US 6101468

I'll research US 6,101,468 with targeted searches, then reconcile against the authoritative full-text I already have.

Now let me check for any litigation or CAFC 2026 docket activity specific to this patent number.

Let me verify the total claim count and check for any reexamination or later legal events.

US Patent 6,101,468 — Analyst Summary

Bottom line: US 6,101,468 is the Dragon Systems "DragonDictate" speech-recognition omnibus patent (a 1997 divisional claiming 1992 priority), granted 8 Aug 2000 and now expired. It has 13 claims, with two independent claims (1 and 10), both directed to case-insensitive alphabetic filtering during word recognition. I found no CAFC 2026 docket activity and no currently pending litigation specifically identifying this patent number — see the caveats at the end.

1. Bibliographic data

2. Abstract (as published)

A word recognition system can: respond to the input of a character string from a user by limiting the words it will recognize to words having a related, but not necessarily the same, string; score signals generated after a user has been prompted to generate a given word against words other than the prompted word to determine if the signal should be used to train the prompted word; vary the number of signals a user is prompted to generate to train a given word as a function of how well the training signals score against each other or prior models for the prompted word; create a new acoustic model of a phrase by concatenating prior acoustic models of the words in the phrase; obtain information from another program running on the same computer, such as its commands or the context of text being entered into it, and use that information to vary which words it can recognize; determine which program unit, such as an application program or dialog box, currently has input focus on its computer and create a vocabulary state associated with that program unit into which vocabulary words which will be made active when that program group has the focus can be put; detect the available computational resources and alter the instructions it executes in response; test if its ability to respond to voice input has been shut off without user confirmation, and, if so, turn that ability back on and prompt the user to confirm if that ability is to be turned off, store both a first and a second set of models for individual vocabulary words and enable a user to selectively cause the recognizer to disregard the second set of models for a selected word; and/or score a signal representing a given word against models for that word from different word model sets to select which model should be used for future recognition.

Important scope note: The abstract and specification describe roughly ten distinct invention aspects (training-sample scoring, command recognition during training, phrase-model concatenation with coarticulation, cross-application vocabulary states, language-context tracking, computational-resource detection, microphone-off confirmation, "forget training" of helper models, and base-vocabulary selection). Only the claim set limits the patent. The granted claims are far narrower than the disclosure — see §4.

3. Specification context (brief)

The patent is a hardware/software disclosure of two Dragon Systems products: MS‑DOS DragonDictate (a TSR program) and DDWin (a Microsoft Windows version). It covers the .VOC/.USR file structures, PIC/PEL acoustic modeling, adaptive and batch training, the tutorial program with its Event Stack, the OOPS buffer and choice lists, WordTraining/TrainNewModel routines, CommandTracking/MenuPick/ControlPick for voice control of other Windows applications, LanguageContextTracking, DSP-board detection, and the "Forget Training" and microphone-confirmation features. It expressly incorporates US 4,783,803; 4,903,305; 4,866,778; 5,027,406; 4,829,578; and 4,837,831, and cross-references parent application 07/976,413.

4. Plain-language overview of the independent claims

Claim 1 — Independent method claim: case-insensitive alphabetic filtering

A computerized method of word recognition with these steps:

1. Receive user-generated word signals (the words to be recognized).
2. Store vocabulary words, and for each store (a) pattern information identifying signals that represent that word and (b) a spelling whose letters can be upper or lower case.
3. Perform pattern matching between incoming word signals and the stored pattern information to select the one or more vocabulary words most probably corresponding to each signal.
4. Receive user-generated alphabetic signals associated with individual word signals.
5. The key limitation: the pattern matching, in response to a given alphabetic signal, must increase the probability that it selects a vocabulary word whose spelling contains the character represented by that alphabetic signal — even though the case of that character in the vocabulary word's spelling is not represented by the user's alphabetic signal.

Plain reading: type (or say) a letter to narrow recognition to words containing that letter, and the system will match regardless of upper/lower case. The specification ties this to re-recognition via the choice list / Oops buffer.

Claim 10 — Independent system claim: "related string" filtering logic

A computerized system for word recognition comprising:

1. Elements receiving user-generated word signals.
2. Elements storing vocabulary words, with pattern information and spelling (upper- or lower-case letters) for each.
3. Elements performing pattern matching to select which vocabulary word(s) most probably correspond to each signal.
4. Elements receiving a user-generated string of one or more textual characters associated with an individual word signal.
5. Filtering elements that respond to that string by causing the pattern matching to select the vocabulary word from a class of words whose spellings each contain a string "related to" (not necessarily identical to) the user's string — where the filtering logic requires the related string be similar to the user string in certain ways, but need not be similar in other ways.

Plain reading: the system claim is drafted deliberately more abstractly than claim 1 — it claims a filter using a partially matching rule, leaving the specific "which ways must match / which need not" to dependent claims 11–13. This is the classic drafting pattern (broad genus in the independent claim, the case rules in dependents).

Dependent claims (context for the two independents)

• 2–4 (depend on 1): acoustic/spoken-word embodiment; spoken "alphabetic" signals (e.g., spelling-alphabet words) that are themselves recognized; and keyboard-entered alphabetic signals.
• 5–7 (depend on 1): signals may represent either case; store the entered character sequence with its case, and selectively output either the vocabulary word's spelling or the user's own entered characters with their case; a lower-case input matches either case, but an upper-case input matches only upper case.
• 8–9 (depend on 1): the alternate-choice group is limited to words whose spellings start with the entered character sequence, but is not substantially limited to matching the case sequence; and the entered characters plus the alternate-choice group are displayed for user selection.
• 11–12 (depend on 10): the filter logic requires the letters to match but not the case; with lower-case user characters matching either case, and upper-case user characters requiring both letter and case to match.
• 13 (depends on 10): the "related string" must correspond to the start of the spelling of the pronunciation of its vocabulary word, and need not correspond to the start of the word's spelling when the spelling includes characters preceding the pronunciation (the specification's "wordname [pronunciation]" bracket convention, e.g., "3 [three]").

5. Litigation / docket check

• No CAFC 2026 docket entry, appeal, or IPR specifically identifying US 6,101,468 was located. I searched by the exact number and by combinations with Nuance, Federal Circuit, and 2026. Results returned other patents' disputes (e.g., US 5,715,468 in the Nuance v. Philosophy/CNU case; the Nuance v. ABBYY/Lexmark and AI Visualize v. Nuance appeals; various 2025–2026 IPRs involving Dialect, InterDigital/Disney-Dolby, and VoiceBox-derived patents) — none of these are US 6,101,468.
• Google Patents carries a "Family has litigation" flag for this family (Darts‑ip family ID 27009904). That flag indicates the family has been litigated historically, not that this specific patent is in an active 2026 case. I could not verify the specific case(s) from the sources retrieved.
• Enforceability: because the patent expired on 13 Nov 2012, it cannot be asserted today; any 2026 "docket" activity concerning it would be unusual (e.g., a lingering validity/ownership or damages issue in a case filed earlier), and I found none.

Do not conflate the following look‑alikes that appeared in search results:

• US 5,715,468 — a different patent (Philosophy/Nuance "natural language" line) asserted against Nuance.
• A Brazilian state commercial-registry entry "6101468 EDUKAR ESTAGIOS..." — an unrelated registration number, not a US patent.
• US 6,101,410, 6,107,562, etc. — unrelated numbers surfaced by substring matching.

