US 10917483

Here's a concise summary of US Patent 10917483:

Title: Automated communication-based intelligence engine
Assignee: Numberai Inc.
Inventors: Daniel Hodges, Tasos Roumeliotis, Joel Grossman, Andrew Ruff, Steven Ginn, Andrew Smith
Filing Date: June 22, 2018
Issue Date: February 9, 2021
Abstract: The patent describes a system that creates and automatically maintains one or more entity models using communication-based information received over a network. This process involves machine learning, a user interface, a sentiment analyzer, a communication monitoring agent, and an automated bot. Bots are initialized from a bot template. The system analyzes new communication data to improve the entity model, keep the bot current, and update other services and products used by the entity. It also allows businesses to cluster other entities to identify competitors and provides analytical insights into user journeys through an entity model. The system automates the maintenance of listings like frequently asked questions within a networked or distributed environment.

CAFC 2026 Dockets: No specific dockets for patent US10917483 were found in the CAFC 2026 scheduled cases or general case information during the search.

Independent Claims Overview:
The full text of the patent, including the claims section, was not provided in the authoritative source. Therefore, a plain-language overview of each independent claim cannot be furnished at this time.

As of April 26, 2026, no specific litigation involving US patent 10917483 was found in the search results from patent litigation tracking sites. The search did not return any cases listing US10917483 as the asserted patent.

Proceedings overview

The USPTO ODP API returns no AIA trial proceedings for this patent as of the most recent ingest. No additional PTAB activity was found through web search. This indicates that US10917483 has not been challenged in an AIA trial.

Strategic summary

As of 2026-06-19, all claims of US10917483 remain untested by AIA trial proceedings. This means there are no canceled or sustained claims due to PTAB decisions. The estoppel provisions of 35 U.S.C. § 315(e)(2) do not apply to this patent, leaving all prior-art grounds available for a potential future petitioner.

The absence of PTAB activity suggests that the patent may not have been aggressively asserted or that potential challengers have not yet deemed it worthwhile to pursue an IPR, PGR, or CBM.

Recommended next steps

No PTAB activity exists for US10917483. A potential defendant facing assertion of this patent would have the full range of prior art and invalidity arguments available for a PTAB challenge, should they choose to pursue one.

Inventors

The named inventors for US Patent 10917483 are:

• Daniel Hodges (Employer at time of filing: Numberai Inc.)
• Tasos Roumeliotis (Employer at time of filing: Numberai Inc.)
• Joel Grossman (Employer at time of filing: Numberai Inc.)
• Andrew Ruff (Employer at time of filing: Numberai Inc.)
• Steven Ginn (Employer at time of filing: Numberai Inc.)
• Andrew Smith (Employer at time of filing: Numberai Inc.)

All inventors appear to be associated with the original assignee, Numberai Inc., at the time of filing. There are no immediate unusual patterns observed, as the reassignment event on 2020-12-28 confirms the assignment from the inventors to Numberai, Inc., which is consistent with typical employment agreements where inventors assign their rights to their employer.

Original assignee

The original assignee named on the issued patent US10917483 is Numberai Inc.

Based on the patent's abstract and detailed description, Numberai Inc. developed a system for an "Automated communication-based intelligence engine." The patent describes a system that creates and automatically maintains entity models using communication-based information, involving machine learning, a user interface, sentiment analysis, communication monitoring, and automated bots. This suggests Numberai Inc.'s primary line of business was in developing and deploying such intelligence engines and related software products.

As of the current date (2026-06-19), publicly available information indicates that Numberai Inc. is an operating company, often associated with conversational AI and automation for businesses.

Assignment timeline

No assignment records for US10917483 were found on the USPTO Patent Assignment Search portal beyond the initial grant to Numberai Inc. The Google Patents legal events section lists an assignment event:

• 2020-12-28 (executed) / recorded 2020-12-28 (based on Google Patents data for "Assigned to Numberai, Inc." from inventors)
  • Conveyance: Assignment (implied, from inventors to assignee)
  • Assignor: Joel Grossman, Tasos Roumeliotis, Andrew Ruff, Daniel Hodges, Andrew Smith, Steven Ginn (the inventors)
  • Assignee: Numberai, Inc.
  • Correspondent: Not specified in Google Patents event, requires USPTO assignment search confirmation.
  • Context: Transfer of invention rights from individual inventors to the corporate entity.

