US 7177798

US 7,177,798 B2 — Verified Record

Note on scope: I searched for the exact number 7177798 and disregarded results for look-alike numbers (e.g., JP 7177798, the unrelated "7177798 CVH CALCADOS LTDA" registry entry, and the '717/'179/'879 patents that surfaced incidentally). All substantive content below traces to the '798 patent record itself or to docket/PTAB records naming patent 7,177,798.

---

1. Bibliographic data

Discrepancy flagged (not auto-corrected): Google Patents and at least one secondary source list a 2000-04-07 priority/prior-art date, which does not match either the April 17, 2000 CIP parent or the May 19, 2000 provisional. The patent's own reference list cites an International Search Report from PCT/US00/09265 "filed Apr. …" — plausibly the source of the 2000-04-07 date, but I cannot confirm that from the material available. Treat the 2000-04-07 date as reported-but-unverified.

---

2. Abstract (verbatim from the patent)

"A method for processing a natural language input provided by a user includes: providing a natural language query input to the user; performing, based on the input, a search of one or more language-based databases; providing, through a user interface, a result of the search to the user; identifying, for the one or more language-based databases, a finite number of database objects; and determining a plurality of combinations of the finite number of database objects. The one or more language-based databases include at least one metadata database including at least one of a group of information types including case information, keywords, information models, and database values."

---

3. Plain-language overview of the independent claims

The patent has 21 claims. The face of the record and the claim-construction record identify claim 1 and claim 9 as independent, with at least one further independent claim in the later-claim range (see caveat below).

Claim 1 — the interface/search method (claims 1–8 group)
A computer-implemented method for handling a user's natural-language question against a database. Steps: (a) accept a natural-language query input from the user; (b) based on the input, without augmentation, search one or more "language-based databases," at least one of which is a metadata database containing at least one of: case information, keywords, information models, and database values; (c) return the search result to the user through a user interface; (d) identify a finite number of database objects for those databases; and (e) determine a plurality of combinations of those finite database objects. In plain terms: instead of parsing English grammar, ground the query in the finite, machine-known set of database objects and metadata, and treat interpretation as searching/enumerating combinations. ("Without augmentation" was construed in later litigation as "based solely on the natural language input without augmentation.")

Claim 9 — the "permutations" method (claims 9–21 group), computer-implemented
(a) Receive a natural-language input; (b) derive from it a plurality of language-based database objects; (c) identify a finite number of permutations of those database objects, with the objects stored in a metadata database comprising at least one of case information, keywords, information models, and database values; and (d) interpret at least one of the permutations to determine a result of the natural-language input. This is the "the domain of interpretation is finite" insight from the specification: a given enterprise database yields only finitely many permutations, so a natural query can be reduced to one of them.

Third independent claim — the graph-search / query-translation method
The specification's third aspect, which the record describes as a claim, recites: determining from the natural-language input a plurality of recognized terms that exist in a data dictionary having a logical graph structure; determining a minimum number of those recognized terms and the associated vertices; determining at least one minimum-cost query graph containing a minimum amount of vertices; if more than one minimum-cost query graph exists, removing redundant graphs to produce a solution set; determining within the solution set at least one cost query graph that is a complete solution; and translating that graph into a query language statement (e.g., SQL). Dependent claims add the reference-dictionary layers, complete-path determination through the graphs, resolution of ambiguity via rules and heuristics, and adding new cases and keywords to the dictionary based on user input.

⚠ Uncertainty on claim numbering: I have high confidence that claims 1 and 9 are independent and that the above is their substance (claim 9's text is quoted in Amazon-complaint analysis and tracked in Apple's IPR petitions). I do not have authoritative confirmation of the exact number of the third independent claim — the record locates it within the 9–21 group and a claim-construction order references a "natural language query input" term in claim 20, which may indicate the third independent claim is claim 20 (or that there are further independents). Do not rely on that number without checking the printed claims.

---

4. Enforcement and PTAB history (all naming patent 7,177,798)

• IPR2014-00077 — Apple Inc. v. Rensselaer Polytechnic Institute, filed Oct. 21, 2013; institution denied (decision Apr. 15, 2014).
• IPR2014-00319 and IPR2014-00320 — Apple; not instituted (procedural).
• IPR2019-01068 and IPR2019-01069 — Unified Patents as petitioner; not instituted (merits).
• District court: N.D.N.Y. 1:12-cv-01579, 1:13-cv-00633 (Apple/Siri, filed June 3, 2013 by RPI and exclusive licensee Dynamic Advances, LLC), 1:18-cv-00549 (Rensselaer Polytechnic Institute v. Amazon.com, filed May 8, 2018, asserting at least Claim 1 and Claim 9 against Alexa); S.D.N.Y. 1:23-cv-00227.
• Federal Circuit: docket numbers 24-1725 and 24-1739 appear in the Google Patents litigation list.
• A secondary (AI-generated, French-language) PatSnap litigation summary states that the Rensselaer Polytechnic Institute v. Amazon.com appeal was filed April 23, 2024, closed February 24, 2026, with the judgment AFFIRMED in RPI's favor.

⚠ Verification caveat on the 2026 CAFC item: I could not independently confirm the CAFC outcome from a primary source (the Federal Circuit's own docket/opinion) in this search. The only source describing a February 24, 2026 disposition is a machine-generated blog digest, and its narration is internally inconsistent (it says the appeal was "filed… in the District of Columbia circuit," whereas the underlying case is N.D.N.Y.). Treat "affirmed, Feb. 24, 2026" as unverified and check the CAFC docket for 24-1725 / 24-1739. In addition, because the patent's adjusted expiration was 2021-11-16, any 2024–2026 appeal can only concern pre-expiration damages or past liability, not ongoing infringement.

Also of note from the Apple-era dispute: Apple settled the Siri litigation in April 2016 for a reported $24.9 million to Marathon Patent Group (Dynamic Advances' parent), with 50% reportedly payable to RPI.

---

5. Key claim-construction points (N.D.N.Y., RPI v. Amazon.com)

A magistrate judge recommended, and the court was asked to adopt, constructions including:

• "natural language query input" = "input that is articulated in a human language in a way that a native speaker could understand and use sensibly";
• "based on the input, without augmentation" = "based solely on the natural language input without augmentation";
• "language-based database" = a "collection of stored data that has a searchable organizational scheme and includes information about structuring, using, or interpreting data."

I have not confirmed whether these recommended constructions were adopted in final form.

---

6. Bottom line

US 7,177,798 is an expired (2021-11-16) Rensselaer Polytechnic Institute patent (Hsu & Boonjing, filed May 21, 2001, issued Feb. 13, 2007) claiming a natural-language database interface that interprets queries by enumerating and ranking finite combinations/permutations of database objects against a four-layer "reference dictionary" (cases, keywords, information models, database values), using a graph model and branch-and-bound search rather than linguistic parsing. It was asserted against Apple (Siri, 2012–2016, settled for ~$24.9M) and Amazon (Alexa, from 2018), and survived every IPR petition filed against it (none instituted). A 2024-filed Federal Circuit appeal (24-1725 / 24-1739) is reported by a secondary source to have been affirmed on February 24, 2026 — that specific 2026 docket outcome should be verified against the CAFC's own records before being relied upon.

I'll search for litigation involving US patent 7177798 across patent litigation databases.

Let me gather more details on the specific cases and outcomes.

Let me get details on the remaining cases including the 2023 action and PTAB proceedings.

Litigation Involving U.S. Patent No. 7,177,798

Patent: U.S. 7,177,798 B2 — "Natural language interface using constrained intermediate dictionary of results" (Application No. 09/861,860; priority Apr. 7, 2000; filed May 21, 2001; issued Feb. 13, 2007). Inventors: Cheng Hsu and Veera Boonjing; original assignee Rensselaer Polytechnic Institute ("RPI").

Ownership note relevant to standing: RPI is the patentee and later exclusively licensed the patent to Dynamic Advances, LLC (a subsidiary of Marathon Patent Group; later styled CF Dynamic Advances LLC). RPI and Dynamic Advances appear as co-plaintiffs in the enforcement actions. Google Patents record also shows a license/security-interest chain (Dynamic Advances, LLC; DBD Credit Funding LLC). The U.S. patent's legal status is listed as expired (adjusted expiration Nov. 16, 2021), though litigation over past damages continued after that date.

Below are the known matters I could confirm. Where sources conflict, I flag it.

