications, wired communications, and combinations of the same.
[0059] In some embodiments, the transceiver 60 and / or the communication port(s) 62 are configured to utilize one or more communication protocols. Examples of wired protocols may include, but are not limited to, Universal Serial Bus (USB) communication, RS-232, RS-422, RS-423, RS-485 serial protocols, FireWire, Ethernet, Fibre Channel, MIDI, ATA, Serial ATA, PCI Express, T-l (and variants), Industry Standard Architecture (ISA) parallel communication, Small Computer System Interface (SCSI) communication, or Peripheral Component Interconnect (PCI) communication, etc. Examples of wireless protocols may include, but are not limited to, the Institute of Electrical and Electronics Engineers (IEEE) 8O2.xx series of protocols, such as IEEE 802.1 la / b / g / n / ac / ag / ax / be, IEEE 802.16, IEEE 802.20, GSM cellular radiotelephone system protocols with GPRS, CDMA cellular radiotelephone communication systems with IxRTT, EDGE systems, EV-DO systems, EV-DV systems, HSDPA systems, Wi-Fi Legacy, Wi-Fi 1 / 2 / 3 / 4 / 5 / 6 / 6E, wireless personal area network (PAN) protocols, Bluetooth Specification versions5.0, 6, 7, legacy Bluetooth protocols, passive or active radio-frequency identification (RFID) protocols, Ultra-Wide Band (UWB), Digital Office (DO), Digital Home, Trusted Platform Module (TPM), ZigBee, etc.
[0060] The display 64 may be any suitable display, and may display the user interface 66. The user interfaces 66 may enable user interaction with the system 2 of FIG. 1. For example, the user interface 66 may be a user interface for an application of a network environment operatorthat allows a user to view and interact with the operator’s website. In some embodiments, a user may interact with the user interface 66 by engaging the input-output devices 58. In some embodiments, the display 64 may be a touchscreen, where the user interface 66 is displayed on the touchscreen.
[0061] The display 64 may include a screen such as, for example, a Liquid Crystal Display (LCD) screen, a light-emitting diode (LED) screen, an organic LED (OLED) screen, a movable display, a projection, etc. In some embodiments, the display 64 may include a coder / decoder, also known as Codecs, to convert digital media data into analog signals. For example, the visual peripheral output device may include video Codecs, audio Codecs, or any other suitable type of Codec.
[0062] The optional location device 68 may be communicatively coupled to a location network and operable to receive position data from the location network. For example, in some embodiments, the location device 68 includes a GPS device configured to receive position data identifying a latitude and longitude from one or more satellites of a GPS constellation. As another example, in some embodiments, the location device 68 is a cellular device configured to receive location data from one or more localized cellular towers. Based on the position data, the computing device 50 may determine a local geographical area (e.g., town, city, state, etc.) of its position.
[0063] In some embodiments, the computing device 50 is configured to implement one or more modules or engines, each of which is constructed, programmed, configured, or otherwise adapted, to autonomously carry out a function or set of functions. A module / engine may include a component or arrangement of components implemented using hardware, such as by an application specific integrated circuit (ASIC) or field-programmable gate array (FPGA), for example, or as a combination of hardware and software, such as by a microprocessor system and a set of program instructions that adapt the module / engine to implement the particular functionality, which (whilebeing executed) transform the microprocessor system into a special -purpose device. A module / engine may also be implemented as a combination of the two, with certain functions facilitated by hardware alone, and other functions facilitated by a combination of hardware and software. In certain implementations, at least a portion, and in some cases, all, of a module / engine may be executed on the processor(s) of one or more computing platforms that are made up of hardware (e.g., one or more processors, data storage devices such as memory or drive storage, input / output facilities such as network interface devices, video devices, keyboard, mouse or touchscreen devices, etc.) that execute an operating system, system programs, and application programs, while also implementing the engine using multitasking, multithreading, distributed (e.g., cluster, peer-peer, cloud, etc.) processing where appropriate, or other such techniques. Accordingly, each module / engine may be realized in a variety of physically realizable configurations, and should generally not be limited to any particular implementation exemplified herein, unless such limitations are expressly called out. In addition, a module / engine may itself be composed of more than one sub- modules or sub-engines, each of which may be regarded as a module / engine in its own right. Moreover, in the embodiments described herein, each of the various modules / engines corresponds to a defined autonomous functionality; however, it should be understood that in other contemplated embodiments, each functionality may be distributed to more than one module / engine. Likewise, in other contemplated embodiments, multiple defined functionalities may be implemented by a single module / engine that performs those multiple functions, possibly alongside other functions, or distributed differently among a set of modules / engines than specifically illustrated in the embodiments herein.
