DCT
6:22-cv-00423
Buffalo Patents LLC v. ZTE
Key Events
Complaint
I. Executive Summary and Procedural Information
- Parties & Counsel:
- Plaintiff: Buffalo Patents, LLC (Texas)
- Defendant: ZTE Corporation (China)
- Plaintiff’s Counsel: Antonelli, Harrington & Thompson LLP
- Case Identification: 6:22-cv-00423, W.D. Tex., 04/27/2022
- Venue Allegations: Plaintiff alleges venue is proper under 28 U.S.C. § 1391(c)(3), which permits a defendant not resident in the United States to be sued in any judicial district.
- Core Dispute: Plaintiff alleges that Defendant’s smartphones and other electronic devices infringe patents related to hybrid flexible display technology and to message recognition systems that use a shared language model for both speech and handwriting inputs.
- Technical Context: The technologies at issue are foundational to modern smartphones: flexible OLED displays that enable curved-edge screens and software that converts user speech and handwriting into text.
- Key Procedural History: The complaint alleges Defendant had pre-suit knowledge of the ’086 Patent via a notice letter dated November 26, 2021. Post-filing records from the USPTO, not mentioned in the complaint, indicate that the asserted claims of two patents-in-suit have been disclaimed following Inter Partes Review (IPR) proceedings. Specifically, Claim 7 of the ’405 Patent was disclaimed (IPR2023-01386), and Claim 13 of the ’737 Patent was disclaimed (IPR2023-01387). These events raise significant questions regarding the continued viability of the infringement counts for the ’405 and ’737 patents.
Case Timeline
| Date | Event |
|---|---|
| 1999-07-17 | ’405 & ’737 Patents - Earliest Priority Date |
| 2001-06-25 | ’086 Patent - Earliest Priority Date |
| 2005-02-15 | ’086 Patent - Issue Date |
| 2005-06-07 | ’405 Patent - Issue Date |
| 2012-06-19 | ’737 Patent - Issue Date |
| 2021-11-26 | Pre-suit notice letter sent to ZTE regarding ’086 Patent |
| 2022-04-27 | Complaint Filing Date |
II. Technology and Patent(s)-in-Suit Analysis
U.S. Patent No. 6,856,086 - "Hybrid Display Device"
- Issued: February 15, 2005
The Invention Explained
- Problem Addressed: The patent’s background section describes the difficulty of achieving precise alignment between two flexible polymeric substrates in display manufacturing, particularly when using roll-to-roll processes, which can compromise display quality (’086 Patent, col. 1:36-44).
- The Patented Solution: The invention proposes a “hybrid” manufacturing approach. A display device is constructed with one panel having a rigid substrate (e.g., glass) and the other having a flexible substrate (e.g., a plastic film). The flexible panel can be efficiently produced using a roll-to-roll process, while the rigid panel is fabricated separately and then precisely aligned and combined with the flexible panel using a "pick and place" operation, aiming to merge the benefits of both manufacturing techniques (’086 Patent, Abstract; col. 2:45-56).
- Technical Importance: This hybrid method sought to combine the manufacturing scalability of flexible film processing with the dimensional stability and high precision afforded by rigid substrates like glass (Compl. ¶11).
Key Claims at a Glance
- The complaint asserts independent Claim 1.
- Essential elements of Claim 1 include:
- A display device comprising a front panel and a back panel with a light control material therebetween,
- wherein one of the panels has a rigid substrate,
- and the other of the panels has a flexible substrate.
U.S. Patent No. 6,904,405 - "Message Recognition Using Shared Language Model"
- Issued: June 7, 2005
The Invention Explained
- Problem Addressed: The patent identifies challenges with contemporary message recognition systems, including the fact that user training (e.g., correcting errors) is typically siloed within a specific device and input modality. It also notes that conventional N-gram language models lack an understanding of global word meaning and context, limiting their accuracy (’405 Patent, col. 3:45-52, col. 4:1-12).
- The Patented Solution: The invention describes a system with multiple message recognizers (e.g., one for speech, one for handwriting) that both utilize a "shared language model." The key innovation is that when a user corrects an error generated from one input type (e.g., a misrecognized spoken word), the resulting training adapts the shared language model, thereby improving the accuracy of all connected recognizers, including the handwriting-to-text function (’405 Patent, Abstract; col. 7:7-21).
- Technical Importance: This unified training approach allows user corrections in one modality to improve performance in another, creating a more efficient and intelligent user adaptation system across different forms of input (Compl. ¶14).
Key Claims at a Glance
- The complaint asserts independent Claim 7.
- Essential elements of Claim 7 include:
- A method for performing message recognition with a shared language model,
- performing message recognition of a first type (e.g., speech) using the shared language model and a first type-specific model (e.g., acoustic),
- performing message recognition of a second type (e.g., handwriting) using the shared language model and a second type-specific model (e.g., handwriting),
- training the shared language model responsive to user correction of an error from either type, thereby improving the accuracy of both.