---

1. District Court Litigation

A. RPI / Dynamic Advances v. Apple Inc. (Siri)

• Plaintiff(s): Dynamic Advances, LLC (exclusive licensee), later joined by Rensselaer Polytechnic Institute
• Defendant: Apple Inc.
• Jurisdiction: U.S. District Court for the Northern District of New York (Albany Division)
• Case number(s): 1:12-cv-01579 (N.D.N.Y.) and 1:13-cv-00633 (N.D.N.Y.; Judge David E. Peebles). Both numbers appear in the records for this controversy; the 2012 action was the initial Dynamic Advances complaint, and the RPI/Dynamic Advances action (1:13-cv-00633) is the captioned case that proceeded to settlement. Treat these as related filings in the same dispute rather than two independent lawsuits.
• Filing date: 2012 (complaint); RPI joined in 2013 (June 3, 2013)
• Accused product: Apple's Siri voice assistant
• Outcome/status: Settled — On April 18–19, 2016, the parties entered a settlement and stipulated dismissal (with prejudice) filed April 19, 2016. Apple agreed to pay $24.9 million ($5 million on dismissal; $19.9 million on satisfaction of conditions), and received a patent license plus a 3-year covenant not to sue. RPI did not initially consent to the proposed royalty split (Dynamic Advances expected to pass ~50% to RPI, counsel, and a predecessor licensee), and that dispute was flagged for potential arbitration.
• Sources: Marathon Patent Group SEC Form 8-K (Apr. 18, 2016); Stipulation and Order of Dismissal, 1:13-cv-633 (N.D.N.Y. Apr. 19, 2016); MacRumors; The Verge.

B. RPI and CF Dynamic Advances LLC v. [Amazon.com, Inc.](/litigations/by-plaintiff/Amazon.com%2C%20Inc.) (Alexa)

• Plaintiff(s): Rensselaer Polytechnic Institute and CF Dynamic Advances LLC
• Defendant: Amazon.com, Inc.
• Jurisdiction: U.S. District Court for the Northern District of New York
• Case number: 1:18-cv-00549 (BKS/TWD) — Judge Brenda K. Sannes
• Filing date: May 8, 2018
• Accused products: Amazon Alexa Voice Software / Alexa-enabled devices (Echo, Echo Dot, Echo Plus, Echo Show, Echo Spot, Echo Look, Amazon Tap, Echo Connect, Fire TV accessories, Fire Tablets, Kindle)
• Outcome/status: Amazon won; patent held invalid. On March 18, 2024, the court granted Amazon's motion for summary judgment (Dkt. 384), denied RPI's motion for summary judgment (Dkt. 390), and dismissed the complaint with prejudice, holding the '798 claims ineligible under 35 U.S.C. § 101 (Alice). No damages awarded.
• Appeal: Rensselaer appealed (Fed. Cir. No. 24-1725, filed April 23, 2024). On Feb. 24, 2026, the Federal Circuit affirmed in a nonprecedential opinion (Judge Dyk writing), holding the claims abstract at Alice step one and lacking an inventive concept at step two, and awarding costs to Amazon.
• Sources: Judgment, Dkt. 444 (N.D.N.Y. Mar. 18, 2024); Fed. Cir. opinion (Feb. 24, 2026) via CCH; Law360; IPWatchdog; VitalLaw; Dinsmore.

C. RPI et al. v. Amazon.com, Inc. — 2023 action

• Plaintiff(s): Rensselaer Polytechnic Institute et al.
• Defendant: Amazon.com, Inc.
• Case number: 1:23-cv-00227
• Filing date: 2023
• Jurisdiction: Reported inconsistently — the Google Patents litigation record lists this as New York Southern District Court, while the Stanford NPE Litigation Database lists it as N.D.N.Y. (Stanford NPE lists two entries under this number for "Rensselaer Polytechnic Institute et al v. Amazon.com, Inc.").
• Outcome/status: Not confirmed from the sources retrieved. This number appears in the litigation databases attached to the '798 patent, but I could not retrieve a docket or disposition for it. It may be a companion/second action tied to the parallel Federal Circuit appeal No. 24-1739 (which appears alongside 24-1725 in the litigation record), but I cannot confirm that linkage.

---

2. PTAB / IPR Proceedings (all directed to the '798 patent)

All five are keyed to Respondent Application No. 09/861,860 / Patent No. 7,177,798.
Sources: Google Patents litigation record for US 7,177,798; GreyB/IPVerse (IPR2014-00319; IPR2019-01068; IPR2019-01069).

---

3. Federal Circuit Appeals

---

4. Related matters and caveats

• Foreign litigation: The Google Patents record links a world-wide patent family to a Darts-ip litigation entry (family 24173121), indicating at least one non-U.S. family member has been litigated. I could not confirm any specific foreign suits from the sources retrieved.
• Discrepancies I am flagging rather than resolving:
  • The Apple case appears under two N.D.N.Y. numbers (1:12-cv-01579 and 1:13-cv-00633); I have not verified whether these were consolidated or sequential.
  • The 1:23-cv-00227 venue is reported as S.D.N.Y. by one source and N.D.N.Y. by another, and I could not confirm its parties/outcome beyond the database labels.
  • The petitioner for IPR2014-00077 is not confirmed in the retrieved records (likely Apple, given the parallel 2014 IPRs).

Bottom line: The '798 patent was asserted in at least two waves — against Apple (settled in 2016 for $24.9M) and against Amazon (final judgment of invalidity under § 101 in 2024, affirmed by the Federal Circuit on Feb. 24, 2026). It also generated five IPR petitions (all institution-denied) by Apple and Amazon.

If you need the 1:23-cv-00227, IPR2014-00077, and CAFC No. 24-1739 specifics verified, those three are the open gaps — I recommend pulling the PACER dockets and the Federal Circuit docket directly, since my searches hit their retrieval limit before confirming them.

I'll research the PTAB proceedings surfaced in the patent's litigation metadata and cross-check against public sources.

PTAB Proceedings — U.S. Patent 7,177,798 ("Natural language interface using constrained intermediate dictionary of results")

Note on sourcing / discrepancy flag. The structured "PTAB proceedings on file" block states that the USPTO ODP API returns no AIA trial proceedings for this patent. That is not the full picture. The patent's own structured bibliographic data (Google Patents litigation field, fetched 2026-09-30) lists five PTAB case numbers, and each one is independently corroborated by public Board documents and secondary coverage surfaced via web search. I am reporting those five below and flagging the ODP gap explicitly. I did not invent any proceeding number — IPR2014-00077, IPR2014-00319, IPR2014-00320, IPR2019-01068, and IPR2019-01069 all appear in the patent's litigation metadata and in Board filings.

---

Proceedings overview

Five AIA trial proceedings are on file; all five ended in denial of institution (0 instituted, 0 final written decisions, 0 claims canceled at the PTAB, 0 settled). Three were denied on the merits — IPR2014-00077, IPR2019-01068, IPR2019-01069, each recorded in the structured data as "Not Instituted - Merits" — and two were terminated on procedural grounds — IPR2014-00319 and IPR2014-00320, recorded as "Not Instituted - Procedural." No petition ever reached trial, so no claim of the '798 patent was canceled or confirmed by the Board. On the PTAB axis alone the patent looks "hardened." But that is the wrong axis for a defendant today: the patent expired 2021-11-16, and on 2026-02-24 the Federal Circuit affirmed summary judgment that the claims are invalid under 35 U.S.C. § 101 as directed to an abstract idea (Rensselaer Polytechnic Institute v. [Amazon.com, Inc.](/litigations/by-plaintiff/Amazon.com%2C%20Inc.), Nos. 2024-1725, 2024-1739). The bottom line is therefore not "the patent survived two IPRs and is hardened" and not "claims 1–5 were canceled" — it is "every claim is judicially invalid and the patent term has run; any demand letter citing the '798 patent is unsupportable."

