Adaptive Classification Tech LLC v. Knovos LLC

I. Executive Summary and Procedural Information

• Parties & Counsel:
  • Plaintiff: Adaptive Classification Technologies LLC (Texas)
  • Defendant: Knovos, LLC (Florida)
  • Plaintiff's Counsel: Bradford Black P.C.
• Case Identification: 8:26-cv-02257, M.D. Fla., 08/04/2026
• Venue Allegations: Venue is alleged to be proper as Defendant is a Florida company with its principal place of business in Tampa, within the district, and maintains a regular and established place of business in the district where it has allegedly committed acts of infringement.
• Core Dispute: Plaintiff alleges that Defendant's eDiscovery software platform infringes a patent related to methods for conducting and terminating a technology-assisted document review (TAR) process.
• Technical Context: The technology concerns machine-learning systems used in legal e-discovery to efficiently find relevant documents within massive data collections, a process known as Technology-Assisted Review (TAR) or predictive coding.
• Key Procedural History: The complaint notes that the inventors' underlying research on Continuous Active Learning (CAL®) has been cited by courts approving predictive coding workflows. The prosecution history of the asserted patent is highlighted, with the complaint stating that the patent examiner found the claims novel over the prior art of record. Plaintiff also alleges sending a pre-suit notice letter to Defendant identifying the patent and the accused product.

Case Timeline

Date	Event
2015-06-19	'374 Patent - Earliest Priority Date
2019-10-15	'374 Patent - Issue Date
2026-03-12	Plaintiff sends notice letter to Defendant
2026-08-04	Complaint Filing Date

II. Technology and Patent(s)-in-Suit Analysis

U.S. Patent No. 10,445,374 - "Systems and Methods for Conducting and Terminating a Technology-Assisted Review"

• Patent Identification: U.S. Patent No. 10,445,374 ("Systems and Methods for Conducting and Terminating a Technology-Assisted Review"), issued October 15, 2019 (the "'374 Patent").

The Invention Explained

• Problem Addressed: The patent's background section addresses what the complaint terms "one of the most vexing problems" in TAR: reliably determining when to stop an iterative document review process Compl. ¶43 '374 Patent, col. 3:20-23 Conventional approaches made it difficult to know when a sufficient number of relevant documents had been found or if the process had converged on a suboptimal result, creating uncertainty about the completeness (or "recall") of the review effort '374 Patent, col. 3:55-61
• The Patented Solution: The invention provides a method to terminate a TAR process with a high degree of confidence Compl. ¶37 '374 Patent, abstract The process involves first identifying a "target set" of known relevant documents Compl. ¶38 '374 Patent, col. 6:63-67 Then, a separate, "independent" iterative search process (e.g., a Continuous Active Learning system) is executed on the entire document collection Compl. ¶40 '374 Patent, col. 7:20-29 The system then terminates the review based on a comparison between the results of this iterative search and the known target set, providing a way to confirm that a desired level of recall has been achieved with a certain statistical probability Compl. ¶42 '374 Patent, col. 18:40-45 The overall process is illustrated in a flowchart in the patent '374 Patent, FIG. 1
• Technical Importance: This method provides a structured, statistically defensible stopping point for a TAR process, moving beyond subjective reviewer judgment or less reliable sampling techniques to offer "quality assurance" for the review's completeness Compl. ¶43 '374 Patent, col. 6:43-45 '374 Patent, col. 6:57-60

Key Claims at a Glance

• The complaint asserts at least independent claim 1 Compl. ¶59
• The essential elements of Claim 1, a system claim, include instructions for a processor to:
  • Receive an identification of a "target set of documents" in a document collection, where these documents were identified as relevant by a "first search strategy".
  • Execute a classification process to train a classifier using the documents in the "target set".
  • Utilize a "second iterative search strategy" (which does not distinguish between documents in the target set and other documents) to classify documents in the collection.
  • "Terminate the classification process based upon a comparison between the results of the second search strategy and a characteristic of the target set", where the process achieves a target level of recall with a certain probability upon termination.
• The complaint does not explicitly reserve the right to assert dependent claims but alleges infringement of "one or more claims" Compl. ¶59

III. The Accused Instrumentality

Product Identification

• The "Knovos Discovery" eDiscovery platform, including its features for Technology Assisted Review ("TAR"), predictive coding, and Continuous Active Learning ("CAL") Compl. ¶¶51-52

Functionality and Market Context

• The complaint alleges that Knovos Discovery uses CAL to classify and prioritize documents for review Compl. ¶52 The workflow allegedly begins by using an initial set of documents, identified through search terms and key players, to "kick-start" an active learning engine Compl. ¶54
• As reviewers code documents, "the model learns, rescans, and ranks the most relevant files first" Compl. ¶56 The system is alleged to provide a "TAR_Check dashboard" that displays metrics including recall, precision, and F-measure, which helps users determine when to end the review process Compl. ¶56 The complaint includes a screenshot from Knovos's materials stating that training continues until a "selected target recall is achieved" Compl. ¶56

