US 5835892

I'll verify details with searches before summarizing.

Let me verify claim text and any litigation/assignment history.

Let me attempt to retrieve the actual claim language.

Patent summary — US 5,835,892

Important caveats up front. I could not query the USPTO PatentCenter/Assignment databases or the CAFC docket system directly; my checks were web searches against patent aggregators (Google Patents, Justia, Espacenet) plus general searches for litigation. The full patent text supplied to me is authoritative but truncated before the claims, so my claim descriptions below come from the specification's "Summary of the Invention" (which closely tracks the claim language) and one claim fragment I retrieved from Justia. Treat exact claim wording as unverified unless I flag it.

Bibliographic data (per Google Patents)

Abstract

A table of derivation elements and a state-transition table indicating applicable strings of derivation types are produced according to pronunciation expanding rules. Each derivation element is composed of (a) a derived sound derived from a key sound at a key position of a question pronunciation character string, (b) a sound position of the derived sound in strings expanded from the question string, and (c) one or more derivation types indicating how the derived sound is derived from the key sound. In the retrieving apparatus, strings of derivation types are produced one by one by arranging derivation types of the table in order of sound position, and each is tested against the applicable strings to determine whether it satisfies the pronunciation expanding rules. Trademark numbers corresponding to satisfying strings of derivation types are then retrieved. Because no similar-pronunciation character string is ever directly enumerated, retrieval is fast.

Plain-language overview of the independent claims

Based on the "Summary of the Invention," the patent appears to claim four core independent aspects (a method and apparatus pair for expansion, and a method and apparatus pair for retrieval), with additional independent claims later in the set (claim 32, discussed below). Descriptions are mine, not verbatim.

1. Similar character string expanding method (method claim). Steps: store similar-sound tables (each mapping a key sound to phonetically similar sounds); store sets of (regular expression + applicable range) derived from pronunciation similarity criteria; receive a question pronunciation character string; produce a set of derivation elements from the tables and regular expressions; produce a set of regular expression elements (regular expression + applicable sound number computed from the applicable range and the question string's sound count); build a finite state automaton representing the applicable strings of derivation types from those elements; output the derivation element set plus the automaton as the "expansion" result. The stated point of novelty: rules are declaratively expressed as regular expressions with numeric applicability ranges rather than as hand-coded procedures, so changing a similarity criterion does not require reprogramming.

2. Similar character string expanding apparatus (means-plus-function apparatus claim). Same substance in apparatus form: similar sound table storing means, regular expression storing means (regular expression + applicable range), question pronunciation character string receiving means, derivation element producing means, and finite state automaton producing means, with the automaton output serving as the stand-in for the (potentially millions of) expanded character strings.

3. Character string retrieving method. Uses the derivation elements (where each derivation element carries a derived sound, a sound position, and a derivation type encoding both the derivation and the positional difference from the remarked sound) and the finite state automaton. Steps: produce possible strings of derivation types by arranging derivation types in sound-position order; test each against the finite state automaton to get "probable" strings; test those against the applicable sound number to get "definite" strings; assemble the corresponding derived sounds into definite character strings; retrieve matching registered character strings from a database.

4. Character string retrieving apparatus. Means-plus-function counterpart: finite state automaton storing means, derivation element set storing means, derivation type string producing means, derivation type string judging means, and character string retrieving means — retrieving registered (e.g., trademark) character strings matching the definite character strings from the database.

5. Claim 32 — "similar character string expanding and retrieving apparatus" (fragment retrieved from Justia). This combines the expansion and retrieval functions into a single claimed apparatus, and notably adds a selectivity feature: the derivation element producing means selects particular regular expressions whose applicable ranges encompass the question string's sound count, and determines an applicable sound number per selected expression. Per Justia, claim 31 (dependent) recites the search-tree traversal mechanics (replacing the final derivation type, decrementing the maximum sound position, adding new derivation types for a newly defined maximum sound position, and stopping when all derivation types for the first position are arranged). I have not verified the full numbering or scope of claims 5–31.

Litigation / CAFC check (negative result)

Searches for USPTO and CAFC 2026 activity tied to 5,835,892 returned no proceedings for this number — no IPR/PGR, no Federal Circuit appeal, no district court case keyed to this patent in the results. Given (a) the patent expired in 2016 and (b) my search was not a direct docket query, I want to be explicit: this is a negative search result, not a certified "no litigation exists" finding. The unrelated hits returned (e.g., Headwater v. Verizon, Spero v. Volkswagen) are different patents and are not relevant.