6. Explicit uncertainties

1. Claim text provenance. The authoritative full text I was given is truncated before the claims section, so I could not read the granted claims from the primary source. The claim language quoted above is taken from two independent secondary reproductions (FreePatentsOnline's claims listing and a PatentHub CPC mirror at trade.patenthub.cn/.../US6101468A), which agree with each other (claims 1–9 on one, 10–13 on the other) and are consistent with the specification. I could not independently confirm from the primary record whether the patent has exactly 13 claims, so treat "13 claims, independents at 1 and 10" as high-confidence but not primary-source-verified.
2. No reexamination/certificate check completed. My attempt to search for a reexamination certificate or PTAB proceeding for this number was cut off, so I cannot affirmatively state whether any reexamination certificate issued. Given the 2012 expiration, any such proceeding would predate that date.
3. Priority-date discrepancy across databases. The authoritative record lists priority as 1992‑11‑13; a Unified Patents portal entry showed 1992‑11‑12 for this patent (likely a time-zone/normalization artifact). I have used 1992‑11‑13 per the authoritative text.
4. Assignee name as recorded. Assignment records list a security-interest holder as "USB AG, STAMFORD BRANCH" (2006). I am reporting this literally as recorded and am not correcting it to "UBS AG," per the instruction to interpret identifiers literally.
5. Apparent family siblings. Search results showed Dragon Systems patents with the same 1992‑11‑13 priority, including US 5,909,666 and US 6,092,043, which are likely divisional siblings from 08/382,752. I did not verify their exact lineage and do not rely on that characterization here.

I'll research litigation involving US patent 6101468. Let me run several searches.

Bottom line

I could not verify any litigation specifically asserting U.S. Patent No. 6,101,468. My searches returned no district-court complaint, docket entry, PTAB proceeding, or ITC filing that names 6,101,468 as an asserted patent. I am reporting that as a negative finding rather than substituting results for other patents in the same Dragon Systems priority family (e.g., U.S. 5,428,707; 5,850,627; 5,915,236; 5,920,836; 5,920,837; 5,983,179; 6,073,097). Those are different patents and must not be conflated with 6,101,468 under the strict "no auto-correction" rule.

What the authoritative record shows for US 6,101,468

Source of the above: https://patents.google.com/patent/US6101468/en and https://pubchem.ncbi.nlm.nih.gov/patent/US-6101468-A

One verified lead (family-level, not patent-level)

The Google Patents record for US 6,101,468 carries a "Family has litigation" flag linking to a Darts‑IP worldwide family litigation dataset:

• https://patents.darts-ip.com/?family=27009904 (family identifier 27009904), referenced from https://patents.google.com/patent/US6101468/en

This flag is at the family level and is a commercial database indicator, not proof that 6,101,468 itself was asserted. It is the single most useful pointer I found, and I could not resolve it to specific case names, courts, or docket numbers within my search budget.

Litigation table (as requested)

What I searched (and what came back)

• Unified Patents litigation portal (portal.unifiedpatents.com, incl. the caselist/patent pages) — returned unrelated patents (e.g., US 9,668,024 B2); no 6,101,468 hit.
• General web / PACER-and-courtlistener-style queries for "6,101,468" and "6101468" as asserted patents — returned only the patent's own bibliographic pages, its citing/cited-art listings, and unrelated litigations involving other patents (NICE Systems v. Witness Systems, D. Del. 1:06‑cv‑00311, which asserted Dictaphone/NICE patents such as 5,274,738; 5,396,371; 5,819,005; 6,249,570 — these are not 6,101,468).
• I could not confirm that Dragon Systems, Inc. v. Dictaphone Corp., or any Nuance/ScanSoft enforcement action, actually named 6,101,468 as an asserted patent. I have a prior general awareness of Dragon-family speech-recognition patent disputes in the late 1990s, but I do not have high-confidence documentation tying the asserted patent numbers in those cases to 6,101,468, so I will not state them as fact.

Recommended verification steps (given the strict no-fabrication rule)

To close this out properly, these sources should be checked directly, because the patent-level docket detail is not available in open web search results:

1. Darts‑IP family 27009904 — the flag source above; a subscription is typically required, and the Darts‑IP portal is the only place the Google Patents page points for the "family has litigation" assertion.
2. PACER / CourtListener RECAP — search the patent number field ("6,101,468" and "6101468") plus assignee strings: Dragon Systems, L&H Holdings USA, ScanSoft, Nuance Communications. Note that pre‑2000 cases are frequently missing from free dockets, and the 1992–2000 window is exactly when Dragon-family enforcement would have occurred.
3. USPTO PTAB (PTAB E2E / Patent Center "Proceedings" tab for 6,101,468) — to confirm no IPR/PGR/CBM or derivation proceeding.
4. USPTO Assignment records (reel/frame entries above) — useful to identify parties who would have had standing to sue in each time window (Dragon pre‑2000; L&H Holdings 2000–2002; ScanSoft 2002–2006; Nuance 2006–2012).
5. ITC EDIS (Section 337) — unlikely for this patent, but a quick negative check.

Confidence statement: I have high confidence that 6,101,468 is a Dragon Systems continuation-in-part in the 1992‑11‑13 priority family and that it expired in 2012. I have low/insufficient confidence about any specific litigation, and therefore report none rather than attribute cases from related Dragon-family patents to this patent number.

I'll start by verifying the structured data against what's publicly searchable.

The first-pass results are returning a different patent (9,679,146 / 8,867,610 — note the "146" ambiguity). Let me run narrower queries keyed to this patent's exact number and owner history.

Let me try patent-number variants and family-litigation leads.

PTAB / AIA Trial Analysis — US Patent 6,101,468

Prepared: 2026-09-28 · Patent: US 6,101,468 B1 ("Apparatuses and methods for training and operating speech recognition systems")
Owner on record: Nuance Communications, Inc. (orig. Dragon Systems → L&H Holdings USA → ScanSoft → Nuance)
Priority 1992-11-13 · Filed 1997-06-26 · Granted 2000-08-08 · Anticipated expiration 2012-11-13 (Expired – Lifetime)
Source of record: Google Patents US6101468A; USPTO ODP "PTAB proceedings on file" block supplied with this task.

Proceedings overview

Zero AIA trial proceedings are on file against US 6,101,468 — 0 active, 0 with claims invalidated, 0 with claims sustained, 0 settled, 0 institution denials. The structured ODP block is the canonical list and returns no IPR, PGR, or CBM petition, and my independent web checks surfaced no petition, institution decision, FWD, or appeal naming this patent. The bottom-line defensive posture is therefore not "hardened by PTAB" and not "claims canceled" — it is that the patent is expired (2012-11-13) and has never been tested at the PTAB. For a defendant, the operative defense is the expiration date and prior art in district court, not an IPR outcome.