[0064] FIG. 3 a flow diagram of three independent data gathering pipelines in accordance with some embodiments. As shown in FIG. 3, Route A, Route B and Route C are independent HTTP routes, not a single coordinated operation. Each route may be called separately, without a server-side “Promise. all” wrapping all three.
[0065] In Route A, a first proximal asset database is queried, using the contextual input data and a proximity distance to identify a plurality of proximal assets representing environmental aspects present within the contextual distance of the object. In step 100, a “ / gather” operation is invoked. In some embodiments, the first proximal asset database is a maps database, such as Google Maps. The database may include an application programming interface (“API”) via whichqueries on the proximal asset database may be performed. Gather 100 may call, e g. the Google Maps Places API, to perform a geographic place search 106, which may be a circular radius search. Geographic place search 106 may be configured to to return data of the type=point of interest within the circular radius of the object. Category specific searches, e.g. via a findPlacesForAHPathways function, may also be performed.
[0066] Each search result may be converted to in a finding (112), which may include findingType = "asset", and latitude and longitude. The first proximal asset database may also return an asset type for each of the surrounding assets. Contextual categories may be inferred from the asset type, e.g. by the convertPlacesToFindings function, from a place type that may have been included with the search result. For example, an asset labeled as a park, may map to one or more categories, such as Natural Environment, Physical Activity, and Stress Reduction. Findings may then be written to storage 118, e.g. as a row a database such as database 14 of FIG. 1.
[0067] In Route B a second contextual database is queried, using the contextual input data, to return a summary environmental data structure relating to the contextual environment within the contextual distance. In some embodiments, the second contextual database may be or include the U.S. Census database, which may have an API for extraction of data. A gather-census function 102 queries the Census API for demographic data, e.g. at the tract level. Demographic data is then retrieved (108) via the census API. An artificial intelligence large language model (“LLM”) may then be used to create (11 ) one or more summary environmental findings data structures relating the summary environmental data to one or more of the taxonomy of contextual categories. Findings may be extracted via a “extractFindingsFromDocument” function, which uses the Al to map findings to valid contextual categories from the full taxonomy of contextual categories. Examples of categories that may be found in demographic census data in a real estate context may include financial wellbeing, industriousness and skill development, and preventive healthcare, the latter of which may be relevant for health and age demographic data. The one or more findings from the summary environmental data may then be stored (120) in the storage medium.
[0068] In Route C, a gather-context function 104 is executed for contextual findings. At least one environmental scope is selected (110), each environmental scope indicating a boundary for environmental context research. Environmental scopes define the boundaries for howproximate to the object the search results should be limited. In a real estate context, environmental scopes can be jurisdictional boundaries such as neighborhood, town, county, city, state, or can be measured by distance, e.g. within three miles. Environmental scopes may be selected by users or in other ways known in the art. An artificial intelligence answer engine is used for contextual research. In some embodiments, the artificial intelligence answer engine may be Perplexity, although other similar engines may be used.
[0069] For each pair of a selected environmental scope and an item on the taxonomy of contextual categories, a natural language query, for determining an extent to which assets matching the item are present within the selected environmental scope, is built. This may occur via a call to buildQuery(scope, pathway), which constructs a targeted natural-language query like "Chicago city -funded affordable housing programs, grants, and policies 2024".
[0070] First a context research cache is checked (116) first to avoid re-querying the artificial intelligence answer engine for the scopes and contextual categories that may already have been used. 3. If the cache hits, cache results are used (118). If the cache misses, each natural language query may be processed using an artificial intelligence answer engine, to create an answer engine output signal comprising the extent to which assets matching the item are present within the selected environmental scope. This may be accomplished via a call (124) to the answer engine, e g. a Perplexity sonar model. In some embodiments, five contextual categories may be used per scope in prompts to the answer engine. In some embodiments, each answer engine output may be processed (126) with the LLM, to create a contextual finding data structure relating to the extent to which assets matching the item are present within the selected environmental scope. This may be implemented via passing (126) Perplexity's raw response and citations to an LLM such as OpenAI gpt-4.1-mini, via distillFindings() , which may extract two to five structured findings, which may be in a JSON format. Findings may then be written (128) to the to the database and results may also be written to the cache.
