accessiBe Ltd v. AudioEye Inc

1. Case Identification

• Case #: IPR2022-00214
• Patent #: 10,845,946
• Filed: December 7, 2021
• Petitioner(s): accessiBe Ltd.
• Patent Owner(s): AudioEye, Inc.
• Challenged Claims: 1-30

2. Patent Overview

• Title: Modular Systems and Methods for Selectively Enabling Cloud-Based Assistive Technologies
• Brief Description: The ’946 patent describes systems for automatically remediating website accessibility issues. The methods involve accessing a website’s code (HTML or DOM), detecting compliance issues such as graphics lacking descriptive attributes, and using an artificial intelligence or image recognition server to generate and assign appropriate textual descriptions that can be read by assistive technologies.

3. Grounds for Unpatentability

Ground 1: Claims 1-28 and 30 are obvious over Springer, Hendry, and Folkens

• Prior Art Relied Upon: Springer (Application # 2004/0148568), Hendry (Application # 2013/0104029), and Folkens (Application # 2015/0286658).
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner argued that Springer taught a system of "checkers" and "fixers" to identify and remediate web accessibility issues, such as images lacking valid "alt" attributes, by modifying a website's Document Object Model (DOM). However, Springer's method for generating descriptive text was rudimentary. Folkens was alleged to cure this deficiency by disclosing an AI-based image processing system that automatically reviews images and generates descriptive tags identifying their content. Hendry was cited for its disclosure of a client-server architecture for document modification, where computationally intensive tasks could be performed on a remote server and remediation results stored for future use.
  • Motivation to Combine: A Person of Ordinary Skill in the Art (POSITA) would combine these references to improve the functionality of Springer's older accessibility system. Petitioner contended a POSITA would integrate the advanced AI-based image recognition from Folkens into Springer's framework to automate the generation of meaningful descriptive attributes, replacing manual or less effective methods. Furthermore, a POSITA would have been motivated to incorporate Hendry’s distributed computing architecture to offload processing to a remote server, enhancing performance and scalability.
  • Expectation of Success: Petitioner asserted that combining these known elements would have been straightforward. Integrating an AI-based image analysis API (Folkens) with a web remediation tool (Springer) using a standard client-server model (Hendry) represented a predictable implementation of established technologies.

Ground 2: Claims 1-28 and 30 are obvious over Lehota, Hendry, and Folkens

• Prior Art Relied Upon: Lehota (Application # 2010/0205523), Hendry (’029 application), and Folkens (’658 application).

• Core Argument for this Ground:

  • Prior Art Mapping: This ground presented Lehota as the primary reference, which disclosed a system for increasing web accessibility using a library of "accessibility widgets." These widgets could be called via JavaScript embedded in a webpage to parse a website's DOM and modify objects to fix compliance issues, such as adding missing ALT tags to images. As in the first ground, Folkens was cited for its AI-based image tagging capabilities, and Hendry was cited for its disclosure of remote processing and storage of accessibility modifications.
  • Motivation to Combine: The motivation was similar to Ground 1 but applied to Lehota's system. A POSITA would improve Lehota’s accessibility widgets by incorporating Folkens' AI to automatically and accurately determine the subject matter of graphics. This would enhance the widgets' ability to assign appropriate descriptive labels. A POSITA would also utilize Hendry’s teachings to implement Lehota’s system in a more robust client-server architecture, allowing for remote processing and personalization while reducing the computational load on the user's computer.
  • Expectation of Success: Petitioner argued a POSITA would have reasonably expected success in combining these references, as it involved applying a known AI solution (Folkens) and a known system architecture (Hendry) to improve a known accessibility widget system (Lehota).

• Additional Grounds: Petitioner asserted additional obviousness challenges for claim 29 (directed to video graphics) based on the combinations of Springer/Hendry/Folkens and Lehota/Hendry/Folkens, each further in view of Venugopalan (a 2015 IEEE article). Venugopalan was cited for teaching an AI model that generates text descriptions for videos by analyzing a sequence of frames, thereby providing the missing element for applying the core invalidity arguments to video content.

4. Key Claim Construction Positions

• Petitioner adopted the claim constructions from a November 23, 2021 order in the related district court litigation. Key adopted constructions included:
  • "alt text tag[s]": construed as "alternative text tag[s]".
  • "adequate descriptive attribute": construed as "descriptive attribute describing a nature of the graphic".
  • "mapping" and "artificial intelligence algorithm": construed according to their "plain and ordinary meaning".

5. Arguments Regarding Discretionary Denial

• Petitioner argued that discretionary denial under §314(a) based on the Fintiv factors would be inappropriate. The core arguments were:
  • The parallel district court litigation was in a very early stage, with fact discovery having only recently opened and no substantive arguments on invalidity presented to the court.
  • The court's tentative trial schedule was approximately 10 months away and highly uncertain, with a high probability of slipping, minimizing any potential overlap with the IPR timeline.
  • The petition raised grounds and arguments that were materially different from those in the district court, and the strength of Petitioner's case weighed heavily in favor of institution to correct an erroneously issued patent.

6. Relief Requested

• Petitioner requested the institution of an inter partes review and the cancellation of claims 1-30 of the ’946 patent as unpatentable under 35 U.S.C. §103.