Verification caveat (important — near-miss patent numbers). My searches returned a large volume of PTAB material referring to "the '146 patent" and "the '610 patent." Those are different patents: IPR2020-01040 concerns U.S. Patent 9,679,146 B2 (PTAB record) and IPR2018-01331 concerns U.S. Patent 8,867,610 B2 (institution decision, 2019-01-31). Neither is US 6,101,468. Do not let "the '146 patent" shorthand in any brief be conflated with 6,101,468 — that is exactly the kind of citation error that draws a sanctions motion under Rule 11 / § 285.

No proceedings to report — what was checked and what was found

Because there are no proceedings, I am listing the verification scope instead of a per-proceeding entry, so the negative finding is auditable:

• ODP structured list (canonical): no AIA trial proceedings. Verified 2026-09-28.
• Web search — patent number: no IPR/PGR/CBM identified for US 6,101,468. Searches keyed to "6,101,468", "US 6,101,468", and "6101468" + PTAB/IPR produced no petition or decision for this patent.
• Web search — parties/technology: Nuance is an active PTAB participant (e.g., MModal-vs-Nuance IPRs over the '398 patent; Nuance-owner IPRs over the '786 patent), but no Nuance IPR over 6,101,468 appeared. Those results are context only and are not attributable to this patent.
• Family-level flag (not a PTAB proceeding): the Google Patents record for US6101468 carries a "Family has litigation / First worldwide family litigation filed" flag pointing to a Darts-ip family entry (family=27009904). This is a litigation indicator at the family level (the '468 patent is in the 1992-11-13 priority family of the parent, US 07/976,413, which issued as US 5,428,707). It is not evidence of an AIA trial against the '468 patent specifically, and I could not, within this session, identify a district-court case asserting 6,101,468 by name. Treat this as an open lead, not a finding.

Why absence of PTAB activity is unsurprising here (and what it means)

Three structural reasons make a null PTAB history the expected result, not an anomaly:

1. Expiration kills the IPR incentive. The '468 patent expired 2012-11-13. IPRs on expired patents are permissible but rare and largely academic: claims cannot be amended, and there is no forward-looking injunctive or royalty exposure to defeat. Defensive aggregators (Unified Patents, RPX, etc.) do not typically spend on expired patents. I found no evidence of a defensive-aggregator challenger in this patent's chain of title.
2. PGR was never available. Post-grant review is limited to patents with an effective filing date on or after 2013-03-16. This patent's chain runs to 1992-11-13 / 1997-06-26 — far outside PGR eligibility.
3. CBM was inapplicable. Covered Business Method review required a financial-services "covered business method" patent. This is a speech-recognition / acoustic-modeling patent (CPC G10L15/06, G10L15/063, G10L15/22, G10L15/285), a technological invention — not CBM-eligible.

Timing window: IPRs became available 2012-09-16; the patent expired 2012-11-13. The theoretical filing window was roughly two months. No filing appears in it.

Strategic summary

Claim status: no claim of 6,101,468 has been canceled or sustained by the PTAB, because no claim has ever been adjudicated there. I therefore cannot list "surviving claims" or "canceled claims" — the FWD record is empty, and I will not manufacture claim numbers or a disposition. Google Patents does not expose the claim set in the material provided to me, so you must read the printed claims off the patent itself before drafting anything. What is established from the record: the patent is expired as of 2012-11-13, and the current assignee of record is Nuance Communications, Inc. (post-chain-of-title: L&H Holdings USA 2002-10-08 → ScanSoft 2002-10-08 → Nuance 2006-03-15; security interests to USB AG Stamford Branch in 2006, with Morgan Stanley releases recorded 2016-05-20 at Reels 017435/0199 and 018160/0909).

Estoppel landscape: § 315(e)(2) estoppel is a non-issue. Because no IPR/PGR was ever instituted, no petitioner and no privy is estopped, and no "ground raised or reasonably could have been raised" was ever locked in. For a defendant being asserted against today, the entire prior-art universe remains available — § 102 anticipation, § 103 obviousness, and § 112 defenses in district court are unconstrained by anything the Board has done. Conversely, there is no FWD to borrow, no Board finding to adopt, and no claim-construction ruling from the Board to invoke. Any invalidity case must be built from scratch in court.

Pattern signals: none of the usual ones are present. No serial petitioner (no petitioner filed even one), no PTAB appeals, no Federal Circuit docket on a PTAB outcome, and no indication of a defensive aggregator in the chain of title or the family history. Nuance's aggregate PTAB footprint is substantial, but it is litigation-driven on other patents; the '468 patent's own posture is one of historical dormancy ending in expiration.

The one thing that matters more than any of this: the patent, as recorded, expired on 2012-11-13 ("Status: Expired - Lifetime"; "Anticipated expiration: 2012-11-13"). That date is driven by the twenty-year term running from the earliest U.S. filing date in the § 120 chain (1992-11-13), consistent with the priority claim from Ser. No. 07/976,413. I state that as what the record shows; confirm the expiration computation and any § 154(b) patent-term adjustment from the face of the patent and PAIR before relying on it in a filing.

Recommended next steps

1. If you are a defendant facing a demand letter citing 6,101,468: the first move is not an IPR — it is the calendar. Confirm the 2012-11-13 expiration on the face of the patent and in Patent Center. If it holds, no ongoing-infringement theory survives, and any pre-expiration damages theory must be tested against the § 286 six-year lookback from the complaint date — which for a 2026 suit reaches back only to 2020, well after expiry. Ask opposing counsel, in writing, to identify (a) the specific expired claims asserted and (b) the infringement period. There is no FWD to link to; the disposition you cite is the expiration of the patent, documented at Google Patents US6101468A and in USPTO Patent Center.
2. Do not file an IPR. On an expired patent you cannot amend claims, and the Board's incentive to institute on a dead patent is low. It is a poor use of budget and, given the null estoppel position, it buys you nothing you don't already have in district court.
3. Run down the family-litigation flag before relying on the null finding. The Darts-ip family pointer (family=27009904) indicates litigation somewhere in this 1992-11-13 family. Pull the Darts-ip/Lex Machina family record and confirm whether 6,101,468 was ever asserted (which would give you a § 282 / § 315(b)-style litigation history and, potentially, prior invalidity rulings) — while remembering that family-level litigation is not patent-level, and that a district-court case is not an AIA trial.
4. Re-verify the ODP null result at institution-cycle milestones. ODP ingest lags. If, on any future check, a petition appears, the statutory clock is: institution decision due within 6 months of the § 315(b) bar date / petition filing, and FWD due within 12 months of institution (§ 316(a)(11)), with a possible 6-month extension for good cause. There is no such clock running today.
5. Attribute carefully in your own papers. If you cite this analysis, cite it as "no AIA trial proceedings on file," and add the affirmative near-miss warning: IPR2020-01040 (US 9,679,146) and IPR2018-01331 (US 8,867,610) are different patents that share only the "146"/"610" shorthand. Misattributing their outcomes to 6,101,468 would be a self-inflicted wound.

Confidence statement: high confidence in the null PTAB finding (the ODP structured list is empty and independent searching corroborates it); high confidence in the expiration date as recorded by Google Patents (2012-11-13) subject to face-of-patent confirmation; low-to-moderate confidence on the family-litigation lead, which I flag as unresolved rather than asserted. I did not locate, and therefore do not state, any claim-level PTAB outcome for US 6,101,468.

I'll start by pulling the authoritative record I already have (the Google Patents legal-events block) and then cross-check against assignment databases.