[0071] When Route A, Route B, and Route C have independently completed, the project may be ready (130) for synthesis.
[0072] The retrieval architecture as disclosed herein, e.g. as disclosed in FIG. 3, has numerous technical advances over a standard approach. The retrieval architecture departs fromconventional vector-embedding-based retrieval in favor of a taxonomy-governed, two-stage approach.
[0073] In standard retrieval-augmented generation (“RAG”) systems, a user query is converted to an embedding vector and compared against a vector index of document embeddings using semantic similarity. This approach is computationally expensive, difficult to audit, prone to retrieving tangentially related content, and cannot enforce categorical boundaries between distinct domains.
[0074] The disclosed system replaces fuzzy semantic retrieval with a deterministic, category-routed architecture. In the first stage, a lightweight classification step receives the user's query and classifies it into one to four pathways from the fixed twenty-pathway taxonomy. This efficient, deterministic operation produces a discrete categorical output rather than a continuous similarity score. In the second stage, the system retrieves findings and pre-computed summaries using a relational join filtered to the classified pathways via the fmding_pathways join table. Only records tagged to those pathways are passed into the response generation context.
[0075] This departure from vector-based retrieval provides significant technical advantages. The industry-standard approach for Al applications operating over a document corpus is to generate vector embeddings and perform similarity search against a vector store. The disclosed system is capable of operating without a vector store and does not require embedding calls. The taxonomy -governed architecture functions to close the semantic gap between how data is tagged at ingestion and how it is accessed at retrieval. The shared constant definition of the taxonomy of contextual categories, used by the ingestion pipeline and the retrieval classifier, improves vocabulary alignment between storage and query, a property that vector-based systems, which derive their retrieval vocabulary implicitly from embedding geometry, cannot structurally provide.
[0076] The system also provides computational benefits. For example, routing and synthesis are separated across two models of different scale and cost, improving efficiency without sacrificing grounding quality.
[0077] FIG. 4 is a flow diagram of a synthesis pipeline in accordance with some embodiments. After gather completes, as discussed with reference to FIG. 3, synthesis can be triggered, either on a contextual category by contextual category basis, or all at once, or in some other grouping.
[0078] For Route A in FIG. 4, GENERATE- SUMMARIES 200 is called. In generating summaries, for each contextual category, the server first pulls all findings tagged to this pathway from the database via a finding pathways join. Findings may then be grouped them environmental scope, e.g. by geographic layer. In some embodiments, radius findings, then city findings, then state findings may be used in that order. Next a prompt may be constructed, which may include all findings, and at least one objective of the object. Next, summary generation 206 may occur, wherein the prompt may be transmitted to the LLM, generating a final structured output, the final structured output including a summary of the object and one or more listings of strengths, weaknesses, and opportunities relating to one or more items on the taxonomy of contextual categories. This may be implemented by a call to OpenAI gpt-4.1-mini , expecting back { summary, strengths, gaps, opportunities, key Assets, recommendedActions }, which may be received in a JSON formet, e.g., in accordance with the prompt. Next, the process stores (212) the final structured output, e.g. the summaries, in the storage medium. This may include upserting, the result into pathway _summaries . Upserting may be defined as inserting if the record does not exist, and / or updating it if it does.
[0079] After summaries are saved, two additional operations may be run inside the same route, each wrapped in its own non-fatal try / catch. generateCrossPathway Sy nthesis() may identify (218) findings shared across multiple pathways and may also synthesize executive-level strategic priorities. The result of step 218 may be written to project_synthesis. narrative, topAssets, whyltMatters, and strategicPriorities. Vision statement generation (224) may also occur at this point in the process.
[0080] Vision statement generation 224 and editing may, in some embodiments, be separated. Vision statement generation 224 takes stakeholder roles, developer goals, uploaded document summaries, and pathway narratives, prompts the LLM, e g. OpenAI, and writes an aspirational two to four paragraph vision into vision statements.