Practical / technical takeaway

The invention targets Japanese-language trademark (and similar) searching, where many phonetically similar katakana variants of an application mark must be matched. Rather than expanding a question string into all similar pronunciations and looking each up (the FIG. 1–4 prior art), it represents similarity criteria as regular expressions over "derivation types," compiles them into a finite state automaton, and guides a database search with derivation elements. The examiner-facing classifications (G06F16/90335, G06F16/90344) frame it as query processing by string matching. From a modern § 101 perspective it would likely raise abstract-idea questions, but I have no record of any such challenge being brought.

Sources: https://patents.google.com/patent/US5835892/en (bibliographic data, abstract, description); https://patents.justia.com/patent/5835892 (claim 31–32 fragment); http://data.epo.org/pise-server/rest/collections/lgpi/EP1380966A3.pdf (EP family member); https://patents.google.com/patent/[US5495409A](/patent/US5495409A)/en (same inventor — Yuji Kanno / Matsushita assignment record REEL/FRAME 007324/0238, indicating a related patent family context, not this patent's own assignment record).

I'll research litigation involving US patent 5835892 across patent litigation databases.

Let me search more targeted litigation databases.

Let me query dedicated litigation databases directly.

Litigation Search Report — U.S. Patent No. 5,835,892

Patent identified: U.S. Patent No. 5,835,892 ("Method and apparatus for expanding similar character strings"), inventor Yuji Kanno, original assignee Matsushita Electric Industrial Co., Ltd. (now Panasonic Holdings Corp.), filed 1996‑04‑03, granted 1998‑11‑10, expired (fee‑related) as of the 2016‑04‑03 anticipated expiration.

Result: No litigation identified

Based on my searches of publicly indexed patent‑litigation sources, I found no record of any litigation (district court, ITC, or Federal Circuit appeal) involving U.S. Patent No. 5,835,892. This means either (a) the patent was never asserted in a court action that is publicly indexed, or (b) any such action is not captured by the sources reachable in this search. I am flagging this as an absence of evidence, not proof of non‑existence — I could not complete a fully authoritative docket‑level verification (see limitations below).

Important caveats on false positives (strict ID interpretation)

The searches surfaced several items containing the digit string "5835892" that are NOT this patent and must not be conflated:

Litigation‑database sources consulted

• Targeted web searches for the patent number combined with litigation terms (lawsuit, infringement, docket, CAFC/PACER, Unified Patents litigation portal).
• CourtListener / RECAP‑indexed opinions and dockets.
• General patent docket and PTAB‑related results.

None returned a case in which US 5,835,892 is named as an asserted patent. Related "Matsushita/Panasonic as defendant or plaintiff" cases that appeared (e.g., Samsung SDI Co. v. Matsushita Electric Industrial Co.) concern entirely different patents and technologies.

Limitations / recommended verification

I was unable to run persistent, account‑gated queries against Docket Navigator, Lex Machina, or the USPTO Patent Litigation Docket Search within this session, and my targeted searches hit a tool step limit. To close the gap with certainty, the following should be checked directly:

1. USPTO Patent Litigation Docket Search (patent number 5835892).
2. PACER / CourtListener RECAP docket search by patent number.
3. Unified Patents litigation portal (portal.unifiedpatents.com/litigation/caselist) — note: the digit‑string hits there in my results related to Unified Patents IPRs on other patents (e.g., 9,357,217; 10,621,228), not the '892 patent.
4. Federal Circuit docket search by patent number.

If, upon those direct queries, no case appears, the correct conclusion is that no known litigation exists for U.S. Patent No. 5,835,892, consistent with its being a 1990s-era Japanese‑language text‑retrieval patent that expired around 2016 and shows no publicly indexed assertion history. I did not find a plaintiff, defendant, jurisdiction, case number, filing date, or outcome because no case was identified.

Proceedings overview

Total AIA trial proceedings on file for US 5,835,892: 0 — the USPTO Open Data Portal ("PTAB proceedings on file") returns no IPR, PGR, or CBM proceeding naming this patent, so the breakdown is 0 active / 0 claims invalidated / 0 claims sustained / 0 settled / 0 institution denials. All claims (1 through the last issued claim) are untested at the PTAB — there is no Final Written Decision, no institution decision, no termination, and no claim that has been canceled in an AIA trial. Bottom line for a defendant: there is no PTAB invalidation record to lean on, but there is also no petitioner estoppel and no hardened, IPR-surviving claim set to worry about. The defensive posture here is not "the patent survived and is hardened" nor "claims 1–5 are canceled"; it is "the patent was never challenged at the PTAB, and it is now expired as of 2016-04-03."