I have solid foundational data. Let me dig for the remaining reel/frame numbers and correspondents.

I've exhausted my search budget. Here is the reconstructed record, grounded in the Google Patents legal-events block for US 6101468 (fetched 2026-09-24) plus corroborating press/SEC-sourced material. I flag gaps explicitly rather than fill them.

Source note: The authoritative Google Patents legal-events block for this patent supplied several reel/frame numbers directly. I was not able to load the USPTO Assignment Center / assignment.uspto.gov record pages themselves, so two data points are missing and are marked as such below: (a) the reel/frame for the L&H Holdings USA → ScanSoft assignment, and (b) all correspondent-of-record data. Those are stated as gaps, not guessed.

Inventors

Application 08/882,914, filed 1997-06-26 (a continuation-in-part of 07/976,413, filed 1992-11-13). Nine named inventors:

All nine are Dragon Systems personnel — consistent with the subject matter (the specification is written around the MS-DOS DragonDictate program and DDWin, Dragon's own products).

Pattern note (partial): No inventor-by-inventor departure dates are recoverable from the sources I could reach, so I cannot confirm the "all inventors gone within 12 months" tell. What is documented: the entire Dragon research staff (~350 employees, 170 scientists/engineers) was rolled into Lernout & Hauspie at the 2000-06-07 merger, and James K. Baker then departed the acquiring chain entirely — he and Janet Baker bought L&H's AudioMining assets for $750,000 through a new entity, Dragon Catalyst LLC (announced at the Dec-2001 bankruptcy auction). That is a documented founder exit, not a founder exit within 12 months of filing.

Original assignee

Two layers, because the patent issued mid-merger:

• Filing assignee (1997-06-26): Dragon Systems, Inc., Newton, Massachusetts — a privately held speech-recognition software company. It shipped products that embody the claims: DragonDictate (MS-DOS) and, from spring 1997, Dragon NaturallySpeaking, the first continuous-speech dictation product. Dragon's 1999 revenue was ~$60M against a ~$22M net loss; its largest shareholder was Seagate Technology (~35%).
• Assignee of record at issuance (2000-08-08): L & H Holdings USA, Inc. Google Patents' "Original Assignee" field lists L&H Holdings USA Inc; the patent is cited elsewhere (e.g., DE69900981T2) under "Dragon Systems, Inc." The merger of Dragon into L&H Holdings USA took effect 2000-06-07, two months before grant, which explains the split. Treat Dragon Systems, Inc. as the originating assignee and L&H Holdings USA, Inc. as the recorded successor-in-interest.

Current status of the chain: Dragon Systems → acquired by Lernout & Hauspie Speech Products N.V. (all-stock, ~$580–593M, closed June 2000) → L&H Chapter 11 / Belgian bankruptcy, November–December 2000 → speech assets (including the Dragon/NaturallySpeaking line) sold at bankruptcy auction, approved 2001-12-11, to ScanSoft, Inc. for ~$39.5M → ScanSoft renamed Nuance Communications, Inc. (2005) → Nuance acquired by Microsoft (announced 2021, closed 2022). Dragon Systems as a standalone entity is dissolved; L&H is bankrupt and dissolved.

Assignment timeline

Chronological by execution/effective date. All entries below are recorded events appearing in the Google Patents legal-events block for US 6101468.

1. 2000-06-07 (executed) / recorded 2002-10-08 — Reel 013362/0732

   • Conveyance: Merger
   • Assignor: Dragon Systems, Inc.
   • Assignee: L & H Holdings USA, Inc.
   • Correspondent: not retrievable from the sources I could access — see gap note.
   • Context: Acquisition — Dragon absorbed into L&H Holdings USA as part of the Lernout & Hauspie all-stock purchase.

2. [reel/frame not retrieved] (executed ~2001-12-11 auction approval) / recorded 2002-10-08

   • Conveyance: Assignment of Assignors Interest
   • Assignor: L & H Holdings USA, Inc.
   • Assignee: ScanSoft, Inc.
   • Correspondent: not retrievable.
   • Context: Bankruptcy fire-sale — sale of L&H's speech/dictation assets out of the Delaware Chapter 11, court-approved 2001-12-11. Note the ~10-month lag between the sale and recordation, and that this and the merger were recorded the same day (2002-10-08) as a chain-of-title clean-up.

3. 2005-10-17 (effective) / recorded 2006-03-15 — Reel 016851/0772

   • Conveyance: Change of Name
   • Assignor: ScanSoft, Inc.
   • Assignee: Nuance Communications, Inc.
   • Correspondent: not retrievable.
   • Context: Change of name only — ScanSoft rebranded to Nuance. No change in beneficial ownership.

4. 2006-03-31 (effective) / recorded 2006-04-07 — Reel 017435/0199

   • Conveyance: Security Agreement
   • Assignor: Nuance Communications, Inc.
   • Assignee: USB AG, Stamford Branch (recorded spelling; widely understood to be UBS AG — I am quoting the record literally)
   • Correspondent: not retrievable.
   • Context: Securitization / collateral — blanket IP security interest securing Nuance's credit facility. Not an ownership transfer.

5. Recorded 2006-08-24 — Reel 018160/0909

   • Conveyance: Security Agreement
   • Assignor: Nuance Communications, Inc.
   • Assignee: USB AG, Stamford Branch
   • Correspondent: not retrievable.
   • Context: Securitization — second/grant collateral filing on the same facility (effectively the same lender-pledge structure as #4).

6. Recorded 2016-05-20 — Reels 017435/0199 and 018160/0909 (releases)

   • Conveyance: Patent Release (two entries — one per pledged reel)
   • Releasing party of record: Morgan Stanley Senior Funding, Inc., as Administrative Agent (i.e., the successor administrative agent on the secured facility)
   • Grantor group named in the release: the Nuance corporate family — including Nuance Communications, Inc., ScanSoft, Inc. (a Delaware corporation), Dictaphone Corporation, SpeechWorks International, Inc., ART Advanced Recognition Technologies, Inc., Telelogue, Inc., and others
   • Correspondent: not retrievable.
   • Context: Collateral release — clears the 2006 security interests. Post-dates the patent's 2012-11-13 expiry; releases were filed as part of a broader facility payoff.

Gap statement: The Assignment Center/assignment.uspto.gov record pages were not reachable during this analysis. Consequently (a) the reel/frame for entry #2 and (b) every correspondent-of-record field are unavailable. Per your own constraint, I am not substituting attorney names from general knowledge. If the correspondent fields are material to the call, this chain must be re-pulled directly at the Assignment Center.

Timeline diagram

timeline
    title Ownership of US 6101468
    1992 : Parent application filed
    1997 : CIP filed by Dragon Systems
    2000 : Dragon merges into LH Holdings
         : Patent issued
         : LH files Chapter 11
    2001 : LH assets sold to ScanSoft
    2002 : Merger and sale recorded
    2005 : ScanSoft renamed Nuance
    2006 : Security agreements to UBS AG
    2016 : Security interests released
    2022 : Microsoft acquires Nuance

NPE / troll-pattern signals

1. Shell-entity transfer — NOT PRESENT. The only recorded assignees are Dragon Systems, Inc.; L & H Holdings USA, Inc.; ScanSoft, Inc.; Nuance Communications, Inc.; and the secured lender USB AG, Stamford Branch. Every operating assignee is a product company. No "IP/Holdings/Ventures/Licensing" single-purpose LLC appears anywhere in reel 013362/0732, 016851/0772, 017435/0199, or 018160/0909.