[0081] Vision statements may be written (226) to vision_statements. In some embodiments, cross pathway synthesis 218 and vision statement generation 224 are non-fatal. If either fails, the pathway summaries already saved are preserved. In some embodiments, a vision alignment check tool may be provided. A user may paste or upload text, e.g., a proposal, a grant application, or marketing brief. The system may then run the submitted text against the project's vision statement using an LLM. Alignment checks may be is stored in alignment_checks and may also be displayed in a Vision & Strategy hub view.
[0082] Routes B and C may be executed independent of one another, any time after Route A for any given object. Route B, “generate-lever-board,” 202 is an independent route which reads already-saved pathway summaries (208) and calls an LLM, e.g. OpenAI, to generate (214) strategic levers, which may include concrete actions at each geographic layer. In some embodiments, only the leverBoard field on project_synthesis may be written (220) as part of Route B, preserving other fields.
[0083] For Route C, POST / API / PROJECTS / :ID / GENERATE-CROSS-LAYER (204) also reads (210) already-saved pathway _summaries and calls (216) detectCrossLayerConflicts(), which looks for contradictions between scopes. Contradictions may include, for example, city zoning that conflicts with a state mandate. In embodiments, only the crossLayerConflicts field on project_synthesis may be written as part of Route C, preserving other fields. Can run in any order relative to generate-lever-board as long as summaries exist. When Routes A, B and C of FIG. 4 have run, synthesis data may be made available (228) in a user interface.
[0084] FIG. 5 is a flow diagram of resident intake and personalized report generation in accordance with some embodiments. In some embodiments, resident routes are public but ratelimited. A shareToken function 300 may be invoked. ShareTokenn 300 fetches (302) project data (findings, contextual categories, vision) by share token. In some embodiments, an internal project UUID is not exposed to the client. A resident sessions row may then be created, wherein a resident may submit a household profile 304. A household profile may contain household inputs. Household inputs may include the following fields. In some embodiments, composition and priorities are required; all other fields are optional and populated as the resident progresses through the intake flow:
[0085] interface ResidentHouseholdlnputs {
[0086] composition. string[];
[0087] priorities: string[];
[0088] members?: HouseholdMember[];
[0089] pets?: HouseholdPet[];
[0090] householdsize?: string;
[0091] ages?: string[];
[0092] ageRanges?: string[];
[0093] workSetup?: string;
[0094] transportpreferences?: string[];
[0095] weekendActivities?: string[];
[0096] lifestyleNotes? : string;
[0097] pathwayRatings?: Record<string, number>;
[0098] pathwayComments?: Record<string, string>;
[0099] selectedStatements?: string[];
[0100] reviewChatTranscript?: { role: string; content: string }[];
[0101] additionalContext? : string;
[0102] enrichmentCompleted?: boolean;
[0103] householdDeepDive?: string;
[0104] }
[0105] A call to an LLM, e.g. a GPT call, that matches the household profile against the project's contextual categories data, may then be called. The HTTP response returns (306) with a sessionToken. The frontend polls (318) with refetchinterval until status = "complete" . The resident sees their personalized fit report 310, which contextual categories match their household, what they will gain, what to be aware of, etc. In some embodiments, the Al prompt, that calls the LLM, includes both household data and the project's pathway summaries, so the generated output is specifically grounded in what this atlas actually offers. The result may then be written to resident_sessions.generatedOutput and rendered as the resident's personalized fit report, which may include the following:
[0106] interface ResidentOutput {
[0107] personalizedlntro: string;
[0108] bestFitAspects: { title: string; description: string; evidence: string }[];
[0109] tradeoffs: { title: string; description: string }[];
[0110] neighborhoodHighlights: {
[0111] name: string; distance: string; relevance: string;
[0112] placeld?: string; lat?: number; Ing?: number; photoUrl?: string;
[0113] }[];
[0114] community Vision: string;
[0115] suggestedQuestions: string[];
[0116] pathway Summaries?: { pathway: string; personalNote: string; keyPlaces: string[] }
[0117] researchcitations?: { url: string; title: string; source?: string; pathway: strin
[0118] }
[0119] In some embodiments, the resident report generator produces a personalized, evidence-bound output through a five-stage pipeline that combines deterministic rule-based processing with a single structured Al generation call.