---

Proceedings (most impactful first)

IPR2019-01068 / IPR2019-01069 — Amazon.com, Inc. v. Rensselaer Polytechnic Institute

(consolidated in a single institution decision; listed as two case numbers in the structured data)

• Type: Inter Partes Review (two coordinated petitions covering all 21 claims)
• Filed: 2019-05-07 (both)
• Status: "Not Instituted - Merits" — denial of institution on the merits under 35 U.S.C. § 314(a)
• Judge panel: Josiah C. Cocks, Trevor M. Jefferson, Miriam L. Quinn, APJs; opinion authored by Jefferson, APJ
• Petition grounds: IPR2019-01068 challenged claims 1–8 (independent claim 1 plus dependents 2–8); IPR2019-01069 challenged claims 9–21 (independent claim 9 plus dependents 10–21) — i.e., the two petitions together covered the entire patent. Obviousness-based § 103 grounds. Prior art of record includes Weber (U.S. Patent 6,532,444) and Meng, plus exhibits including U.S. Patent 5,966,126, Zhang, and the Lieberman materials, supported by an expert declaration from Dr. Jaime Carbonell. Petition counsel: Knobbe Martens (Colin B. Heideman et al.); Patent Owner counsel: Cyrus Morton.
• Institution decision: Denied 2019-11-14. The Board adopted the construction "case information" = "information about prior instances of use of the natural language processing method" (carried over from IPR2014-00077 and the parallel district court claim construction), then held the petitioner had not persuasively shown that the prior art (e.g., Weber's clarifying prompts) discloses that limitation. Because the alleged absence of "case information" was the only basis for the merits challenge, no ground had a reasonable likelihood of prevailing. (Institution decision, IPR2019-01068/-01069)
• Final Written Decision: None issued — the case never instituted, so nothing was adjudicated claim-by-claim.
• Settlement / termination: Not settled. Terminated by denial of institution.
• Rehearing: Amazon filed requests for rehearing on 2019-12-16 under 37 C.F.R. § 42.71(d), arguing the Board "misapprehended the meaning of 'case information' by excluding clarifications from users." The Board disagreed and denied rehearing, holding it had not narrowed the construction and that the denial rested on the sufficiency of Amazon's evidence as to Weber/Meng. (Rehearing decision, IPR2019-01068; Amazon's rehearing request, IPR2019-01069) — I could not confirm the exact entry date of the rehearing denial; the decision is public in the 2020 timeframe.
• Appeal: None from the PTAB denial (a denial of institution is unreviewable under § 314(d) / Thryv). Amazon instead won on § 101 in the parallel district court and at the Federal Circuit (below).
• Defensive value: Low value as an IPR roadmap, high value as a cautionary tale. These petitions were the closest the patent ever came to trial, and they failed on a single claim-construction term. A defendant cannot assume this art set was weak — it was never tested at trial, and because no FWD issued, no § 315(e)(2) estoppel attached to Amazon. Practically, though, this ground is now moot: the claims are invalid under § 101 and the patent has expired.

---

IPR2014-00077 — Apple Inc. v. Rensselaer Polytechnic Institute and Dynamic Advances, LLC

• Type: Inter Partes Review
• Filed: 2013-10-21 (corrected petition)
• Status: "Not Instituted - Merits" — Decision Denying Institution of Inter Partes Review, 37 C.F.R. § 42.108
• Judge panel: Josiah C. Cocks, Bryan F. Moore, and Miriam L. Quinn, APJs; opinion authored by Moore, APJ
• Petition grounds: Challenged claims 1–21 (independent claims 1 and 9). Five primary references: Shwartz, Janas, Bouchou, Dar, and Warthen, with a supporting declaration from Dr. James F. Allen. The grounds were § 102 anticipation and/or § 103 obviousness. Patent Owner filed a preliminary response on 2014-01-30.
• Institution decision: Denied 2014-04-15. The Board analyzed the merits under § 314(a) and found no reasonable likelihood of prevailing on any challenged claim. Critically for the patent's later history, the Board construed "case information" to mean "information about prior instances of use of the natural language processing method," and expressly stated that "instances of use" of the system did not mean "indications cover[ing] instances of use of the system, e.g. how the system interprets words or phrases." That construction is what defeated every subsequent petitioner. (Decision Denying Institution, IPR2014-00077)
• Final Written Decision: None issued.
• Settlement / termination: Not settled; terminated by denial.
• Appeal: No appeal of the denial (unreviewable institution decision).
• Defensive value: This is the decision that built the patent's moat. Any IPR petitioner on this patent in 2014–2019 had to attack through the "case information" construction, and no one could. Because there was never a FWD, the "moat" was only ever a merits/threshold moat — it evaporated the moment a district court took a § 101 pass at the claims.

---

IPR2014-00319 and IPR2014-00320 — Apple Inc. v. Rensselaer Polytechnic Institute (and Dynamic Advances, LLC)

• Type: Inter Partes Review (two follow-on petitions filed concurrently with a § 315(c) joinder motion)
• Filed: 2014-01-03 (both)
• Status: "Not Instituted - Procedural" (both) — denial of institution, with repeated rehearing attempts failing
• Judge panel: Not confirmed from the sources reviewed; the more granular panel composition was not public in the records I could retrieve.
• Petition grounds: Same patent, same claims (1–21), overlapping art, and the same expert (Dr. Allen) as IPR2014-00077. Apple itself described the two new petitions as addressing the same grounds: IPR2014-00319 addressed Shwartz and Janas; IPR2014-00320 addressed Bouchou, Dar, and Warthen. Apple moved under § 315(c) / 37 C.F.R. § 42.122(b) to join both new proceedings with IPR2014-00077 and with each other. (Petitioner's Motion for Joinder, IPR2014-00319)
• Institution decision: Not instituted — procedural. The dominant issue was the § 315(b) one-year bar. Apple was served with the original Dynamic Advances, LLC v. Apple, Inc. complaint (No. 1:12-cv-01579, "Dynamic I") on 2012-10-23; Dynamic I was dismissed without prejudice on 2013-07-22 as part of a Rule 42 coordination with RPI & Dynamic Advances v. Apple, No. 1:13-cv-00633 (N.D.N.Y.). Apple argued the Rule 41(a)-style voluntary dismissal left the parties "as though the action had never been brought" (per Macauto), restarting the clock; RPI argued the Rule 42 consolidation meant the clock kept running. The Board agreed with the Patent Owner, distinguishing dismissals without prejudice under Rule 41(a) from dismissals incident to Rule 42 consolidation, which the Board treated as analogous to a conditional dismissal. The joinder motion also failed because there was no instituted IPR to join. (Contemporaneous practitioner analysis of the denials and rehearing history)
• Post-denial activity: Apple sought rehearing and then a second request for rehearing before an expanded panel in IPR2014-00320. In a decision dated 2014-09-11, the Board held the request was unauthorized and ordered it expunged from the record. Rehearing of the IPR2014-00319 institution decision was denied (decision dated 2014-08-07), and a second request was likewise denied. (Faegre Drinker PTAB docket listing; Patent Lawyer Blog, "PTAB to Apple: no third or fourth bite at the apple," 2014-09-16)
• Final Written Decision: None issued in either case.
• Settlement / termination: Not settled; terminated by denial (IPR2014-00320's second rehearing request expunged 2014-09-11).
• Appeal: None.
• Defensive value: These two files show the Board's procedural tolerance limit on serial petitions against this patent — Apple filed three petitions and racked up an expunged filing for its trouble. For a defendant in 2026 this is historical context only; the substance is dead under § 101.

---

Strategic summary

Claim status across all forums. At the PTAB, no claim was ever canceled, and no claim was ever adjudicated on the merits past the institution threshold — all 21 claims (independent claims 1 and 9 plus dependents 2–8 and 10–21) simply never made it into trial. In the district court, the picture inverted: RPI and CF Dynamic Advances asserted the patent against Amazon in N.D.N.Y. (No. 1:18-cv-00549, filed 2018-05-08, and again in No. 1:23-cv-00227, filed 2023-02-21), and the district court granted Amazon summary judgment that the claims are invalid under § 101. The Federal Circuit affirmed on 2026-02-24 in a nonprecedential opinion, holding that "[g]eneric use of AI without other parameters … is abstract," that "[t]he use of case information in the claimed database simply involves the use of AI as applied to a new environment and cannot render the claims non-abstract," and that "[a] conventional application of case-based reasoning, even to a novel environment, is abstract." Claim 1 was treated as the representative claim. (Justia opinion, Nos. 24-1725/24-1739; full-text PDF) The patent also expired 2021-11-16 (adjusted expiration; status "Expired - Lifetime"). I am not aware of any post-2026 Federal Circuit or en banc activity, and I did not find one.

Estoppel landscape — the important nuance. Because no IPR ever reached a final written decision under § 318(a), § 315(e)(2) estoppel never attached to Apple or Amazon on any ground they raised or reasonably could have raised. Apple, for instance, is not estopped as to Shwartz, Janas, Bouchou, Dar, or Warthen; Amazon is not estopped as to Weber, Meng, or the related art. That is unusual — a patent that survived only institution denials leaves prior-art defenses fully intact for the original challengers. But the point is academic: the validity fight was resolved on a ground that is not available in an IPR at all (§ 101), the patent has expired, and no PTAB-based defense is needed.