2. Known asserter in the chain — NOT PRESENT. No assignee matches the supplied list (Acacia, Marathon, IV, IPNav, Wi-LAN, Mosaid/Conversant, Vringo, Pendrell, Innovatio, MPHJ, Lumen View, Round Rock, Document Generation, Spangenberg entities). Dragon/L&H/ScanSoft/Nuance are vendors; USB AG and Morgan Stanley Senior Funding are lenders/agents, which appear only as collateral holders, not owners.

3. Repeat correspondent across the chain — UNKNOWN / NOT DETERMINABLE. This is the one signal I cannot score, and I will not pretend otherwise: correspondent-of-record data was not retrievable from the sources available to me. No attorney or firm name is asserted here. Action item for a fuller pull: the recurring-firm test applies to three recordation clusters recorded by the same Nuance in-house/outside team — the 2002-10-08 pair (013362/0732 + the un-retrieved ScanSoft reel), the 2006 pair (017435/0199 and 018160/0909), and the 2016-05-20 release pair. If one correspondent appears across all three clusters, that is a Nuance corporate-IP-household pattern, not an NPE pattern.

4. Cascading transfers — NOT PRESENT (as defined). The test is multiple chained LLCs inside 24 months. Here there were three ownership events (Dragon→L&H 2000-06-07; L&H→ScanSoft 2001; ScanSoft→Nuance name change 2005-10-17) spread over ~5 years and driven by a merger, a bankruptcy, and a rebrand — corporate succession, not an engineered LLC ladder. The 2002-10-08 same-day double recording is a title clean-up, not a cascade.

5. Pre-litigation transfer — UNCLEAR. Google Patents flags "Family has litigation" for this family (Darts-IP link on the record page), indicating at least one suit somewhere in the family. I could not retrieve the docket, the parties, or the filing date, so I cannot test the "assignment within 6 months before first suit" window. The last ownership-changing event is 2005-10-17 (name change only), and the patent expired 2012-11-13 — 20 years from the 1992-11-13 priority date — so no post-2012 assertion of this patent is possible. Score as unresolved rather than as a negative.

6. Bankruptcy fire-sale — PRESENT (strongest signal in this chain). Lernout & Hauspie filed Chapter 11 on/about 2000-12-01 after an accounting fraud; the U.S. Bankruptcy Court for the District of Delaware approved the sale of the speech-technology assets — including the Dragon line — to ScanSoft on 2001-12-11 for ~$39.5M ($10M cash, a $3.5M note, ~$26M in ScanSoft stock). L&H had paid ~$580–593M in stock for Dragon roughly 18 months earlier. The recorded assignment from L & H Holdings USA, Inc. to ScanSoft, Inc. at the 2002-10-08 recordation date is the title step implementing that court-approved sale. This is a textbook distressed sale — but it moved assets between operating companies, not to an NPE.

7. Privateering — NOT PRESENT. No operating company transferred this patent to an NPE to assert against competitors. The acquirer (ScanSoft/Nuance) continued commercializing the very product the claims describe.

8. Defensive aggregator (anti-NPE) — NOT PRESENT. The chain does not terminate at RPX, AST, LOT, Unified Patents, or OIN. It terminates at Nuance Communications, Inc. (now a Microsoft subsidiary) subject to released bank collateral.

Additional lead, expressly not scored as a finding: search surfaced a recorded "Assignment Nuance to Voice Inventions" instrument executed effective 2021-01-01 (Nuance Communications, Inc., by David Greenbaum, VP Litigation & IP, assigning a portfolio of Nuance patents to a "Voice Inventions" entity). The portions I could retrieve list Nuance's own 7,3xx,xxx–9,7xx,xxx patents; US 6101468 does not appear on the visible list, and in any event this patent expired 2012-11-13, nine years before that instrument. I therefore do not place this patent in that transfer. Treat "Voice Inventions" as worth checking against RPX/Unified lists for other Nuance patents, not for this one.

Verdict

Operating-company assertion — with a stated caveat.

The entire recorded chain of title is a corporate-succession chain among product companies: Dragon Systems, Inc. → L & H Holdings USA, Inc. via merger effective 2000-06-07 (reel 013362/0732), → ScanSoft, Inc. via the court-approved L&H bankruptcy sale (recorded 2002-10-08), → Nuance Communications, Inc. via change of name effective 2005-10-17 (reel 016851/0772). The only other recorded items are Nuance's own bank pledges (reels 017435/0199 and 018160/0909) and their 2016-05-20 releases by Morgan Stanley Senior Funding as administrative agent — collateral, not ownership. Zero NPEs, zero shell LLCs, zero defensive-aggregator terminations across those reels.

Two caveats keep this from being a slam-dunk: (i) the chain contains a genuine bankruptcy fire-sale (L&H Chapter 11 → ScanSoft, ~$39.5M against a ~$580M purchase price), so "the assignee bought it cheap out of a bankruptcy" is true — it just went to an operating buyer, not an asserter; and (ii) the correspondent-of-record data was not retrievable, so the strongest remaining troll tell could not be tested. The patent expired 2012-11-13, which is dispositive for present-day assertion risk — whatever this portfolio's history, this particular patent cannot be asserted now.

Verify/re-pull at: https://assignmentcenter.uspto.gov/ (search 6101468) and https://patents.google.com/patent/US6101468/en (legal events).

I'll search for the specific patent at USPTO and for its cited references.

Let me get the full citation list for US 6,101,468 from the patent's front page and Google Patents.

Important up-front caveats (read before the analysis)

Three sourcing facts constrain what I can responsibly assert, and under the strict no-fabrication rule I am flagging them rather than papering over them:

1. The authoritative full text you supplied does not contain the claims, and it does not contain the front-page "[56] References Cited" list. It truncates in the DDWin discussion (around FIG. 46/DDWin ca…) before reaching the claims. So I cannot map cited references to specific numbered claims of US 6,101,468 with confidence. Any claim-number mapping below is a feature-level hypothesis, explicitly labeled as such.
2. The citation list I retrieved comes from a secondary mirror (PubChem's patent page), not from the patent PDF's own front page. PubChem lists document identifiers only, with no issue-date column. I could not retrieve the front-page date table, so the "publication date" column below is populated only where I verified it from another source; elsewhere it reads not retrieved rather than a guessed date.
3. Do not conflate the bare string "6101468" with other records. Searching the raw number returns, among other things, an unrelated OSTI record (OSTI ID 6101468 → US 4,815,431, "Oil heating apparatus for internal combustion engine," Nippon Soken — https://www.osti.gov/biblio/6101468). That is a different document; I have excluded it and am noting it precisely because the "no auto-correction" rule requires me to show the collision rather than bury it.

Also relevant: the USPTO/Google record confirms US 6,101,468 is a continuation-in-part claiming priority (1992‑11‑13) from US 07/976,413, which issued as US 5,428,707 (per the family/priority entry in your authoritative text: Priority claimed from US07/976,413 … patent/US5428707A/en). That gives two candidate critical dates — 1992‑11‑13 (original disclosure) and 1997‑06‑26 (actual filing / CIP new matter). This matters enormously for § 102, as shown below.