[0120] Stage 1: Pathway relevance mapping. Household attributes are mapped to a relevant subset of the twenty pathways through deterministic rule-based functions. These functions apply demographic heuristics and keyword matching against the resident's notes to identify which pathways are relevant for this household. For example, a household including members aged zero to ten with an education priority will map to the Education and Health Literacy pathway. This filtering step is performed without Al inference, ensuring the subsequent generation call operates on a curated, rule-derived pathway scope rather than an unbounded context.
[0121] This deterministic mapping replaced an earlier approach in which residents were asked to score each of the twenty pathways on a numeric one-to-five scale, the straightforward approach for capturing stated preferences across a fixed taxonomy. That approach was abandoned because residents either assigned identical mid-range ratings to all pathways, producing undifferentiated signals, or assigned ratings that correlated poorly with what they expressed in free text. The current approach derives pathway signals from demographic attributes and the resident's own words, producing signals grounded in what the resident actually said rather than how they responded to an abstract rating instrument.
[0122] Stage 2: Finding prioritization and boosting. Keywords are extracted from the resident's intake notes and chat transcript, for example, "diabetes," "dog," or "remote work", and findings whose title or detail fields match those keywords are boosted to the top of the findings list. Approximately the top one hundred twenty findings are then retained for the generation context. This is the primary mechanism by which the evidence base is made household-specific rather than generic: the same corpus of findings is re-ranked differently for each household based on the resident's own words.
[0123] Stage 3: Context assembly. A single large prompt is constructed containing: only the pathway summaries for pathways mapped as relevant in Stage 1; the boosted, top-ranked findings with names and verified distances; the full demographic context; the resident's intake; and the project's vision statement, community synthesis, and preferred partner links.
[0124] Stage 4: Single structured generation call. One call is made to the generation model. The system prompt then instructs the model to apply the household's life-stage as a lens through which each pathway summary is reinterpreted, so the Transportation pathway summary reads as bench availability and sidewalk quality for a retiree, and as commute time and transit access for a remote worker, and to reference only real place names and distances from the supplied findings, never inventing. Personalization is therefore not achieved through a second Al pass over the summaries; it is achieved through filter, reframe, and selectively boosted evidence within a single generation call.
[0125] Stage 5: Evidence binding and post-processing. Every place name appearing in the generated output's neighborhood highlights and pathway key places is fuzzy-matched back against the original findings database. Only matched places receive a verified place identifier, coordinates, and confirmed distance attached to the output record. Unmatched names cannot receive coordinates and therefore cannot appear on the resident-facing map. This post-processing step creates a closed validation loop between generation and the verified data corpus: the structural mechanism ensures that only places the system can independently confirm appear as navigable, mapped results. This structural enforcement replaced an earlier reliance on prompt-based instructions alone, instructing the model not to hallucinate, which proved insufficient because it relied on model behavior rather than output validation.
[0126] The final output includes a personalized introduction, four to six best-fit aspects each with an evidence field citing a specific finding, two to four honest trade-offs, six to ten neighborhood highlights with verified coordinates and distances, a community vision narrative, four to six personalized suggested questions, and per-pathway summaries each carrying one to three key places.
[0127] Prompts for Al models, as generated by some embodiments of the presently disclosed system, are generated by a geospatial conversational prompt pipeline.
[0128] This pipeline is an improvement over two more conventional approaches. The first was a static, hand-authored chip list- the same set of conversational prompts displayed to every resident regardless of property location. This approach was not sufficient because geographically distant properties with entirely different community assets produced identical intake experiences,defeating the purpose of location-grounded personalization. The second considered approach was Al-generated chip phrases, which uses a language model to write each conversational prompt from the underlying place data. This approach can be inefficient, as per-chip LLM calls are expensive, introduce latency, and produce inconsistent surface language across chips within the same intake session. Deterministic templating, rotating a fixed set of phrase patterns over verified place names, is cheaper, faster, produces consistent language, and is auditable in a way that generatively produced prompts are not. Using deterministic templating rather than Al generation has measurable operational advantages.