Pattern signals. (1) Repeat petitioners: Apple filed three petitions (IPR2014-00077, -00319, -00320) within ~10 weeks in late 2013/early 2014; Amazon filed two coordinated petitions (IPR2019-01068, -01069) in 2019 covering all 21 claims. That is a classic two-wave attack pattern — a 2013 wave during the Apple/Siri litigation and a 2019 wave during the first Amazon litigation. (2) Patent owner behavior: RPI/CF Dynamic Advances litigated aggressively, won all five institution fights, and took the Apple case to the eve of trial before settling; it then pursued the Amazon case to the Federal Circuit and lost on appeal. (3) Defensive aggregators: There is no defensive aggregator in the petitioner chain on this patent — the petitioners of record are Apple Inc. and Amazon.com, Inc. "Unified Patents" appears only as the third-party litigation-data source cited in the Google Patents litigation field (and the Unified Patents portal page for this patent), not as a petitioner. I want to be explicit about that so the record is not misread. (4) The Google Patents record also shows a PTAB case IPR2019-01068 entry with a blank "Petitioner:" field; the actual petitioner in the Board's own decision is Amazon.com, Inc.

---

Recommended next steps

1. If you are a defendant holding a demand letter on the '798 patent, the demand is unsupportable on its face. Quote the disposition: the Federal Circuit affirmed summary judgment that the claims are directed to ineligible subject matter under § 101. Key language: "A conventional application of case-based reasoning, even to a novel environment, is abstract. Therefore, the application of case-based reasoning to natural language processing does not provide an inventive concept sufficient to render the claims patent-eligible." Full opinion: Justia (Nos. 2024-1725, 2024-1739, decided 2026-02-24) / CCH full text PDF. Add the expiry: the patent expired 2021-11-16, so there is no forward-looking royalty period.
2. Do not build a defense around PTAB outcomes. No FWD exists to cite, and a denial of institution carries no estoppel and no preclusive effect. If a counterparty asserts that "the PTAB upheld these claims," that is inaccurate — the Board merely declined to institute, which leaves "the patent owner in possession of unaltered patent claims and the petitioner free to challenge those claims" (see the § 314(d) framing in the Apple v. Maxell line, Fed. Cir. blog copy).
3. If you need the claim-construction record (the "case information" construction carried from IPR2014-00077 into the 2014 N.D.N.Y. claim construction and then into the 2019 Amazon IPRs), pull: IPR2014-00077 institution decision, 2014-04-15, IPR2019-01068/-01069 institution decision, 2019-11-14, and the 2019 rehearing decision. Note that the 2014 construction is now largely academic given the § 101 affirmance.
4. Verify on PTAB E2E before filing anything. All five proceedings are decades-old, terminated, and no longer on any active trial schedule, so there are no institution deadline / oral hearing / FWD-due-date milestones to track. The ODP block in this prompt reports no AIA proceedings, which conflicts with the five case numbers above — I would reconcile that discrepancy directly against PTAB E2E (or the PTAB's public "Patent Trial and Appeal Board End to End" search) and, if the ODP record is incomplete, note it, because it is exactly the kind of gap that can surface in a diligence memo.
5. One correction worth flagging to whoever prepared the underlying data: the prompt's default position — "no PTAB activity on file" — is wrong for this patent. There were five petitions. The correct statement is: five AIA petitions were filed; all five were denied institution; none proceeded to a final written decision; no claim was canceled by the PTAB; the asserted claims were instead invalidated under § 101 in district court and affirmed on appeal.

I'll reconstruct this from the authoritative patent text plus live records, and I'll verify the assignment chain rather than assume it.

Key finding emerging: RPI appears to have remained the owner while Dynamic Advances was an exclusive licensee. Let me verify the recorded documents themselves.

Now let me pin down the recorded reel/frame entries and the asserter-directory listings.

Ownership Chain Reconstruction — US 7,177,798 B2

Patent: "Natural Language Interface Using Constrained Intermediate Dictionary of Results" · App. 09/861,860 · Filed 2001‑05‑21 · Issued 2007‑02‑13 · Expired – Lifetime (adjusted expiration 2021‑11‑16) · Assignee of record on the face of the patent: Rensselaer Polytechnic Institute.

Sourcing note (important): Google Patents' legal‑events tab and the court record expose the events and parties, but do not publish reel/frame numbers or the correspondent of record. I could not retrieve the individual Assignment Center record pages (reel/frame + correspondent fields) in this session, so those fields are reported only where a primary document in the file history supplies them, and are otherwise marked not retrieved. I have not invented reel numbers. The one reel/frame in this chain that is documented in a primary source is the inventor→RPI assignment (Reel 012195/0213).

---

Inventors

Priority chain: CIP of Ser. No. 09/544,676 (filed 2000‑04‑17, abandoned) plus provisional 60/205,725 (filed 2000‑05‑19). Note a literal‑reading discrepancy: Google Patents lists the priority date as 2000‑04‑07, while the specification states the parent was filed Apr. 17, 2000. I flag it rather than reconcile it.

Unusual‑pattern check — no departure anomaly. Both inventors assigned to RPI as a matter of university policy (the specification itself notes RPI's institutional ownership practice), and the first‑named inventor stayed at RPI for the entire 20‑year life of the patent. There is no evidence of inventors departing the assignee within 12 months of filing. The title did not move off RPI until the 2012–2013 licensing event — roughly 12 years post‑filing — which is a licensing decision, not an inventor‑departure fire sale.

---

Original assignee

Rensselaer Polytechnic Institute (RPI), Troy, New York.

• Entity type / line of business: private research university (founded 1824). RPI does not ship a product embodying the claims; it is a non‑operating patent owner by definition. The claimed subject matter (a metadata‑database natural language interface) was never commercialized by RPI.
• Current status: operating; and critically, RPI is still the patentee. The N.D.N.Y. held on 2023‑09‑15 (1:18‑cv‑00549, Dkt. 423) that "RPI did not transfer all substantial rights to CF Dynamic and remains the patentee of the '798 Patent."
• Patent status: expired 2021 (portfolio schedule in Marathon's SEC filing lists expiry 11/06/21; Google Patents records an adjusted expiration of 2021‑11‑16; 578 days of patent term adjustment). Claims were additionally held patent‑ineligible under §101 (see verdict section).

---

Assignment timeline

Recorded events as reflected in Google Patents legal events / Assignment Center event data, cross‑checked against the patent's file history exhibits and the litigation record:

1. 2001‑09‑24 (executed) / recorded — Reel 012195 / Frame 0213

• Conveyance: Assignment of assignors' interest
• Assignor: Cheng Hsu and Veera Boonjing (individually)
• Assignee: Rensselaer Polytechnic Institute
• Correspondent: Not published by Google Patents for this record. Flag: the reel/frame pair is confirmed only via the patent's file history — a USPTO assignment cover sheet bearing filer Tarek N. Fahmi, Customer No. 23628, receipt date 25‑APR‑2012 — which lists this patent among four RPI patents (App. 09/861,860 → Reel 012195/0213) with "Inventors → RENSSELAER POLYTECHNIC INSTITUTE."
• Context: Routine university employment/institutional assignment, executed four months after filing.

2. 2013‑05‑10 — Reel/frame not retrieved

• Conveyance: License (recorded under the Assignment Center's license category; described on Google Patents as "Assigned to DYNAMIC ADVANCES, LLC — LICENSE")
• Assignor: Rensselaer Polytechnic Institute
• Assignee / recorded party: Dynamic Advances, LLC (Texas; direct parent DA Acquisition LLC, itself a 100%‑owned subsidiary of Marathon Patent Group, Inc. per Dynamic Advances' First Amended Rule 7.1 Corporate Disclosure Statement, 1:13‑cv‑00633, July 18, 2014)
• Correspondent: Not retrieved.
• Context: Transfer‑to‑asserter. This is the Amended and Restated Exclusive License Agreement between RPI and Dynamic Advances (Marathon's monetization vehicle); the courts treated the amended ELA as conveying exclusionary rights sufficient to sue. Do not record this as an ownership change — RPI retained title.