IV. Analysis of Infringement Allegations

Claim Chart Summary: '374 Patent Infringement Allegations

Claim Element (from Independent Claim 1)	Alleged Infringing Functionality	Complaint Citation	Patent Citation
receive an identification of a target set of documents in a document collection, wherein the target set of documents consists of documents identified as relevant as part of a first search strategy...	The Knovos Discovery platform uses documents identified via search terms and key players to "kick-start active learning at the outset of the attorney review." This initial set is alleged to be the claimed "target set."	¶54	col. 18:28-32
execute the classification process that enables training of a classifier using documents in the target set...	The platform's active learning engine is trained using the "kick-start" set of documents, which allegedly gives it "a solid understanding of what's relevant, based on the coding of relevant documents."	¶55	col. 18:33-34
...the classification process utilizes a second iterative search strategy which does not distinguish between documents in the target set and documents in the document collection...	The accused "TAR 2.0 with AI Tag Classification" system allegedly "learned from subject-matter experts' decisions and predicted tags across the broader set," applying its classifier to the entire collection without excluding documents based on their membership in the initial target set.	¶55	col. 18:35-39
terminate the classification process based upon a comparison between the results of the second search strategy and a characteristic of the target set...wherein the classification process achieves a target level of recall with a certain probability upon termination.	Knovos allegedly instructs users to compare TAR results with existing coding and provides a "TAR_Check dashboard" to evaluate recall. The process is alleged to terminate when a "selected target recall has been achieved." A screenshot of this dashboard is provided in the complaint.	¶56; ¶56, p. 19	col. 18:40-45

Identified Points of Contention

• Scope Questions: A central question may be whether the accused system's "kick-start" process (based on search terms) and its subsequent "Continuous Active Learning" process constitute a "first search strategy" and a "second iterative search strategy" that are sufficiently distinct as required by the claim. The defense may argue these are two phases of a single, integrated workflow.
• Technical Questions: The complaint alleges the accused product's "TAR_Check dashboard" and its termination condition based on achieving a "selected target recall" perform the claimed "comparison" Compl. ¶56 An issue for the court may be whether this general recall measurement is equivalent to the patent's more specific requirement of a "comparison between the results of the second search strategy and a characteristic of the target set" that ensures a "certain probability" of success. The complaint includes a screenshot of the Knovos "TAR_Check" dashboard, which displays metrics like "Review Percentage," "Recall Percentage," and "Richness," that will be central to this analysis Compl. ¶56, p. 19

V. Key Claim Terms for Construction

The Term: "first search strategy" and "second iterative search strategy"

Context and Importance

• The claim structure requires two distinct strategies. The plaintiff's infringement theory relies on mapping the defendant's initial document seeding to the "first strategy" and its CAL engine to the "second." The construction of these terms will be critical to determining if the accused system's architecture meets the claim's structural limitations.

Intrinsic Evidence for a Broader Interpretation

• The patent does not appear to strictly define "search strategy," and the specification states the "second" strategy can be a "TAR process" or a "CAL approach" '374 Patent, col. 7:26-29, which may suggest flexibility. A party could argue that any method for identifying an initial set of documents qualifies as the "first strategy."

Intrinsic Evidence for a Narrower Interpretation

• The specification describes the second search as "independent" of the first '374 Patent, col. 6:38-39 A party could argue that this requires a level of operational separation not present if the CAL engine (second strategy) is directly and continuously trained on the seed set (first strategy).

The Term: "comparison between the results of the second search strategy and a characteristic of the target set"

Context and Importance

• This term defines the core mechanism for the patented termination step. Whether the defendant's system performs this specific "comparison" will be a focal point of the infringement analysis.

Intrinsic Evidence for a Broader Interpretation

• The complaint presents a screenshot from a Knovos guide describing "Recall Measurement" as a process to estimate the proportion of relevant documents found by TAR Compl. ¶56, p. 18 A party may argue that any process that tracks how many documents from a known-relevant set have been found by the iterative process meets the plain meaning of "comparison."

Intrinsic Evidence for a Narrower Interpretation

• The patent describes terminating the classification when a "sufficient number of documents in the target set have been classified as relevant" by the second strategy '374 Patent, claim 5 A party could argue this requires a direct, document-level check against the target set, rather than a more generalized statistical recall estimate which Knovos's public-facing materials appear to describe Compl. ¶56

VI. Other Allegations

Indirect Infringement

• The complaint alleges induced infringement, asserting that Knovos provides "product webpages, guides, and white papers" that instruct customers on how to use the allegedly infringing TAR and CAL functionality Compl. ¶65 The allegation is that Knovos continued this conduct with specific intent to cause infringement after receiving the plaintiff's notice letter Compl. ¶65

Willful Infringement

• The willfulness allegation is based on Knovos's alleged continued infringement after receiving actual notice of the '374 Patent via a letter dated March 12, 2026 Compl. ¶60

VII. Analyst's Conclusion: Key Questions for the Case

• A core issue will be one of architectural mapping: Does the accused Knovos Discovery workflow-which uses a "kick-start" set to initiate a continuous learning process-embody the two distinct "first search strategy" and "second iterative search strategy" steps required by Claim 1, or does it operate as a single, unified process outside the claim's scope?
• A key evidentiary question will be one of functional equivalence: Does the accused system's "TAR_Check dashboard" and its process for stopping review based on a "selected target recall" perform the specific "comparison" between the iterative search results and the target set as claimed in the patent, or is there a material difference in the technical operation of the termination criteria?