No proceedings to report

I searched and found no AIA trial proceeding — no Inter Partes Review, Post-Grant Review, or Covered Business Method review — naming U.S. Patent No. 5,835,892 as the challenged patent.

Because the structured source lists zero proceedings, there are no proceeding sections to populate: no proceeding number, no petitioner, no patent owner, no filing date, no judge panel, no institution decision, no FWD, no settlement, and no appeal can be reported. Inventing any of those fields would be fabrication. Explicitly:

• Type / Filed / Status / Panel / Grounds / Institution / FWD / Settlement / Appeal: not applicable — no proceeding exists.
• Claim-level outcome: none. No claim of US 5,835,892 has been canceled, confirmed, or held unpatentable by the PTAB. I am not going to attribute a claim disposition to this patent that no decision supports.

Why this absence is credible rather than a search artifact

This is not merely a tooling gap; several structural facts make "zero PTAB activity" the expected result:

1. Timing. The AIA trial regime opened 2012-09-16. This patent issued 1998-11-10 and reached its anticipated expiration on 2016-04-03 (Google Patents legal-status field: "Expired – Fee Related"). That leaves roughly a 3.5-year window in which the patent was both live and IPR-eligible — and any IPR filed would have had to be worth prosecuting against a patent with a very short remaining life.
2. Subject matter. The patent is a Japanese-language (katakana) pronunciation-expansion and trademark-retrieval patent (see the specification's similarity criteria at FIG. 3 and expanding rules at FIG. 4). It is not a "covered business method" candidate under AIA § 18, so CBM is essentially off the table, and PGR was unavailable for a 1998-issued patent.
3. No assertion history to trigger a defensive filing. The companion litigation section of this analysis found no district court, ITC, or Federal Circuit case asserting this patent. IPRs overwhelmingly follow assertion or competitive threat; a never-asserted, near-expiry, Japanese-language text-retrieval patent attracts neither.

False positives I encountered and am rejecting

Per the strict-identifier rule, the following surfaced in search and are not proceedings on US 5,835,892:

Limitations / recommended direct verification

My web searches did not return a docket-level confirmation from PTAB E2E or the ODP trials endpoint; the conclusion rests primarily on the structured "no AIA trial proceedings" feed plus the negative search results above. To close the last gap:

1. PTAB E2E (https://e2e.uspto.gov/) — search by patent number 5835892, all trial types (IPR/PGR/CBM/DER).
2. USPTO Open Data Portal trials/proceedings API (https://data.uspto.gov/ and the API catalog at https://developer.uspto.gov/api-catalog) — query by the patent number directly.
3. Docket Navigator / Lex Machina / Unified Patents litigation portal — filter by patent number 5,835,892 across PTAB and district court.
4. CourtListener / RECAP (https://www.courtlistener.com/) — search 5,835,892 and 5835892 as CAFC docket text.

If those return nothing, the correct conclusion is no known PTAB activity on US 5,835,892, which is exactly what the structured feed reports.

Strategic summary

Claim status. No claim of US 5,835,892 has been canceled, confirmed, amended, or otherwise adjudicated at the PTAB, because no AIA trial was ever instituted. Every claim is best characterized as UNTESTED — and now EXPIRED. There is therefore no "surviving claim set" narrowed by IPR to report; the patent simply ran out. For a defendant, this means the usual IPR-based playbook (point to a cancelled independent claim and tell the plaintiff their theory is dead) is unavailable — not because the claims are strong, but because nobody ever tried. The invalidity case, if any, has to be built from scratch in district court or via ex parte reexamination (though reexam of an expired patent offers the requester no ongoing freedom-to-operate benefit, only a retrospective clean-up of asserted past damages).

Estoppel landscape. § 315(e)(2) estoppel is person-specific: it bars only a petitioner (and its real parties in interest/privies) from raising in a civil action grounds it raised or reasonably could have raised in an instituted IPR. With zero petitioners, there is zero estoppel on this patent. Concretely for a defendant being asserted today: every prior-art ground is still available — § 102 and § 103 combinations in any forum, unconstrained by anything any other party argued at the PTAB. Conversely, you also have no FWD findings (no adopted claim construction, no "teaching away" findings, no objective-indicia determinations) to borrow from.