3. 2015‑02‑02 — Reel/frame not retrieved

• Conveyance: Security Interest
• Assignors: Dynamic Advances, LLC and Marathon Patent Group, Inc.
• Assignee: DBD Credit Funding LLC
• Correspondent: Not retrieved.
• Context: Securitization / collateral. The RPI license was pledged as collateral for a loan to the Marathon group; this is the first link to DBD Credit Funding, which later becomes the destination of the license itself.

4. 2017‑01‑11 — Reel/frame not retrieved

• Conveyance: Security Interest
• Assignors: the Marathon/CF IP‑holding family — Marathon IP GmbH, TLI Communications GmbH, Marathon Ventures S.à.r.l., MedTech Development Deutschland GmbH, Motheye Technologies, LLC, Munitech IP S.à.r.l., Nyanza Properties, Orthophoenix, LLC, Synchronicity IP GmbH, Synchronicity IP LLC, Traverse Technologies Corp., Vermilion Participations, 3D Nanocolor Corp., Bismarck IP Inc., Magnus IP GmbH
• Assignee: DBD Credit Funding LLC, as collateral agent
• Correspondent: Not retrieved.
• Context: Portfolio‑wide collateral package across the entire Marathon affiliate group — the signature of a group‑level restructuring/securitization, not a patent‑specific transaction.

5. 2017‑09‑29 (executed; not shown as a recorded assignment) — n/a

• Conveyance: Consent + transfer of the Amended RPI Exclusive License (assignment of the license, not of the patent)
• Assignor: Dynamic Advances, LLC (with RPI's written consent)
• Assignee: CF Dynamic Advances LLC — "a designee of DBD"; corporate parent CF Marathon, LLC (per CF Dynamic Advances' Rule 7.1 statement filed 2018‑05‑08 in 1:18‑cv‑00549)
• Correspondent: n/a
• Context: Distressed/securitized transfer. Executed by RPI's consent letter of 2017‑09‑29 (N.D.N.Y. 1:18‑cv‑00549, Dkt. 92‑4; R&R Dkt. 103). I could not confirm that this transfer was separately recorded at the Assignment Center — treat it as documented in litigation, not verified as a recorded assignment.

No further recorded assignments. Nothing after 2017. The patent expired in Nov 2021 with RPI still holding legal title and CF Dynamic Advances LLC (DBD‑designated) holding the exclusive license.

Litigation overlay (for signal analysis, not part of the assignment record): 2012‑10‑19 Dynamic Advances, LLC v. Apple Inc., 1:12‑cv‑01579 (N.D.N.Y.) — first '798 suit, dismissed on standing questions · 2013‑06‑03 RPI & Dynamic Advances v. Apple, 1:13‑cv‑00633 · IPR2014‑00077 / ‑00319 / ‑00320 (Apple, not instituted) · 2016‑04‑18 Apple settlement $24.9M (Marathon 8‑K, 2016‑04‑18) · 2018‑05‑08 RPI & CF Dynamic Advances v. Amazon, 1:18‑cv‑00549 (Alexa) · 2018‑10‑10 CF Dynamic Advances LLC v. TechDev Holdings (N.Y. Sup.; license‑ownership fight with Spangenberg‑linked entities) · IPR2019‑01068 / ‑01069 (Unified Patents, not instituted) · 2024‑03‑18 §101 summary judgment for Amazon · Fed. Cir. 24‑1725 / 24‑1739 consolidated, non‑precedential opinion affirming (docketed March 2026).

---

Timeline diagram

timeline
    title Ownership and assertion of US 7177798
    2001 : Filed by Hsu and Boonjing
         : Assigned to Rensselaer Polytechnic Institute
    2007 : Patent issued
    2012 : Dynamic Advances sues Apple
    2013 : Exclusive license recorded to Dynamic Advances
    2015 : Security interest to DBD Credit Funding
    2016 : Apple pays 24.9 million to settle
    2017 : License shifted to CF Dynamic Advances
    2018 : RPI and CF Dynamic sue Amazon
    2021 : Patent expires
    2024 : Claims held ineligible
    2026 : Federal Circuit affirms ineligibility

---

NPE / troll-pattern signals

1. Shell-entity transfer — PRESENT.
The exclusionary rights were moved from a non‑operating university (RPI) into a licensing‑only Texas LLC, recorded 2013‑05‑10 (Google Patents legal event, conveyance "License," assignee Dynamic Advances, LLC). Concrete support beyond the name: (a) Dynamic Advances' own Rule 7.1 Corporate Disclosure Statement (July 18, 2014, 1:13‑cv‑00633 Dkt. 110) states its parent is DA Acquisition LLC, 100% owned by Marathon Patent Group, Inc., a publicly traded patent acquisition/monetization company; (b) it had no products — its first act was to file an infringement suit; (c) the license later passed to a second single‑purpose entity, CF Dynamic Advances LLC, whose disclosed parent is CF Marathon, LLC (Rule 7.1 statement, 2018‑05‑08).

2. Known asserter in the chain — PRESENT.

• Marathon Patent Group, Inc. (OTCQB/NASDAQ: MARA), a self‑described "patent acquisition and monetization company," is the disclosed indirect parent of Dynamic Advances, LLC (Rule 7.1 statement, July 18, 2014) and is the entity that received the Apple settlement proceeds (Marathon 8‑K filed 2016‑04‑18: Apple to pay $24.9M to Dynamic Advances, a wholly owned subsidiary).
• Erich Spangenberg / IPNav link: in CF Dynamic Advances LLC v. TechDev Holdings (N.Y. Sup. Ct., 2019 NY Slip Op 32068(U)), the plaintiffs alleged the defendant entities "are controlled by … Spangenberg," and Spangenberg's affidavit states his communications with RPI about the patent were made "through IP Nav as the retained advisor" and that Dynamic Advances LLC was held by TechDev Holdings / Acclaim Financial Group (majority‑owned since 2006 by his former spouse). Whichever ownership account is credited, both point to NPE‑side principals controlling the licensee.
• The same collateral list includes TLI Communications GmbH, a known high‑frequency NPE filer.
• Caveat: ownership of Dynamic Advances, LLC is contested — Marathon said it owned it; Spangenberg's affidavit says TechDev/Acclaim did. I report both rather than pick one.

3. Repeat correspondent across the chain — UNCLEAR (insufficient data).
I have exactly one correspondent datum in this chain: the assignment cover sheet filed by Tarek N. Fahmi, Customer No. 23628, receipt date 25‑APR‑2012, listing this patent's inventor→RPI assignment (Reel 012195/0213) among a four‑patent RPI batch. That is a single appearance ~6 months before the first Apple suit, and I cannot establish that he acted for the NPE side rather than for RPI — one appearance is not the recurrence the signal requires. The Assignment Center correspondent fields for the 2013, 2015 and 2017 recordings were not retrievable in this session. Separately, do not conflate litigation counsel with assignment correspondents: Skiermont Puckett LLP (Dallas) filed the Apple actions for Dynamic Advances; Heslin Rothenberg Farley & Mesiti (Albany) and Robins Kaplan/MoloLamken appeared for RPI later.

4. Cascading transfers — PRESENT.
Four sequential rights‑transfer events in roughly 54 months, three of them inside a 24‑month window, all revolving around DBD Credit Funding LLC: license recorded 2013‑05‑10 → security interest 2015‑02‑02 (Dynamic Advances + Marathon → DBD Credit Funding) → group security interest 2017‑01‑11 (Marathon/CF IP‑holding family → DBD Credit Funding as collateral agent) → license assignment to CF Dynamic Advances LLC, DBD's designee, with RPI's consent dated 2017‑09‑29. The shared counterparty (DBD) and the single‑purpose nature of the transferees supply the "common principals" element.

5. Pre-litigation transfer — PRESENT.
The exclusive license was recorded 2013‑05‑10, yet Dynamic Advances, LLC filed suit alone against Apple on 2012‑10‑19 — i.e., the assertion preceded the recordation (standing was then cured by joining RPI in the 2013‑06‑03 complaint). For the Amazon action, the license transfer to CF Dynamic Advances was consented on 2017‑09‑29 and suit filed 2018‑05‑08 — roughly seven months, just outside a strict 6‑month window but clearly sequenced to establish standing before filing. The April 2012 assignment‑recordation submission is a further pre‑suit chain‑of‑title cleanup.