[0129] Raw geospatial API responses are noisy, inconsistently structured, and not directly presentable in a conversational interface. The disclosed system processes these responses through a multi-stage deterministic pipeline. In the first stage, results are filtered for quality: only records carrying a verified third-party place identifier, a valid computed distance from the project address, and a place name passing the validation rules (length between three and thirty-five characters, no leading digits, no statistical or demographic terms such as "Median," "Poverty," or "Capital") are retained. In the second stage, retained results are tagged to their relevant Impact Pathways and ranked by proximity within each pathway category, with per-category caps applied. In the third stage, ranked results are converted into conversational phrases using a rotating template system, producing outputs such as "Love hanging out at Casa Garcia's" or "Feel safe near the Fire Department", rather than raw data labels. In the fourth stage, dynamically generated phrases are merged with a static fallback set to ensure that all user-interface topic areas retain coverage even when local data is sparse for a given pathway.
[0130] The validation rules and the verified place identifier requirement emerged from a security hardening process after earlier versions of the public story endpoint were found to expose unverified internal data, including census statistics and demographic findings, through the chip surface to anonymous users. The verified place identifier gate exists because the failure mode was discovered and corrected through iterative enforcement: it is an engineering response to a documented data leakage problem, not an anticipated design choice.
[0131] The result is an end-to-end deterministic pipeline from raw API data through structured normalization through ranked filtering through validated name cleaning throughtemplated conversational output, with quality and trust gates enforced at each stage before any content reaches the user. Ranking, filtering, name validation, and templating are deterministic operations, making the pipeline's outputs auditable and reproducible in a way that purely generative conversational content is not.
[0132] FIG. 6 is a structural diagram of an LLM 600 formed in a transformer architecture, utilized by systems and methods described herein, in accordance with some embodiments. Generative Al uses natural language processing (NLP) and machine learning to create natural language data or content. The LLM 600 includes a deep learning neural network configured to perform natural language processing (NLP) tasks, e.g., text generation, summarization, translation, text classification, and answering questions, thereby enabling Generative Al. Compared with a normal language model, the LLM 600 includes more than 100 million parameters and is pretrained with on large corpora of text. In some embodiments, the LLM 600 is implemented with a transformer architecture, and configured to shift through large datasets and recognize patterns and relationships between words or phrases. The transformer architecture includes anmechanism that weighs the importance of different words or phrases in a given context.
[0133] The transformer architecture of the LLM 600 includes an encoder network 602 and a decoder network 604. The encoder network 602 is configured to receive an input sequence 606 and generate a sequence of hidden states 608. Each hidden state 608 includes a vector that encodes contextual information of a word in the input sequence 606 based on their relative positions. The decoder network 604 is configured to receive portions 610P of a target sequence 610 successively and use an output 612 of the encoder network 602 to generate the target sequence 610. In some embodiments, the decoder network 604 starts with a starting token (e.g., “start”) and generates one prediction at a time. The decoder network 604 uses the output 612 produced by the encoder network 602 to understand the context of the input sequence 606. For each word of the target sequence 610 to be predicted, the decoder network 604 uses cross-attention mechanisms to focus on corresponding portions of the output 612 of the encoder network 602. As each word of the target sequence 610 is generated, the decoder network 604 updates its state and predicts a next word, until the entire target sequence 610 is generated.
[0134] In some embodiments, the LLM 600 applies a self-attention mechanism, and each position in a sequence (e.g., a natural language query) is attended to all positions in the same sequence. Self-attention helps the LLM 600 to understand and interpret the sequence by considering the entire sequence. For instance, when processing the natural language query, selfattention allows each word to be contextualized in relation to every other word in that natural language query. Alternatively, in some embodiments, the LLM 600 applies a transformer architecture including multihead attention 620 (also called multihead self-attention). Each attention head 622 learns a respective attention mechanism so that multihead attention 620 as a whole can learn more complex relationships. For example, referring to Figure 3, multihead attention 620 is applied in both the encoder network 602 and the decoder network 604 of the LLM 600 implemented in the transformer architecture.
[0135] It should be understood that various changes, substitutions, and alterations can be made herein without departing from the spirit and scope of the invention as defined by the appended claims. It should also be apparent that individual elements identified herein as belonging to a particular embodiment may be included in other embodiments of the invention. Moreover, the scope of the present application is not intended to be limited to the particular embodiments of the process, machine, manufacture, and composition of matter, means, methods and steps described in the specification. As one of ordinary skill in the art will readily appreciate from the disclosure herein, processes, machines, manufacture, composition of matter, means, methods, or steps, presently existing or later to be developed that perform substantially the same function or achieve substantially the same result as the corresponding embodiments described herein may be utilized according to the present invention.