U.S. Patent No. 8,204,737 - "Message Recognition Using Shared Language Model"
- Issued: June 19, 2012
Technology Synopsis
As a continuation of the technology in the ’405 Patent, this patent describes a method for multi-modal input processing on a device. The method involves receiving distinct inputs like freehand stylus writing and audio from a microphone, selecting a user-specific message model from a plurality of models, and adjusting a local language model based on the selected user profile. This adjusted local model, which includes language, handwriting, and acoustic components, is then used to generate converted text data from either handwriting or speech inputs (’737 Patent, Abstract; Compl. ¶14).
Asserted Claims
The complaint asserts independent Claim 13 (Compl. ¶51).
Accused Features
The Gboard application on ZTE smartphones, which allows users to select personal or work profiles, receives both handwriting and voice input, and uses on-device personalization to adjust its language model based on user input and corrections to improve text generation (Compl. ¶54-65). The complaint provides a screenshot of a "Choose profile" setting on an Android device to support this allegation (Compl. p. 37).
III. The Accused Instrumentality
Product Identification
ZTE smartphones with curved or flexible OLED displays (e.g., ZTE Axon 30 Ultra) and/or that include the Gboard application or similar virtual keyboard technology (e.g., ZTE Blade A7 Prime) (Compl. ¶19, ¶32, ¶50).
Functionality and Market Context
- For the ’086 Patent: The accused functionality is the physical construction of the display. The complaint alleges these displays use a multi-layer structure that includes a rigid top layer of Corning Gorilla Glass 5 and a flexible AMOLED screen built upon a flexible substrate like polyimide. The light-emitting OLED material is situated between these layers (Compl. ¶23-25). The complaint includes a diagram illustrating the layered structure of an OLED display to support its technical description (Compl. p. 8).
- For the ’405 and ’737 Patents: The accused functionality is the Gboard software. Gboard provides users with multiple input modalities, including voice-to-text and handwriting-to-text. The complaint alleges it employs a shared n-gram language model for these functions. The software also features "on-device personalization," where it learns from user inputs, selections, and corrections to locally update and train its models, thereby improving future prediction accuracy for both speech and handwriting (Compl. ¶36, ¶43, ¶60-61).
IV. Analysis of Infringement Allegations
U.S. Patent No. 6,856,086 Infringement Allegations
| Claim Element (from Independent Claim 1) | Alleged Infringing Functionality | Complaint Citation | Patent Citation |
|---|---|---|---|
| a display device comprising a front panel and a back panel with a light control material therebetween | The ZTE Axon 30 Ultra smartphone is a display device with a top layer ("front panel"), a bottom layer ("back panel"), and an OLED material ("light control material") situated in between. | ¶22, ¶24 | col. 2:45-49 |
| wherein one of the panels has a rigid substrate | The top layer of the display includes Corning Gorilla Glass 5, which the complaint alleges is a "rigid substrate." | ¶23 | col. 2:45-49 |
| and the other of the panels has a flexible substrate | The bottom layer is a flexible plastic substrate, such as polyimide, which is a component of the "AMOLED flexible curved screen." | ¶25 | col. 2:45-49 |
Identified Points of Contention
- Scope Questions: A central question may be whether the term "panel", as used in the patent, reads on the highly integrated layers of a modern smartphone display. The patent’s specification describes a process of joining separately fabricated panels with a "pick and place" operation (’086 Patent, col. 2:53-56), which raises the question of whether integrated display stack layers, such as the cover glass and the underlying flexible screen assembly, constitute distinct "panels" in the sense contemplated by the patent.
- Technical Questions: The analysis may focus on the definitions of "rigid substrate" and "flexible substrate". While the complaint alleges Gorilla Glass is rigid and the AMOLED's base layer is flexible (Compl. ¶23, ¶25), a defendant could raise questions about the degree of rigidity or flexibility of these materials and whether they meet the claimed definitions.