6. Bankruptcy fire‑sale — NOT PRESENT (distressed securitization instead).
I found no Chapter 7/11 filing by RPI, Marathon Patent Group, Dynamic Advances, or DBD Credit Funding. The mechanism here was a secured loan/collateral arrangement (the 2015‑02‑02 and 2017‑01‑11 "Security Interest" recordings) followed by a transfer of the license to the lender's designee in 2017, disputed in state court in 2018–19. That is distressed monetization, not a bankruptcy §363 sale. Marathon's public trajectory (later renamed Marathon Digital Holdings and pivoted to cryptocurrency) is consistent with an exit from patent assertion, but I have no filing to cite for a bankruptcy event.

7. Privateering — NOT PRESENT in the strict sense.
Privateering requires an operating company transferring patents to an NPE to assert against its competitors. The transferor here is a university, not an operating company. Note the closest analogue: this was a revenue‑sharing monetization arrangement in which the licensee was contractually obligated to pay roughly 50% of gross proceeds back to RPI (Marathon 8‑K, 2016‑04‑18), which is a university‑licensing monetization model rather than competitor‑directed privateering.

8. Defensive aggregator (anti‑NPE) — NOT PRESENT.
The chain never reaches RPX, AST, LOT Network, Unified Patents, or OIN. Note the inverse relationship: Unified Patents attacked this patent (IPR2019‑01068 and IPR2019‑01069, both Not Instituted), and ODP/PTAB records show Apple (IPR2014‑00077, ‑00319, ‑00320) and Amazon (IPR2019 petitions) as challengers. The record is one of third parties defending against the patent, not aggregating it.

---

Verdict

NPE — high confidence.

The asserted exclusionary rights in US 7,177,798 were held by two successive single‑purpose, non‑producing Texas LLCs — Dynamic Advances, LLC (recorded exclusive license, 2013‑05‑10; disclosed parent Marathon Patent Group, Inc., Rule 7.1 statement July 18, 2014) and then CF Dynamic Advances LLC (RPI consent 2017‑09‑29; parent CF Marathon, LLC) — with no products, assertion as the sole business purpose ($24.9M Apple settlement, 2016‑04‑18), and documented Spangenberg/IPNav involvement (Spangenberg affidavit, CF Dynamic Advances LLC v. TechDev Holdings, 2019). These are corroborated by the collateral‑security chain to DBD Credit Funding LLC (2015‑02‑02 and 2017‑01‑11) and the pre‑recordation 2012‑10‑19 Apple filing. The patent is now expired (Nov 2021) and its claims were held §101‑ineligible (N.D.N.Y. 2024‑03‑18), affirmed by the Federal Circuit in consolidated appeals 24‑1725 / 24‑1739 — so the asserted asset has been judicially neutralized. The one honest qualification: legal title never left RPI, a university rather than an NPE, and the 2023 holding that "RPI … remains the patentee" means this is a license‑driven NPE chain, not an ownership‑transfer NPE chain.

Verify at: https://assignment.uspto.gov/patent/index.html (or https://assignmentcenter.uspto.gov/) — search patent number 7177798 or application 09/861,860; the inventor assignment should appear at Reel 012195 / Frame 0213. Reel/frame and correspondent for the 2013, 2015 and 2017 recordings should be read directly off those record pages — I was unable to retrieve them here and have deliberately not guessed them.

I'll research US patent 7177798 and its cited prior art references.

I found the key reference lists. Let me verify the complete "References Cited" list and check the IPR art, which is central to invalidity analysis.

Let me get the claim set and check what prior art was actually asserted against this patent in the PTAB challenges.

Prior-Art Analysis — US 7,177,798 B2

Identifier interpreted literally: "7177798" = US 7,177,798 B2 (not 7,177,978, 7,177,798-family variants, etc.). All dates below are as printed and are not corrected.

---

1. The patent under analysis

Contested history (context, not prior art): PTAB IPR2019-01068 and IPR2019-01069 (both not instituted — merits); IPR2014-00320 and IPR2014-00319 (not instituted — procedural); IPR2014-00077 (not instituted — merits). District-court suits in N.D.N.Y. (1:12-cv-01579; 1:13-cv-00633; 1:18-cv-00549), S.D.N.Y. (1:23-cv-00227), and CAFC appeals 24-1739 / 24-1725. Apple's Siri dispute settled in 2016 for $24.9M.

Claim architecture (basis for the §102 mapping)

I could not retrieve verbatim claim text within this session. From the patent's own Summary, the 21 claims are organized around three independent claims (approximately claims 1, 13 and 20) and their dependents:

• Independent claim A (≈ cl. 1): receive a natural-language query → search one or more language-based databases including a metadata database containing at least one of case information, keywords, information models, database values → return a result through a user interface; dependents add keyword identification, mapping to combinations of a finite number of database objects, reference dictionary, ambiguity resolution, and rules/heuristics for optimal interpretation.
• Independent claim B: provide database objects → identify a finite number of permutations → interpret a permutation to determine the result; dependents add data types / instances / operators, mapping to a database query (SQL), path-finding through the reference dictionary, elimination of permutations, and adding new cases/keywords at runtime.
• Independent claim C: determine recognized terms against a logical-graph data dictionary → determine minimal recognized terms/vertices → find minimum-cost query graph → remove redundant graphs → determine a complete solution → translate to a query-language statement.

⚠️ All claim mappings below are provisional and keyed to these three claim families, not to verified claim numbers.

---

2. Mandatory caveats before reading the tables

1. Two very different §102 questions. The user asked which citations "potentially anticipate under §102." In fact:
   • Almost every reference on the face of a patent is applied by the examiner as §103 art (alone or in combination), not as §102 anticipation. For this patent, no cited reference was found to have been used as a standalone §102 rejection of the issued claims, and the four PTAB petitions targeting it were not instituted.
   • References that postdate the ~April 2000 priority date cannot be §102(a)/(b) prior art at all. Several face-of-patent entries have 2002–2004 issue/publication dates (e.g., 6,356,905; 6,505,175; 6,505,183; 6,546,388; 6,556,983; 6,625,583; 6,711,585; 6,718,324; 6,766,320; pubs. 2002/0087578 and 2004/0236778). They are only citable, if at all, under §102(e) (filed before, published/issued after) — which requires confirming each one's actual filing date, which I did not verify.
2. Purpose contamination. Several US patent references (payments/agents/e-commerce) appear to have been carried over from the parent application 09/544,676 "System and method for accessing personal information." Those references are irrelevant to the language-interface claims and I flag them as such rather than treating them as §102 art.
3. Method note: I could not directly query the USPTO full-text/PatFT or PatentCenter APIs with the tools available. The lists below are compiled from the patent's own front page as reproduced in the IPR exhibit (Exhibit 1001, Docket Alarm) and consistent public mirrors (Google Patents / Justia). I could not fetch the IPR2019-01068/-01069 petition exhibits in this session, so I cannot confirm which references petitioners actually asserted.
4. Descriptions marked [verified] I am confident of; those marked [class/date only] are compiled from the number, date, inventor and US classification and should be re-checked against the document.

---

3. Table A — Most relevant prior art (the references that actually bear on the claims)

Also cited [verified bibliographic]: Weizenbaum 1966 (ELIZA, CACM 9(1):36–44); Waltz 1978 (CACM 21(7):526–539); Codd 1978 (RENDEZVOUS, in Databases: Improving Usability and Responsiveness, pp. 3–28); Gross, Appelt, Martin & Pereira 1987 (TEAM, ACM Transactions 32:173–243); International Search Report PCT/US00/09265.

---

4. Table B — Remaining US patent references on the face of the patent

These are cited but carry little to no §102 weight against the language-interface claims. Most are payment, personal-information, agent, or e-commerce references apparently inherited from parent application 09/544,676. I list them for completeness with dates as printed.

Additional NPL cited from the parent's ISR (agent/commerce literature, not language-interface art): Burns; Decker et al. (1996); Gregg; Hadidi et al. (1998); He et al. (1997); Moore et al.; Pannu et al.; Sameshima et al. (1997); Sugumaran; Sycara et al.; Zeng et al. (1995); Zhao.

---

5. Bottom line on §102

• Nothing on the face of US 7,177,798 appears to be a true standalone §102 anticipatory reference. The claims are characterized by the combination of (i) a metadata database integrating cases, keywords, information models and database values, (ii) enumeration of finite permutations / minimum-cost query graphs over a logical structure, and (iii) case-based learning feedback. No single cited reference teaches all of these.
• The closest art is Shimazu (CAPIT, 1992) for keywords + case-based parsing, Owel (1997) for NL-query filtration into a conceptual model, Kamel (1990) for graph-based knowledge retrieval, and Braden-Harder (US 5,933,822) for dictionary generation. These are the references a challenger would combine under §103 (and were likely the type relied on in the non-instituted IPRs).
• The 2002–2004 entries cannot be §102(a)/(b) art; treat them as potential §102(e) references only after verifying their actual filing dates. Likewise, the payment/agent references from the parent application have no bearing on the language-interface claims.