Since the USPTO Assignment Center search did not return any records for this patent (https://assignmentcenter.uspto.gov/patents/10917483), it can be stated that there are no recorded post-issuance assignments beyond the initial assignment from the inventors to Numberai Inc. at the time of grant. The event noted in Google Patents ("Assigned to Numberai, Inc. 2020-12-28") typically refers to the recordation of the assignment of inventorship rights to the company prior to or at the time of issuance.

Timeline diagram

timeline
    title Ownership of US 10917483
    2018 : Application filed
    2020 : Inventors assigned to Numberai Inc
    2021 : Patent issued to Numberai Inc
    2026 : Currently owned by Numberai Inc

NPE / troll-pattern signals

1. Shell-entity transfer — Not present. The patent remains with Numberai Inc., which appears to be an operating company developing products related to the patented technology. There is no record of transfer to a known shell entity or a company with "IP / Patents / Licensing" in its name.
2. Known asserter in the chain — Not present. Numberai Inc. is not identified as a known NPE/asserter by RPX or Unified Patents.
3. Repeat correspondent across the chain — Not present. There is only one implied assignment (from inventors to Numberai Inc.) and no subsequent recorded assignments to observe a pattern of recurring correspondents.
4. Cascading transfers — Not present. No multiple consecutive assignments have occurred.
5. Pre-litigation transfer — Not present. There is no record of litigation involving this patent, and no transfers have occurred to suggest a pre-litigation setup.
6. Bankruptcy fire-sale — Not present. Numberai Inc. is an active operating company, and there is no indication of bankruptcy proceedings or associated patent sales.
7. Privateering — Not present. There is no evidence of a transfer to an NPE on behalf of an operating company.
8. Defensive aggregator (anti-NPE) — Not present. The patent is currently held by Numberai Inc. and has not been transferred to a defensive aggregator like RPX or AST.

Verdict

Insufficient data.
There are no recorded assignments for US10917483 in the USPTO Assignment Center (https://assignmentcenter.uspto.gov/patents/10917483) beyond the initial implied assignment from the inventors to Numberai Inc. at the time of filing/grant. Therefore, there is insufficient data to identify any NPE/troll-pattern signals. The patent appears to be held by its original operating assignee, Numberai Inc.

To identify the most relevant prior art for US patent 109174483, I need to search the USPTO database for the patent and then examine its cited references.

The provided patent text itself does not include the full claims section. Therefore, while I can identify prior art, I cannot definitively state which specific claims are anticipated without the full text of the claims. I will provide the prior art information and note this limitation.

Here's the plan:

1. Search the USPTO database for US patent 10917483 to access its full details, including citations.
2. For each patent citation listed, gather the full citation, publication/filing date, and a brief description.
3. Discuss how each reference potentially anticipates claims under 35 U.S.C. § 102, noting the absence of the full claim text from the prompt. Under 35 U.S.C. § 102, a patent claim is anticipated if every element of the claim is found, either explicitly or inherently, in a single prior art reference.

Since I cannot directly search the live USPTO database in this environment, I will rely on the information from the provided patent text and a Google Patent search to identify cited prior art.

Based on the provided patent text, the "Prior art keywords" are: "entity, likelihood, conversion, communication, communication session." This suggests the general technical areas of the prior art.

The patent mentions "the prior art includes a rather primitive example of automating the following of a user's journey through an entities communication-based information path" and "E-commerce websites have made use of funnel analytics to better understand why and how potential customer make or do not make purchases." These are general descriptions of prior art concepts but do not provide specific patent citations.

To find the most relevant cited prior art, I'll leverage the Google Patents link provided in the original request, as it typically lists cited references.

I will proceed by making an assumption that the Google Patents page provides the necessary citation information.

Based on the initial Google Patents page for US10917483B2, I can see a "Prior art date" of 2017-06-22. However, this refers to the priority date of the application itself, not a specific prior art reference. To find specific prior art, I need to look for "Cited by" or "References Cited" sections, which are typically found further down the Google Patents page.

Since the provided text does not include a list of "References Cited" from the patent document itself, and I cannot perform a live deep search on the USPTO website to extract this information, I must state this limitation.

Limitation: The full list of "References Cited" (prior art documents) for US patent 10917483 is not directly available in the provided text. Therefore, I cannot provide a detailed analysis of specific prior art patents and their potential anticipation of claims under 35 U.S.C. § 102.