Claims
ClaimsWhat is claimed is:
1. A system comprising:a networked computing environment including one or more servers configured to implement service-oriented architecture (“SOA”) components for producing an artificial intelligence based structured assessment of an object using contextual factors relating to the object, wherein the SOA components comprisea first SOA back-end component of the one or more servers configured to:receive an input comprising contextual input data indicating a contextual environment in which the object exists;query a first proximal asset database, using the contextual input data and a proximity distance to identify a plurality of proximal assets representing environmental aspects present within the proximity distance of the object, the first proximal asset database returning an asset type for each of the plurality of proximal assets;for each of the plurality of proximal assets, infer a contextual category, from a taxonomy of contextual categories, from the asset type; andstore each result from the first proximal asset database, and the inferred contextual category, as a proximal asset finding in a storage medium;a second SOA back-end component of the one or more servers configured to:query a second contextual database, using the contextual input data, to return a summary environmental data structure relating to the contextual environment within the proximity distance;use an artificial intelligence large language model (“LLM”) to create one or more summary environmental findings data structures relating the summary environmental data structure to one or more of the taxonomy of contextual categories, using a deterministic, category-routed architecture; andstore the one or more summary findings relating to the summary environmental data structures in the storage medium;a third SOA back-end component of the one or more servers configured to:select at least one environmental scope, each environmental scope indicating a boundary for environmental context research;for each pair of a selected environmental scope and an item on the taxonomy of contextual categories, build a natural language query for determining an extent to which assets matching the item are present within the selected environmental scope;process each natural language query using an artificial intelligence answer engine, to create an answer engine output signal comprising the extent to which assets matching the item are present within the selected environmental scope;process each answer engine output with the LLM, to create a contextual finding data structure relating to the extent to which assets matching the item are present within the selected environmental scope; andstore the one or more structured contextual finding data structures to the storage medium; anda fourth SOA back-end component of the one or more servers configured to, for each proximal asset finding, each summary finding, and each structured contextual finding:sort the findings by environmental scope;construct a prompt including all findings, and at least one objective of the object; transmit the prompt to the LLM, generating a final structured output, the final structured output including a summary of the object and one or more listings of strengths, weaknesses, and opportunities relating to one or more items on the taxonomy of contextual categories; andstore the final structured output in the storage medium.
2. The system of claim 1, the third SOA back-end component further configured to, prior to processing each natural language query using the artificial intelligence answer engine, process the natural language query on a context cache to determine if the natural language query had already been processed by the artificial intelligence answer engine; andstore the one or more structured contextual findings to the context cache.
3. The system of claim 1, the fourth SOA back-end component further configured to identify multi -category findings in the storage medium that apply to multiple items on the taxonomy of contextual categories, and synthesizing an executive-level strategic priority based on the multicategory findings.
4. The system of claim 1, the fourth SOA back-end component further configured to prompt the LLM to generate a vision statement for the object.
5. The system of claim 4, the fourth SOA back-end component further configured to provide an editing interface for the vision statement.
6. The system of claim 5, the fourth SOA back-end component further configured to evaluate a user-submitted text for alignment with the vision statement.
7. The system of claim 1, the fourth SOA back-end component further configured to detect contradictions between a first of the findings having a first environmental scope and a second of the findings having a second environmental scope.
8. The system of claim 1, further comprising a fifth SOA back-end component configured to determine compatibility between the object and a potential user of the object, the determining compatibility comprisingreceiving a user consumption input relating to potential consumption of the object by the potential user, the user consumption inputs comprising user consumption information and user priority information;using the LLM to generate user preference data relating to at least some of the taxonomy of contextual categories; andprompting the LLM to create a user fit report, wherein the prompting comprises pointing the LLM to the user preference data and the findings and structing the LLM to create a comparison report comparing the user consumption information, the user priority information, and at least one of the proximal asset findings, the summary finding, and the structured contextual findings.
9. The system of claim 1, the fourth SOA back-end component further configured to prompt the LLM to generate a structured set of actional strategies organized by contextual category and by scope.