---

6. Confidence & limitations

• High confidence: the identity of the patent, its filing/grant/priority data, claim count, assignee, litigation/PTAB history, and the existence and bibliographic data of the Table A NPL items.
• Medium confidence: the complete membership of the "References Cited" block, the mapping of individual references to specific claim families, and issue-date-to-filing-date relationships for the post-2000 references.
• Not verified in this session (explicitly unknown): (a) the verbatim text and numbering of claims 1–21; (b) the specific prior-art references and grounds asserted in IPR2019-01068 / IPR2019-01069 and the earlier 2014 petitions; (c) the actual filing dates needed to qualify the post-2000 patents as §102(e) art. If you need the exact IPR grounds or verbatim claim chart, those must come from the PTAB record (petitions/exhibits) and the issued claim set, which I was unable to retrieve here.

I have not auto-corrected any number: 7,177,798 is analyzed as printed, and the near-miss strings that appeared during research (e.g., 7,177,978) were not substituted.

I'll start by grounding both the patent's claims and the prior-art references cited on the page.

Now let me retrieve the actual claim set and the PTAB record on the art that was asserted.

Obviousness Analysis — U.S. Patent No. 7,177,798 B2

Prepared: April 26, 2026 · Basis: Google Patents record for US7177798B2 as fetched, plus PTAB/docket records located by search. All alphanumeric identifiers are reproduced literally as they appear in the sources.

---

1. Scope, identifiers, and a caveat about the "Prior Art" section

Important note on my source material. The Google Patents page as fetched does not render the examiner-cited patent references list, nor the "Similar Documents"/"Cited By" tables. The only prior-art content actually present on the page is (a) the machine-extracted "Prior art keywords" list — query, natural language, database, keywords, input — and (b) the narrative prior-art discussion in the Background/Detailed Description, which cites specific non-patent literature. The analysis below is built from (b). I have deliberately not invented patent numbers for references I could not retrieve. A complete §103 opinion would require the file wrapper and the IPR exhibits, several of which I identify in §7.

Note on the priority date. The page lists a "prior art date" of 2000-04-07, while the specification recites an Apr. 17, 2000 parent and a May 19, 2000 provisional. I flag rather than reconcile this: the discrepancy affects only which side of the critical date the inventors' own 1996 publications fall — and they fall on the prior-art side under every candidate date.

Statutory framework. Because application 09/861,860 was filed May 21, 2001, pre-AIA 35 U.S.C. § 103(a) governs, with the Graham v. John Deere factors and KSR Int'l Co. v. Teleflex Inc., 550 U.S. 398 (2007), supplying the motivation-to-combine rubric. The inventors' own published work (Hsu 1994, 1996) is §102(b) statutory-bar art.

---

2. The claims to be analyzed

Three independent claims (1, 9, 21); the balance are dependent.

Claim 1 — a method comprising: (i) providing a natural language query input to the user; (ii) performing, based on the input, a search of one or more language-based databases including at least one metadata database comprising at least one of a group of information types comprising case information, keywords, information models, and database values; (iii) providing, through a user interface, a result of the search to the user.

Claim 9 — a computer-implemented method comprising: (i) receiving a natural language input; (ii) providing from said natural language input a plurality of language-based database objects; (iii) identifying a finite number of permutations of the plurality of database objects, the database objects being stored in a metadata database comprising at least one of a group of information comprising case information, keywords, information models, and database values; (iv) interpreting at least one of the permutations to provide determination of a result of the natural language input. (Claim 9's text as quoted in the 2012 infringement complaint, as reported by The Register; treat as a secondary reproduction, not the certificate text.)

Claim 21 — determining recognized terms in a data dictionary having a logical graph structure; determining a minimum number of the recognized terms and their vertices; determining at least one minimum-cost query graph containing a minimum number of vertices; if more than one, removing redundant query graphs to produce a solution set; determining within the set at least one complete-solution query graph; and translating that graph to a query-language statement.

Key construction point that lowers the §103 burden: claims 1 and 9 recite a Markush-style "at least one of a group comprising case information, keywords, information models, and database values." On its face that requires only one of those four layers. The N.D.N.Y. magistrate judge recommended declining to construe five method limitations, including "determining a plurality of combinations of the finite number of database objects" (claim 1), "resolving ambiguity between the keywords and the plurality of database objects" (claim 5), "identifying a finite number of permutations of the plurality of database objects" (claim 9), and "interpreting at least one of the permutations…" (claim 9), finding them "readily understood as stated" (Report & Recommendation, RPI v. Amazon, 1:18-cv-00549). Broad, plain-meaning scope makes the prior art read on the claims more easily.

---

3. The prior art available from the page

Non-patent literature cited in the Background (all pre-date every candidate priority date):

1. Weizenbaum, J., "ELIZA — A Computer Program for the Study of Natural Language Communication between Man and Machine," CACM 9(1):36–44 (1966) — template/keyword-driven response generation.
2. Waltz, D.L., "An English Language Question Answering System for a Large Relational Database," CACM 21(7):526–539 (1978) — syntax-based NL→relational-DB query translation.
3. Codd, E.F., Arnold, R.S., Cadiou, J-M., Chang, C.L., Roussopoulos, N., RENDEZVOUS Version 1: An Experimental English Language Query Formulation System for Casual Users of Relational Data Bases, IBM Research Report RJ2144 (Jan. 1978); also Codd, "How About Recently? (English Dialog with Relational Databases Using Rendezvous Version 1)" (1978) and Codd, "Access to Relational Data Bases for a Casual User" (full text; mirror; ACM copy).
4. Hendrix, G.G., Sacerdoti, E.D., Sagalowicz, C., Slocum, J., "Developing a Natural Language Interface to Complex Data," ACM TODS 3(2):105–147 (1978) (semantic-grammar approach); also Hendrix, "Natural-Language Interface," Am. J. Computational Linguistics 8(2) (1982).
5. Gross, B.J., Appelt, D.E., Martin, P.A., Pereira, F.C.N., "TEAM: An Experiment in the Design of Transportable Natural-Language Interfaces," Computational Linguistics 32:173–243 (1987).
6. Janus, J.M., "The Semantics-Based Natural Language Interface to Relational Databases," in Bolc & Jarke (eds.), Cooperative Interfaces to Information Systems, Springer-Verlag, pp. 143–187 (1986).
7. Motro, A., "FLEX: A Tolerant and Cooperative User Interface to Databases," IEEE Trans. Knowledge and Data Engineering 2(2):231–246 (1990).
8. Guida, G. & Tasso, C., "NLI: A Robust Interface for Natural Language Person-Machine Communication," Int'l J. Man-Machine Studies 17:417–433 (1982).
9. Allen, J., Natural Language Understanding, Benjamin/Cummings (1994) — the standard graduate textbook on NLU (cited in the spec as incorporated by reference).
10. Hsu, C., Enterprise Integration and Modeling: The Metadatabase Approach, Kluwer (1996); Hsu, C., et al., "The Metadatabase Approach to Integrating and Managing Manufacturing Information Systems," J. Intelligent Manufacturing 333–349 (1994); Hsu, C., et al., "The Model-Assisted Global Query System for Multiple Databases in Distributed Enterprises," ACM Trans. Information Systems 14(4):421–470 (Oct. 1996) (the MSQL system).
11. Microsoft MindNet (conceptual-network logical-structure approach); MIT JUPITER; CMU MOVIELINE; ASK.COM / Ask Jeeves.

This is an unusually rich prior-art record, and — critically — items 10 are the inventors' own pre-1998 published work. Together with the specification's own admissions (§4), they supply most of the "reference dictionary" concept as a matter of record.