If I were able to access the "References Cited" section of US10917483 from the USPTO database or a comprehensive Google Patents view, I would proceed as follows:

For each cited prior art patent:

• Full Citation: Patent number, inventor(s), issue/publication date.
• Publication/Filing Date: The relevant date for 35 U.S.C. § 102 analysis (usually the earliest effective filing date or publication date).
• Brief Description: A summary of the invention disclosed in the prior art patent.
• Potential Anticipation (35 U.S.C. § 102): I would then compare the description of the prior art to the independent claims of US10917483. For each independent claim, I would analyze if every element is present in the prior art reference. Since the independent claims are not provided, this section cannot be completed.

General Interpretation of 35 U.S.C. § 102:
Under 35 U.S.C. § 102, an invention is not novel (and thus unpatentable) if it was known or used by others, patented, described in a printed publication, or in public use or on sale before the effective filing date of the claimed invention. The America Invents Act (AIA) modified this statute, creating two main categories of prior art under Sections 102(a)(1) and 102(a)(2). These sections broadly define what constitutes prior art, including disclosures made before the effective filing date of the claimed invention. Exceptions exist, for example, for disclosures made by the inventor within one year of the filing date or if the prior art and claimed invention were commonly owned.

Without the explicit list of prior art references from the patent document, I cannot fulfill the request for "most relevant prior art" with specific examples and anticipation analysis.

The obviousness analysis for US Patent 10917483, as of April 26, 2026, focuses on combinations of prior art elements that a person having ordinary skill in the art (PHOSITA) would have been motivated to combine to achieve the claimed invention. Since the full claims text is not provided, this analysis is based on the patent's abstract, summary, definitions, and detailed description, which outline the scope of the invention and its improvements over the described prior art. The "Prior Art" discussed here refers to the technologies and methods described in the Background section of the US10917483 patent as existing before the priority date.

PHOSITA Definition: A PHOSITA in 2017 (the patent's priority date) would be an individual with expertise in software development, machine learning, natural language processing, data aggregation, network communication protocols (including APIs for telecom, messaging, and social networks), and business intelligence systems focused on customer communication management and competitive analysis.

Motivation for Combination: The patent's Background section explicitly details several significant problems in the prior art that businesses faced [cite: "BACKGROUND"]. These problems include:

• The difficulty in keeping communication-based information (e.g., phone numbers, addresses, hours, FAQs) up-to-date and accurate, leading to customer loss [cite: "Modern entities face major problems with trying to keep their communication-based information up-to-date and accurate. Incorrect or outdated information can easily drive customers or users away while enabling competitors to become more effective."].
• The significant manual effort and human judgment required for creating and maintaining FAQs, which often grew stale quickly without any automated update mechanism or identification of information gaps [cite: "In the prior art FAQs took significant effort in identifying and answering even common questions.", "In addition, FAQs often grow stale almost as soon as they are published. ... In the prior art there was just no way to automate updating FAQs or to analyze entity communication-based information to identify information gaps in FAQs or even to automate the delivery of communication-based information based on an analysis of prior communications."].
• Inefficient, inaccurate, and non-automatable methods for identifying competitors, relying on manual categorization or self-identification [cite: "The once rather simple problem being faced by a multitude of modern businesses is identifying who its competitors actually are. In the prior art this was fairly easily handled by manually assigning businesses into certain categories... The foregoing no longer works effectively.", "Other prior art approaches to identifying competitors include allowing customers to identify such competitors, by asking businesses to self-identify, to manually search the internet, to track information from your own suppliers, and to obtain information from on-line services. Those prior art approaches are inefficient, can be manipulated, are costly, are prone to significant errors and cannot be automated."].
• The existence of only "primitive" automation for tracking user journeys, such as e-commerce funnel analytics, with limited scope [cite: "The prior art includes a rather primitive example of automating the following of a user's journey through an entities communication-based information path. E-commerce websites have made use of funnel analytics to better understand why and how potential customer make or do not make purchases."].
• Basic services from online organizations (e.g., Yelp, Facebook, Google) for businesses to update facts, but these were "incomplete and ha[d] little value beyond assisting data entry and form creation" [cite: "For example, Yelp, Facebook, and Google enable businesses to update their open hours and other business facts so that they are both prominently presented and accurate. While such is useful it is also incomplete and has little value beyond assisting data entry and form creation."].

These recognized deficiencies would have provided clear and compelling motivations for a PHOSITA to combine known technologies to develop a more sophisticated, automated, and intelligent solution for managing communication-based business intelligence.