What the record shows for US 6,101,468 itself

Sources: https://patents.google.com/patent/US6101468/en (authoritative text supplied) and https://pubchem.ncbi.nlm.nih.gov/patent/US-6101468-A.

References cited in US 6,101,468 (as retrieved)

Source: https://pubchem.ncbi.nlm.nih.gov/patent/US-6101468-A (PubChem "Citations" section — mirrors USPTO bibliographic data). The list returned 47 US patent documents + 4 non-patent items. I am reproducing it verbatim-by-identifier; I did not independently verify that this is the complete front-page list, and PubChem does not flag which were examiner-cited vs. applicant-cited.

US patent documents cited

Non-patent literature cited

§ 102 analysis

Why I cannot give you "which claims it anticipates" — and what I can give instead

A proper § 102 map requires (a) the exact claim text of 6,101,468 and (b) each reference's disclosure. I have (b) only partially (identifiers, no verified disclosure summaries) and I do not have (a), because the authoritative text cuts off before the claims. Inventing claim numbers would violate the operating rules. So below I give a feature-cluster → candidate-reference hypothesis table, keyed to the invention clusters that the patent's own abstract attributes to its claims (abstract retrieved from PubChem; wording is the Office's own summary):

The two-date problem is the single most important § 102 issue

Because 6,101,468 is a CIP with a 1992 priority claim, each claim must be tested twice:

• Claims entitled to 1992‑11‑13: only art published/patented before 1992‑11‑13 can be § 102(a)/(b) art. Notably, none of the four NPL items qualifies (1994, 1994, 1994, and Word 2.0c), and none of the US patents listed that issued after 1992‑11‑13 qualifies unless relied on under § 102(e) as of their own earlier US filing date.
• Claims entitled only to 1997‑06‑26 (new CIP matter): the 1994 NPL items (Gliedman, Kurzweil guide, Lane) and the Microsoft Word 2.0c printout become available under § 102(a)/(b), as do the 1995–1997-issuing US patents (e.g., #44's siblings, #45–47).

Practical consequence: the references that look most dangerous in a citation list (the recent ones, #45–47) are ones the examiner already had before him and allowed over; the references that could actually knock out a claim are the new-matter claims, and the 1994 NPL items are the only 102(b)-eligible non-patent art in the whole list. That is where I would focus any invalidity investigation.

Same-family references that must be excluded from any prior-art list

US 6,101,468 sits in a large continuation family. These documents appear in the same Google Patents family/priority records and are not prior art to it:

• US 5,428,707 (parent, from US 07/976,413) — though it is cited on the face of the patent.
• US 5,850,627, US 5,900,666, US 5,915,236, US 5,920,836, US 5,920,837, US 5,960,394, US 5,983,179, US 6,073,097, US 6,092,043 — all appear as family members in the EPO search-report family table retrieved from WO 2006/044023 A1 and in the assignee's portfolio listing (GoodIP). Note in particular US 6,092,043, "Apparatuses and method for training and operating speech recognition systems" (granted 2000‑07‑18), a sibling of the patent-at-issue.

Under the no-auto-correction rule I am listing these as distinct documents, not as substitutes for 6,101,468.

Bottom line

• Verified: 6,101,468 cites at least 47 US patent documents and 4 non-patent items; the only reference I could confirm by title/date/assignee is US 4,783,803 (Dragon Systems, issued 1988‑11‑08). US 5,428,707 is the same-family parent and is not prior art.
• Not verified (and therefore not asserted): title, issue date, and disclosure of the other 46 patent references; and the full front-page citation set as it appears on the printed patent.
• Cannot be responsibly answered from the material available: a claim-by-claim § 102 anticipation mapping, because the authoritative text you supplied omits the claims.
• Highest-value § 102 targets, if any exist, are the four 1994/Word-2.0c non-patent items — and they only bite against claims that are not entitled to the 1992‑11‑13 priority date.

Verification steps to close the gaps

1. USPTO Patent Center / Patent Full-Text (PatFT) for 6,101,468 — pull the printed front page ([56] References Cited with dates) and the full claim set. This is the single missing input.
2. USPTO Patent Public Search — look up each of the 47 US references for title/issue date/assignee to build the date table above.
3. File wrapper (Patent Center "Documents") — identify Form PTO‑892/1449 to see which citations were examiner-cited vs. applicant IDS, and any intervening § 102 rejections, which would tell you exactly which references the examiner considered closest.
4. Family check — confirm the § 120 benefit chain to US 07/976,413 (now US 5,428,707) so the 1992 vs. 1997 date split is fixed claim-by-claim.

Confidence: high on the identity/date of US 6,101,468 and US 4,783,803, on the parent/priority relationship to US 5,428,707, and on the completeness caveat for the citation list. Low on descriptions/dates of the other 46 references and nil on claim-number-specific anticipation.

I'll research the claims and the cited prior art for US 6,101,468 before doing the §103 analysis.

I have the full claim set and the patent's own "References Cited" list. Let me now verify the content of the most probative cited references so the §103 grounds are grounded in what those references actually disclose.

I have the complete claim set (claims 1–13), the patent's full "References Cited" list (46 U.S. patents + 8 non‑patent items), and verified content for the two most probative references. Here is the §103 analysis.

Obviousness Analysis of U.S. Patent No. 6,101,468 under 35 U.S.C. § 103

Analyst note on scope and sourcing. The task directs me to use the Prior Art section of the patent page. The Google Patents full-text excerpt supplied to me did not contain the citation tables, so I reconstructed the "References Cited" list from the patent's own front-page data as mirrored at uspto.report and EveryPatent for US 6,101,468, plus the PubChem "Citations" list. Every reference I treat as prior art below is drawn from that list. Where I could not verify a reference's disclosure (as opposed to its existence, number, title, and date), I say so explicitly rather than characterizing it.

I. What is actually claimed (and therefore what must be obvious)

Despite an abstract and specification describing roughly ten distinct inventions, the claims of this divisional are confined to one of them: case‑insensitive alphabetic filtering of word recognition.

Two drafting consequences matter for §103:

1. Claim 1(e) is a string‑comparison limitation, not an acoustic limitation. It speaks of the "vocabulary word's associated spelling." The case‑insensitivity therefore lives in the text matching step, not in the acoustic model. That makes ordinary text‑search prior art directly on point and defeats any "different field" argument.
2. Claim 10 is drafted broadly and functionally ("logic … similar in certain ways, but need not be similar in other ways"). Under any reasonable construction it covers any partial/approximate string filter, which enlarges the prior art that reads on it (and arguably raises §112(b) questions, though that is outside this task).

II. Threshold issue that dominates the whole analysis: the effective critical date

This is where I must flag an assumption and a risk.