---

4. The specification's own admissions

A §103 analysis may rely on the applicant's characterization of the art. The Background here concedes:

• five named approaches already existed (template, syntax, semantics-grammar, intermediate-representation-language, semantics-model) — i.e., every architectural family, including the semantics-model approach that "provide[s] a semantic model or a dictionary as the roadmap to generate possible interpretations";
• the finite-domain insight was already understood: "users are bound to refer… to these known database objects (types or semantic models, instances or values, and operators)";
• the problem was already known: keywords "could increase exponentially as the number of users and usage patterns increase";
• the entity-relationship model is adopted as the modelling substrate because "it is both sufficient… and other database models can be translated into it" — i.e., ER modelling is conventional;
• the interpretation problem is "related to the vast tradition of constrained optimization (e.g., scheduling and traveling salesman)" — i.e., branch-and-bound is applied, not invented.

Under KSR, these admissions define a finite design space in which the remaining work is the predictable combination of known elements.

---

5. Element mapping

---

6. Combinations that render the claims obvious

Combination A (core, hits claims 1–8): RENDEZVOUS (1978) + Motro FLEX (1990) + Hsu metadatabase (1994/1996)

• RENDEZVOUS discloses the entire claim-1 architecture: free-form English input, interpretation against the relational schema, clarification dialog when "ambiguity," "erroneous," or "incomprehensible" input is encountered, multiple-choice fallback, English restatement for user confirmation, and translation into a formal data sublanguage (DEDUCE). It also expressly aimed to be "readily adaptable… to any relational data base" by making code "as independent as possible of the domain of discourse."
• FLEX supplies the semantic-model/dictionary layer that RENDEZVOUS lacked, plus tolerance for imprecise and ambiguous queries and cooperative disambiguation — i.e., claim 5's "resolving ambiguity between the keywords and the plurality of database objects" and claim 7's rules/heuristics.
• Hsu supplies the enterprise metadata repository — information models, rules, database values — and the shortest-path global-query mechanism, i.e., the "metadata database" and the inference of implied elements.

Motivation. All three are in the same field (NL access to relational/enterprise databases) and address the same two recognized problems: ambiguity/incompleteness of natural articulation, and portability of the interface across domains. RENDEZVOUS itself supplies the express motivation to make the query-formulation component domain-independent — which is precisely the function FLEX's semantic model and Hsu's metadatabase perform. Combining them is "the combination of familiar elements according to known methods" yielding "predictable results" (KSR).

Combination B (hits claims 1–9): Hendrix et al. (1978) + Gross et al. TEAM (1987) + Hsu metadatabase (1994/1996)

• Hendrix teaches that domain semantics can drive NL interpretation for complex data — the semantic grammar is the information model.
• TEAM teaches the transportable variant: a domain lexicon acquired from the database, kept separate from the linguistic component, with translation through an intermediate representation — essentially an "intermediate dictionary of results" as the patent's own title phrases it.
• Hsu supplies the enterprise-scope metadata layer and global-query reformulation.

Motivation. TEAM's stated purpose is portability between database domains; Hendrix supplies the semantic interpretation; Hsu supplies the enterprise repository. A PHOSITA seeking both naturalness and portability — the exact tension the '798 Background describes — would combine them as a matter of routine design.

Combination C (hits the "case information" and learning limitations): Guida & Tasso (1982) + Motro FLEX (1990) + Hsu (1994/1996) + case-based reasoning as known in the art

• Guida & Tasso's "robust interface" and FLEX's "tolerant and cooperative" design both describe interfaces that accept incomplete input and adapt across sessions.
• Case-based reasoning (problem definition → solution → outcome) was a mature paradigm before the priority date; the patent itself describes it as "typically" having three components, confirming its conventionality.
• Adding a case layer alongside keywords/models/values in the metadata repository is the predictable application of a known technique to a known structure.

Motivation. The field's recognized need for "closure" of failed queries and for reducing the number of candidate interpretations supplies the rationale; CBR is the known mechanism for reusing past solutions.

Combination D (hits claim 21 and its dependents): Hsu MSQL (1996) + RENDEZVOUS/TEAM/Hendrix + branch-and-bound / graph search as known in the art

• Claim 21's "minimum cost query graph… minimum amount of vertices" is squarely disclosed in concept by Hsu's MSQL, which "determine[s] the shortest path when alternative paths exist."
• "Translating the at least one cost query graph to a query language statement" is disclosed by RENDEZVOUS (→DEDUCE), TEAM (→intermediate formal language), and Hendrix.
• The redundancy-removal and complete-solution steps are ordinary graph-search pruning. The specification itself concedes the interpretation problem "relates to the vast tradition of constrained optimization," making branch-and-bound application a known technique for a known purpose (KSR factor 4).

Combination E (front end only): ASK.COM + Motro FLEX

ASK.COM discloses free-format NL input, recognized-keyword capture, retrieval of related keywords, and presentation of candidate questions/results through a UI — claims 1(i)/(iii) and the keyword layer. Its admitted deficiency (the patent calls it "a text classification system" that "does not actually interpret natural language queries") is cured by FLEX's semantic-model interpretation. Under KSR, "the improvement of one reference by another" is a classic rational combination.

Aggregate conclusion. Claims 1 and 9 are, on this record, strongly vulnerable: every recited element appears in the art, the Markush "at least one of" language requires only one metadata layer, and the district court has signaled that the limitations are plain-meaning. Claims 2–8 fall with claims 1/9 given Hsu's finite-object model, Motro's ambiguity resolution, and Hsu's rules. Claim 21 is the most defensible, because it requires a specific ordered algorithm (minimum terms → minimum vertices → minimum-cost graphs → redundancy removal → complete solution → translation) that no single reference steps through; it is nonetheless vulnerable to Combination D.

---

7. Countervailing record — why this is not a clean win

1. Every PTAB challenge failed at the threshold. The page itself records IPR2014-00077 (not instituted – merits), IPR2014-00319 and IPR2014-00320 (not instituted – procedural), and IPR2019-01068 / IPR2019-01069 (not instituted – merits). In the 2019 decision, the Board concluded "the information presented does not show there is a reasonable likelihood that Petitioner would prevail in establishing the unpatentability of any challenged claim" of claims 1–8 (IPR2019-01068) or 9–21 (IPR2019-01069) (decision; case record). Institution denials are not merits adjudications and have no preclusive effect, but they are meaningful evidence that the art actually asserted did not clearly teach the claimed subject matter.
2. The asserted art is not the art I analyzed. Declarations of Dr. James F. Allen — the author of the very Natural Language Understanding textbook the '798 patent incorporates by reference — appear in the 2014 Apple petitions (IPR2014-00077 and IPR2014-00320, Ex. 1005). A claim chart in the 2014 record frames analysis of claims 1–21 "in view of the prior art." I could not retrieve the substantive references asserted in those petitions, and I will not guess at them. A complete opinion must review those exhibits and the Amazon petition's grounds.
3. Parallel §112 attack. In RPI v. Amazon (1:18-cv-00549, N.D.N.Y.), Amazon moved for summary judgment on patentability and enablement with a Daubert challenge to its own expert admissibility, and the plaintiffs cross-moved to exclude defense expert testimony (docket). This indicates the defendants' lead theory may be §112 enablement/written description rather than §103.
4. Live appellate activity. The page lists 2024 Federal Circuit appeals Nos. 24-1739 and 24-1725 arising from this family. Their substance is unknown to me; any validity conclusion is provisional pending them.
5. Claim 21's specificity remains the best non-obviousness argument, and the "reference dictionary" as an integrated four-layer graph with cross-references (the patent's stated point of novelty over each individual layer) is the strongest secondary argument.

---

8. Bottom line

• Independents 1 and 9: likely obvious over Combination A (RENDEZVOUS + FLEX + Hsu) or Combination B (Hendrix + TEAM + Hsu), reinforced by the applicant's own admissions of the five known approaches, the known finite-domain insight, and the known keyword-explosion problem. The Markush "at least one of" metadata language and the district court's plain-meaning treatment materially narrow the gap the patentee must close.
• Dependents 2–8: obvious with the addition of Motro's ambiguity resolution and Hsu's rule/governance layer (Combination C).
• Independent 21: the hardest claim; best attacked by Combination D (MSQL shortest-path + a known branch-and-bound/graph-reduction technique + a known NL→query-language translation), but the ordered six-step algorithm is the patentee's strongest redoubt.
• Caveats I cannot resolve from the sources provided: the verbatim text of claims 2–8 and 10–20; the examiner-cited patent references on the Google Patents page (not rendered in my fetch); and the specific prior-art references asserted in IPR2014-00077 / IPR2014-00319 / IPR2014-00320 and IPR2019-01068 / IPR2019-01069. Those are the materials that would convert this screening analysis into a defensible opinion.