U.S. Patent No. 6,904,405 Infringement Allegations
| Claim Element (from Independent Claim 7) | Alleged Infringing Functionality | Complaint Citation | Patent Citation |
|---|---|---|---|
| A method for performing message recognition with a shared language model... | The accused ZTE smartphones implement a method of message recognition using the Gboard application, which employs a shared n-gram language model. | ¶35, ¶36 | col. 7:7-9 |
| performing message recognition of a first type...to provide text data in accordance with both the shared language model and a first model specific to the first type... | Gboard performs voice-to-text conversion ("first type") using a speech recognizer that combines an acoustic model ("first model") with the shared n-gram language model. | ¶37, ¶38 | col. 7:10-14 |
| performing message recognition of a second type...to provide text data in accordance with both the shared language model and a second model specific to the second type... | Gboard performs handwriting-to-text conversion ("second type") using a handwriting model ("second model") in combination with the same shared n-gram language model. | ¶40, ¶41 | col. 7:14-17 |
| training the shared language model responsive to user correction of error...thereby improving accuracy of each of the first and second types... | Gboard uses "on-device personalization" to learn from user corrections and preferences, which trains and updates the language model locally, thereby improving the accuracy of both speech and handwriting recognition. | ¶42, ¶43, ¶45 | col. 7:17-21 |
Identified Points of Contention
- Scope Questions: Practitioners may focus on the scope of a "shared language model." The complaint provides screenshots showing Gboard offers both voice and handwriting input (Compl. pp. 15-16), but a key question will be whether the underlying software architecture truly uses a single, shared model as claimed, or if it uses distinct, separately tuned instances of a language model for each modality.
- Technical Questions: The infringement theory may turn on whether Gboard’s "on-device personalization" (Compl. ¶43) constitutes "training the shared language model" as required by the claim. A defendant may argue that personalization features, such as updating a personal dictionary, are technically distinct from the more fundamental adaptation or retraining of the core statistical parameters of the language model itself.
V. Key Claim Terms for Construction
The Term: "panel" (’086 Patent, Claim 1)
- Context and Importance: The claim requires a "front panel" and a "back panel." The definition is critical because the accused product is a highly integrated display stack, whereas the patent describes an assembly process that suggests more discrete components. Whether the layers of the accused display, such as the cover glass and the flexible screen assembly, legally constitute separate "panels" will be a central issue.
- Intrinsic Evidence for Interpretation:
- Evidence for a Broader Interpretation: The claim language itself does not specify that the panels must be fabricated separately or be physically distinct before assembly, potentially allowing "panel" to be construed as a major functional layer within the final display structure (’086 Patent, col. 22:30-34).
- Evidence for a Narrower Interpretation: The specification repeatedly describes a manufacturing method where the rigid panel is fabricated separately from the web of flexible panels and later combined via a "pick and place" operation, which suggests a narrower definition of "panel" as a discrete, pre-fabricated component (’086 Patent, Abstract; col. 2:53-56).
The Term: "training the shared language model" (’405 Patent, Claim 7)
- Context and Importance: This term is the central mechanism for the claimed invention's improvement in accuracy. The infringement case hinges on whether the accused Gboard's "on-device personalization" feature meets this definition. Practitioners may focus on this term because "training" can have a specific technical meaning in machine learning that may be narrower than general "personalization" or "learning."
- Intrinsic Evidence for Interpretation:
- Evidence for a Broader Interpretation: The patent states that the system "adapts language, acoustic, and/or handwriting models that reside in data storage" in response to user training, which could be interpreted to cover a range of adaptive behaviors including personalization that updates user-specific data used by the model (’405 Patent, col. 3:52-56).
- Evidence for a Narrower Interpretation: The specification discusses adapting a "syntactic model" in response to user correction, which may imply a more fundamental alteration of the model's core probabilistic or grammatical rules rather than simply augmenting it with user-specific preferences or dictionary entries (’405 Patent, col. 31:4-8).
VI. Other Allegations
- Indirect Infringement: The complaint alleges both induced and contributory infringement for the ’086 and ’737 patents. Inducement is based on allegations that ZTE provides instructions and advertising that encourage end-users to use the products in an infringing manner (Compl. ¶72). Contributory infringement is based on allegations that the accused products contain special features (e.g., the hybrid display structure, the multi-modal keyboard) that are not staple articles of commerce and have no substantial non-infringing uses (Compl. ¶87-88).
- Willful Infringement: The willfulness allegation is based on alleged pre-suit knowledge of the ’086 patent from a notice letter dated November 26, 2021, and knowledge of all patents-in-suit as of the complaint's filing. The complaint also alleges willful blindness, claiming ZTE has a "policy or practice of not reviewing the patents of others" (Compl. ¶91, ¶94).
VII. Analyst’s Conclusion: Key Questions for the Case
- A primary issue will be one of definitional scope: can the term "panel," described in the ’086 Patent’s specification in the context of discrete, separately fabricated components, be construed to cover the highly integrated layers of a modern smartphone’s flexible OLED display assembly?
- A key evidentiary question will concern functional operation: does the accused Gboard software's "on-device personalization" feature perform the specific, inventive step of "training the shared language model" as required by the ’405 Patent, or does it represent a technically distinct form of user adaptation?
- A threshold procedural question will be one of patent viability: given that the specific claims asserted against the ’405 and ’737 patents were disclaimed in IPR proceedings after the complaint was filed, a dispositive issue will be whether these infringement counts can proceed or are now moot.