• US 6,101,468 (appln. 08/882,914, filed 1997‑06‑26) is a division of 08/382,752 (filed 1995‑02‑01), which is a CIP of 07/976,413 (filed 1992‑11‑13, issued as US 5,428,707). (Verified: parent-case text on the face of the patent, as mirrored at uspto.report/EveryPatent.)
• The case‑insensitive filtering disclosure appears only in the DDWin (Windows) portions of the specification. The patent says of the MS‑DOS choice list: "Although not disclosed, similar case-independent alphabetic filtering could be used with the choice list and Oops buffer of the MS‑DOS DragonDictate." That is an admission that the case‑independent feature was not in the earlier system.

Ramification: claims 1–13 (all of which require case‑insensitivity, directly or through dependencies) most likely receive an effective filing date of 1995‑02‑01, not 1992‑11‑13. The pre‑AIA §102(b) critical date would then be 1994‑02‑01.

That makes the §102 status of the cited art:

I flag this explicitly: I could not verify whether 08/382,752's specification and US 5,428,707 contain the case‑insensitivity disclosure. If 5,428,707 does disclose it, the critical date moves back to 1993‑11‑13 and the Sept./Oct. 1994 NPL items drop out entirely — but the Word 2.0c, Roszkiewicz, Schmandt, Tough and Roberts references survive either way, so the grounds below do not depend on the 1994 NPL.

Also note: pre‑AIA §103(c) cannot remove US 5,027,406 despite common ownership by Dragon Systems, because §103(c) disqualifies only art under §102(e)/(f)/(g). A §102(b) patent stays available regardless of common ownership. That is important, because US 5,027,406 is the single most damaging reference here.

III. Level of ordinary skill

A PHOSITA at the 1994–1995 critical date would have had a bachelor's degree in EE/CS (or equivalent) and ~2 years' experience with speech‑recognition user interfaces and interactive text‑editing/word‑processing software, or a master's degree with less experience. Critically, that person would have been familiar with both (i) large‑vocabulary discrete‑utterance recognizers with correction/choice lists (DragonDictate, Kurzweil Voice were shipping products) and (ii) contemporaneous word‑processor search/replace interfaces, including case‑sensitive vs. case‑insensitive "Find" (Word 2.0c, WordPerfect). This dual familiarity is the linchpin of the motivation analysis below.

IV. Ground 1 — Claims 1, 2, 4, 5, 8, 9, 10, 11

Combination: US 5,027,406 (Roberts et al.) in view of Microsoft Word 2.0c case‑insensitive "Finding Text" (as documented in the applicant's own IDS and the Word 2.0c help printout cited on the face of the patent).

US 5,027,406 is verified, and it is not merely close — it is the same inventors'‑assignee predecessor teaching the same feature, and it is cited on the face of the '468.

Claim chart for '468 claim 1:

So every element of claim 1 except the case‑insensitivity clause is literally in US 5,027,406, and the case‑insensitivity clause is a textbook case‑insensitive string compare.

The Word 2.0c evidence in the record is unusually strong because the applicant put it there:

"Statement contained in the Information Disclosure Statement for this application that text insensitive searching existed in Microsoft Word Version 2.0c, Copyrighted by Microsoft Corporation 1989‑1992." and "Printout of 'Finding Text' help window from Microsoft Word Version 2.0c, Copyrighted by Microsoft Corporation 1982‑1992."

Whatever "text insensitive" was meant to say, the accompanying help‑window printout of a "Finding Text" dialog is the classic Match‑Case toggle — i.e., a user‑selectable option under which a typed character string is matched against stored text "independently of case." Word 2.0c is undisputedly a printed publication and prior art under §102(b).

Motivation to combine (KSR‑consistent):

1. Same problem, same user population. 5,027,406 is expressly aimed at "a user with limited typing skills" who is dictating, and it uses typed letters only as a filter, never as content. A user typing a single filter letter has no reason to care about case; requiring case fidelity adds friction and error with zero functional benefit. Word 2.0c had already solved precisely that friction problem for text search.
2. Simple substitution of a known technique for a known technique. Replacing a case‑sensitive string comparison with the art‑recognized case‑insensitive string comparison (Word's default "Find" mode) is the paradigm of a predictable substitution; it changes no principle of operation of the recognizer, adds no new hardware, and yields the expected result.
3. The patentee's own articulated rationale is a convenience rationale, not a technical one. The specification justifies the feature as: "This enables the user to use alphabetic filtering to limit recognition to words starting with a desired letter without having to worry about whether the case of the letter he enters matches that of the letter in the desired word." A stated benefit that consists of relieving the user of an unnecessary input constraint, achieved by a known input convention, is very weak evidence of non‑obviousness.
4. Same field of use, and overlapping disclosures. The '468's own system types recognized words into, and corrects them inside, third‑party word processors. The Word 2.0c help printout is in the applicant's IDS precisely because the applicant itself considered it relevant.
5. Claims 2 and 4 add only that the word signals are acoustic and that alphabetic signals come from a keyboard — both expressly disclosed or rendered obvious by 5,027,406 (which teaches both typed and spoken entry of the filter string).
6. Claim 5 (alphabetic signals "of either upper or lower case") is met by the combination: the stored spelling is case‑capable and the Word 2.0c dialog accepts either case.
7. Claims 8 and 9 (limit the alternate‑choice group to spellings starting with the sequence, ignoring case; display the characters and the choices; let the user select) map onto 5,027,406 claims 20 (initial characters) and its claimed "menu of character sequences" with user selection/editing (cl. 1), plus its FIGS. 10–24 display-sequence disclosure — with case‑indifference again from Word 2.0c.
8. Claims 10 and 11 are the system‑claim mirror of claims 1 and 1(e) and are met by the identical combination; claim 10's "similar in certain ways, but need not be similar in other ways" language is broad enough that a case‑insensitive prefix filter literally satisfies it.

V. Ground 2 — Claim 6

Claim 6 requires (i) storing the typed sequence with its respective case and (ii) outputting either the recognized vocabulary word's spelling or that stored sequence, depending on conditions.

US 5,027,406 claim 1 requires "means for enabling the user to edit the selected character sequence to make it represent a different sequence of characters" and "means for associating said edited character sequence with said acoustic model"; and claim 7 states the output rule almost verbatim: "produces as its output for each spoken word either the word recognized or the word edited and selected for that spoken word."

A typed/edited character sequence is stored as characters with case inherent; and where the user's typed capitalization is not matched by any vocabulary word's spelling, outputting the user's own typed string (which is exactly what DDWin's ChoiceList step 1457 does — treating the StartString itself as the first choice word when no word matches the capitalization) is the natural fallback. Combination: 5,027,406 (output the edited string) + Word 2.0c ("no case match found" handling) + the ordinary design incentive to never lose the user's typed input. This is essentially a design‑choice claim; under KSR, where the prior art discloses both branches of an "either/or" output rule, selecting between them is within the skill of the art.

VI. Ground 3 — Claims 7 and 12 (the asymmetric case rule) — the most vulnerable claims

Claims 7 and 12 require the asymmetric rule: lower‑case input matches either case; upper‑case input matches upper case only. This is exactly the FIG. 55A alternate embodiment of the '468 (conditions 2118 and 2120), and the specification supplies the sole rationale: "if a user goes to the trouble of typing a letter in upper case, he or she probably wants that letter to be upper case."