Obviousness Arguments based on Combinations of Prior Art:

1. Automated Entity Model Creation and Dynamic Updates

• Prior Art Elements:

  • Primitive user journey tracking/funnel analytics in e-commerce: Known for analyzing customer paths [cite: "The prior art includes a rather primitive example of automating the following of a user's journey through an entities communication-based information path. E-commerce websites have made use of funnel analytics to better understand why and how potential customer make or do not make purchases."].
  • General machine learning (ML) and Natural Language Processing (NLP): Widely used by 2017 for data analysis, pattern recognition, sentiment analysis, and understanding human language in text or transcribed voice. The patent itself mentions using NLP in the user interface 42 [cite: "the system 10 is programmed to understand a customer's question such as “What time do you close on Tuesdays?” by using Natural Language Processing (NLP) techniques in the user interface 42."].
  • Automated bots/chatbots: Basic automated response systems were in use, particularly for customer service.
  • API integration with communication platforms: Standard practice for connecting applications to telecommunication carriers, messaging platforms, social networks, and online services to monitor and publish data [cite: "The entity intelligence engine 12 is configured to acquire data from a processor-enabled telecommunication carrier application program interface (“telecom API”) 14 , from a messaging platform API 16 , from a social network platform API 17 , from an online services host 19 , and from a device monitor 18 that handles communications to and from communication devices 20 ."].

• Rationale for Combination: Faced with the problem of manually maintaining outdated FAQs and other business information, and the general need for up-to-date communication-based information [cite: "Modern entities face major problems with trying to keep their communication-based information up-to-date and accurate."], a PHOSITA would be motivated to combine these existing technologies.

  • It would be obvious to leverage ML and NLP to analyze a broader range of "communication-based information" (e.g., texts, voice calls, social media messages) from various channels (via APIs) beyond just e-commerce funnel data, to automatically extract factual information about an entity [cite: "Communication-based information can include texts, voicemails, voice conversations, messages on messaging platforms such as Facebook MessengerTM and Apple's iMessageTM, emails, social media posts and broadcasts, and other social media messaging including photos and direct messages."].
  • Using this analysis to build and maintain an "entity model" encompassing facts like hours, services, and tone would be a logical application of information extraction and knowledge representation techniques.
  • Further, integrating automated bots to handle common queries and then feeding back accepted user responses or corrections into the entity model via ML (a "self-scoring process" or "automated feedback") would be an obvious approach for continuous improvement and to overcome the "stale FAQs" problem [cite: "The foregoing process may also include receiving a first plurality of queries via a network, suggesting a first plurality of responses to the first plurality of queries based on the entity model, accepting selections of the suggested first plurality of responses, and updating the entity model based on the accepted selections.", "the system 10 implements a self-scoring process based on predictions."]. This closed-loop learning mechanism is a fundamental aspect of many intelligent systems. Publishing these updates to various platforms via APIs would be a known method for disseminating information [cite: "That update is then applied to the particular platform (step 310 ).", "The system 10 also implements a publishing server in the telecom carrier interface 32 that publishes the updated fact to any connected services and products that are listening for updates."].

2. Automated Entity Categorization and Competitor Identification

• Prior Art Elements:

  • Manual competitor identification: As noted, prior methods were inefficient and prone to error [cite: "Other prior art approaches to identifying competitors include allowing customers to identify such competitors, by asking businesses to self-identify, to manually search the internet, to track information from your own suppliers, and to obtain information from on-line services. Those prior art approaches are inefficient, can be manipulated, are costly, are prone to significant errors and cannot be automated."].
  • Web crawling: Established technology for gathering publicly available data from the internet (websites, directories).
  • Data analytics and clustering algorithms: Common techniques for grouping data points based on similarities, applied in various fields including market research.

• Rationale for Combination: Given the "major problems" with identifying competitors [cite: "The once rather simple problem being faced by a multitude of modern businesses is identifying who its competitors actually are."], a PHOSITA would be motivated to automate and improve this process.

  • It would be obvious to combine web crawling (to gather published data and initial entity types) with the comprehensive monitoring of communication-based information (metadata like number, time, length, channel, and content via NLP/sentiment analysis) of secondary entities [cite: "analyzing the plurality of communications to determine at least one of a number of communications, time of communications, length of communications, channel of distribution of communications, or communications content of the plurality of secondary entities"].
  • Applying clustering algorithms to this rich, aggregated data would logically group similar entities, thereby identifying competitors more effectively than manual methods [cite: "The system 10 clusters secondary entities that have similar metadata patterns together. For example, businesses that field about the same volume of communications at relatively similar times in similar manners with similar types of customers are clustered together into a competitor cluster."].
  • Furthermore, known techniques for improving data quality and efficiency, such as identifying errors in clustering by cross-referencing published data and communications, and then performing targeted "re-crawls" for specific entities rather than entire networks, would be obvious optimizations to conserve resources and improve accuracy [cite: "The system 10 also identifies errors in existing secondary entity clusters.", "A re-crawl of the network is beneficially performed when such conflict is determined to properly cluster a particular entity. By re-crawling for a specific entity based on such conflict rather than re-crawling a large portion of a network at periodic or frequent intervals, system processing and network bandwidth resources are conserved."].