• The universe of options is two: treat case symmetrically, or asymmetrically. This is the KSR "finite number of identified, predictable solutions" situation, and the art already contained both poles (Word 2.0c's Match‑Case on = case‑sensitive; off = case‑insensitive).
• The selection criterion is not technical but inferential about user intent — the "capital letter is the marked form" convention that any typist knows. Indeed, the examiner could take judicial notice of the capitalization convention (a user who shifts for a specific letter intends that letter capitalized), and the patent itself states the inference, demonstrating that it required no inventive insight.
• Secondary evidence would be required to save these claims, and the nexus problem is severe (see §IX).

I regard claims 7 and 12 as the strongest obviousness candidates and claims 1, 11 as the next strongest (because of US 5,027,406).

VII. Ground 4 — Claim 3

Claim 3 requires the alphabetic signals to be acoustic — i.e., spelled‑out letter words recognized by the recognizer. US 5,027,406 expressly discloses that "the user may type or speak the initial letters of the word," and includes claims covering entry of the filtering string by voice and even selection of a candidate filtering string by speech recognition against the same portion of speech (claims 18 and the dependent clauses in the '406 disclosure). The '468's own embodiment (International Communications Alphabet words, "a [alpha]," "b [bravo]" …) is one implementation of that idea.

Combination: 5,027,406 (spoken filter entry), alone or with a spelling‑alphabet teaching. This ground is comparatively strong on the entry mechanism, but note that claim 3 also inherits the case‑insensitivity of claim 1, so Word 2.0c or an equivalent case‑insensitive‑compare teaching should be added.

VIII. Ground 5 — Claim 13 (weakest link — escalate with caution)

Claim 13 requires that the "related strings" correspond to the start of the spelling of the pronunciation and need not correspond to the start of the word's spelling "when said spelling includes characters before the start of the spelling of said pronunciation." This corresponds to the '468's own "[Command Mode]"-style convention and to "3 [three]"/", [comma]" spellings, with the FIG. 55A condition 2122 that a match may begin immediately after a "[" character.

The record's cited art maps imperfectly:

• US 5,027,406 claim 19 is the broader "spellings which contain the string" filter; claim 20 is the "initial characters" filter. Neither expressly addresses a stored spelling whose leading characters are not pronounced.
• The Kurzweil Voice User's Guide, Rel. 1.0, pp. 60–61 (©1994) is cited on the face of the patent; Kurzweil Voice was a direct DragonDictate competitor of the same era with a correction/word‑list UI. I could not verify what pp. 60–61 disclose, and I will not assert that they teach bracketed pronunciations. If they do (Kurzweil‑style command words and punctuation/number‑name spellings are a plausible subject of those pages), claim 13 falls to 5,027,406 + Word 2.0c + that guide; if they do not, claim 13 is the claim most likely to survive.
• The Schmandt (1990) and Tough (1991) articles are relevant to speech input into window systems and command vocabulary, but I have not verified any bracketed‑pronunciation content in them.

Recommended evidentiary step: before asserting claim 13 in any challenge, retrieve the Kurzweil Voice User's Guide pp. 60–61 and the DragonDictate/L&H "command word" documentation; also consider any prior art teaching a word dictionary in which a stored entry contains non‑pronounced leading characters (e.g., number/punctuation name entries).

IX. Anticipated rebuttals, and how strong they are

A. "Different field — word processing vs. speech recognition."
Weak. Claim 1(e) compares the alphabetic signal to the vocabulary word's "associated spelling" — a string. The '468 system exists to insert text into word processors and to correct that text; the IDS itself placed Word 2.0c's "Finding Text" in the record. Also, under KSR, familiarity with the art of text searching would be imputed where the claimed advance is a text‑matching convention.

B. "No reference teaches or suggests ignoring case in a recognizer."
This is the patentee's best argument, and it is really an "absence of explicit motivation" argument rather than a teaching‑away argument. It weakens under KSR rationales (known‑technique substitution; two predictable options) and is undercut by the patent's own convenience‑only framing of the benefit.

C. Secondary considerations.
DragonDictate's commercial success and the industry's adoption of alphabetic "spell‑to‑correct" filters (in Dragon, IBM ViaVoice, Kurzweil, Philips) cut against the patentee on nexus: success in this art is overwhelmingly attributable to large‑vocabulary acoustic modeling, adaptive training and the Oops/choice‑list correction architecture — not to case‑folding a filter string. Absent a showing that the case‑insensitive feature drove purchases, secondary evidence is unlikely to carry these claims.

D. Prosecution history.
The claims were allowed. That is not, by itself, evidence of non‑obviousness, but the file history (especially the applicant's own IDS submissions and any examiner remarks on the Word 2.0c material) should be examined before relying on Ground 1: if the examiner considered and distinguished Word 2.0c, there will be a stated rationale to rebut.

X. Practical and procedural cross‑references to the earlier sections

• The litigation analysis in the prior section found no verified assertion of 6,101,468 and a family‑level Darts‑IP litigation flag (family 27009904). A §103 attack of the Ground‑1 type (own‑assignee predecessor patent cited on the face + case‑insensitive text search) is the natural invalidity theory a defendant would run if these claims were ever asserted, so the family flag remains the place to look for any actually‑litigated validity ruling.
• The patent expired 2012‑11‑13 (anticipated expiration, per the Google Patents record). IPR of an expired patent is not automatically barred, but the practical remedy is limited to cancellation of claims (no prospective injunctive value), and any §103 defense would most likely arise in a damages‑only or Section 337 posture.
• Contradiction check: none found. The previously generated section's bibliographic and chain‑of‑title data is consistent with the claim/continuity data I verified here (division of 08/382,752; CIP of 07/976,413; parent issued as US 5,428,707).

XI. Bottom line, by claim

XII. Explicit confidence and verification gaps

Verified: the complete claim set 1–13; the References Cited list; the continuity data; the disclosure of US 5,027,406 (abstract and claims 1, 2, 7, 8, 9, 16, 19, 20, including the "type or speak the initial letters … recognition algorithm is called again satisfying the initial letters" passage and FIG. 10‑24 display sequence); US 4,776,016 (Hansen, "Voice control system" — keyboard/voice control with a keyboard‑priority switch and a confirmation dialog; it does not disclose case‑insensitive alphabetic filtering, and I therefore do not rely on it); US 4,829,575 (Lloyd, a Viterbi/finite‑state‑machine engine usable in a speech recognizer — background only).

Not verified (do not rely on without checking): the content of US 4,823,306, 4,829,576, 4,990.../5,003,603/5,236,670‑type items whose titles I did not confirm; the content of the Kurzweil Voice User's Guide pp. 60–61; the content of the Gliedman, Lane, Schmandt, Tough and Roszkiewicz articles beyond their titles/dates; whether US 5,428,707 or application 08/382,752 discloses case‑insensitive filtering (this determines the critical date); the Word 2.0c "Finding Text" printout itself (I rely on the applicant's IDS characterization plus the reference's inclusion on the face of the patent); and the prosecution history of 08/882,914.

One caution on method: the "text insensitive" wording in the IDS entry, as transcribed in the sources I retrieved, may be a transcription artifact for "case insensitive." I have treated the Word 2.0c evidence as case‑insensitivity evidence because the accompanying "Finding Text" dialog printout is the artifact that supports it — but the actual printout and the Word 2.0c user documentation should be obtained before this ground is relied on in a filing.