3. Intelligent Online Conversation Routing based on Conversion Likelihood

• Prior Art Elements:

  • E-commerce funnel analytics: Used to understand why customers make or don't make purchases [cite: "E-commerce websites have made use of funnel analytics to better understand why and how potential customer make or do not make purchases."].
  • Automated response systems/bots: Capable of handling initial customer interactions.
  • Predictive analytics/machine learning: Applied to forecast business outcomes, including sales or customer conversions.
  • Customer service routing systems: Existed to direct customer inquiries to appropriate agents.

• Rationale for Combination: To enhance the effectiveness of customer engagement and prioritize high-value interactions, a PHOSITA would be motivated to apply predictive intelligence to online conversations.

  • It would be obvious to extend the concept of "conversion" from e-commerce funnel analytics to real-time communication sessions (e.g., chat, messaging) [cite: "The system 10 is trained by using a sample of conversations. Ideally these are actual conversations between the primary entity and a secondary entity, such as a customer. Each conversation is tagged as being successful or not based on some definition of success. ... A success is considered a 'conversion'."].
  • Training a "likelihood of conversion model" using ML based on historical communication data and observed conversions would be a direct application of known predictive modeling techniques to a new data set [cite: "training a likelihood of conversion model based on the plurality of communications and on the plurality of conversions"].
  • Integrating this model with an automated response system (bot) to analyze ongoing communications and dynamically decide whether to continue bot interaction or "hand off" the session to a human agent based on the calculated likelihood of conversion would be an obvious optimization strategy for maximizing business outcomes and efficiently allocating human resources [cite: "analyzing the particular communications to determine a likelihood of conversion based on the conversion model, and handing off the particular communication session from the artificial response system based on the likelihood of conversion."]. This effectively translates the "funnel analytics" concept into a real-time, proactive communication management system.

4. Automated Bot Provisioning

• Prior Art Elements:

  • Basic chatbots and bot templates: Existing bots often had challenges with initial setup and provisioning of information [cite: "One challenge with using bots is their initial set up and provisioning with information."]. Bot templates were available to streamline deployment for common use cases.
  • Lookup services: Online services (e.g., Foursquare, Yelp, Google search, reverse phone number lookups) for finding public information associated with identifiers like phone numbers or addresses.
  • Web crawling and data extraction (including OCR for documents): Standard methods for gathering and structuring information from diverse online sources.

• Rationale for Combination: To address the "challenge with using bots" in terms of setting them up and provisioning them with knowledge [cite: "One challenge with using bots is their initial set up and provisioning with information."], a PHOSITA would be motivated to automate this process.

  • It would be obvious to use a "bot template" as a foundation and then automatically populate it with relevant business information by "seeding" with a basic identifier (like a phone number) and initiating network lookup services and web crawls [cite: "the bot template 45 preferably uses old fashion phone numbers as seed information when setting up and provisioning the bot 43 .", "the bot template 45 uses the entity intelligence engine 12 to actively seek out additional information for the entity model 21 and for the entity datastore 26 as well as for provisioning the bot 43 ."].
  • Extracting information from various sources, including processing documents like menus using optical character recognition (OCR) and natural language processing (NLP) to structure the data, and comparing information from multiple datastores to identify mistakes, are well-known data integration and validation techniques that a PHOSITA would readily apply to create a robust and accurate bot knowledge base [cite: "a lookup service may find a portable document format (“PDF”) file or photo of a menu on a website or through another source (such as YelpTM). That menu PDF can be processed by the user interface 42 to extract structured information about the primary entity. ... a restaurant menu can be processed using optical character recognition (OCR) and natural language processing (NLP) techniques at the user interface 42 to extract the menu's contents."].

In conclusion, the innovations described in US10917483, while presenting a comprehensive system, represent the application of well-known information technology, machine learning, and natural language processing techniques to address long-standing and explicitly identified problems in business intelligence and customer communication management. A PHOSITA, motivated by these recognized problems and equipped with knowledge of the available prior art tools, would have found it obvious to combine and integrate these elements in the ways described to achieve the functionalities of the claimed invention.