Pattern signals. There is no pattern to detect: no repeat petitioner, no joinder, no defensive aggregator (no Unified Patents or RPX filing visible), no patent-owner appeal activity, and no PTAB-to-CAFC chain. Combined with the litigation analysis finding no asserted case, the picture is that of a dormant, expired, never-contested Japanese-language text-retrieval patent from Matsushita's 1990s portfolio, now held by Panasonic Holdings Corp. That is the opposite of the profile of a patent that has "survived two IPRs and is hardened."

One timing caution that cuts in the defendant's favor. Because the patent's anticipated expiration is 2016-04-03, and the fee-related status is "Expired," any damages theory is necessarily limited to the statutory look-back on past infringement (max six years pre-complaint) and cannot reach post-expiration conduct. That is a stronger defensive datum than anything a PTAB record would give you here.

Recommended next steps

1. Do not expect a PTAB silver bullet. There is no PTAB activity on file for US 5,835,892 — plainly stated. The absence is itself a signal: this is not a well-asserted, IPR-magnet patent; it is an expired, uncontested one. Assertion of an expired 1998 Japanese-language retrieval patent should be scrutinized for whether it can support any viable damages theory at all.
2. Confirm the docket-level negative before you rely on it. Run PTAB E2E (https://e2e.uspto.gov/) and the ODP trials endpoint for patent number 5835892, and a CourtListener/RECAP search for 5,835,892 and 5835892, to verify both the "no PTAB" and "no litigation" conclusions at the source.
3. Check terminal disclaimer / maintenance-fee history. The status "Expired – Fee Related" should be verified against USPTO Patent Center (https://patentcenter.uspto.gov/) to pin down the exact lapse date, since it governs the outer bound of any damages window. Patent Center is also the source for any ex parte reexamination or reissue having been filed (none surfaced here).
4. If a demand letter arrives, attack the damages premise. With no IPR record to cite and no post-2016 damages available, the highest-value defensive questions are (a) when exactly the patent lapsed for non-payment, (b) what pre-lapse, pre-complaint acts are alleged, and (c) whether the accused functionality even maps to the FIG. 3 similarity criteria / FIG. 4 rules as claimed. I found no FWD to quote and no canceled claim to link, and I will not manufacture either.

Verification links used/available:

• Patent text (authoritative): https://patents.google.com/patent/US5835892/en
• PTAB E2E: https://e2e.uspto.gov/
• USPTO Open Data Portal: https://data.uspto.gov/
• CourtListener: https://www.courtlistener.com/
• USPTO Patent Center: https://patentcenter.uspto.gov/

Nothing in this section should be read as asserting that any claim of US 5,835,892 has been invalidated. No PTAB proceeding exists, so no such finding exists.

I'll research the assignment record for US 5835892, starting with targeted searches on the ownership chain.

Let me dig for the specific assignment reel/frame data and the Matsushita→Panasonic name-change record.

I found a lead on a recorded name-change. Let me pull details on that assignment and hunt for the original 1996 inventor assignment.

Ownership & Assignment Analysis — U.S. Patent No. 5,835,892

Scope note on sourcing. I worked from the authoritative full text/front-matter supplied for US 5,835,892 (Google Patents legal-events block) plus indexed mirrors of the USPTO assignment record (PlainSite's rendering of USPTO Patent Assignment data). I could not open the Assignment Center's own result page or the USPTO Patent Assignment Search API within this session (tool step limit), so where a reel/frame or correspondent was not surfaced by the indexed sources, I say so explicitly rather than inventing one.

Inventors

• Only one inventor is named. The application was filed 1996-04-03; the assignment to Matsushita was recorded 1996-05-28 (≈8 weeks post-filing).
• No unusual departure pattern determinable. The record shows a single, ordinary employee-invention assignment executed contemporaneously with filing. There is no evidence of the inventor departing the assignee within 12 months, no second inventorship-level transfer, and no subsequent inventor-side assignment. I found nothing suggesting a portfolio fire-sale trigger.

Original assignee

• Entity on the issued patent: Matsushita Electric Industrial Co., Ltd. (1006 Oaza Kadoma, Kadoma-shi, Osaka, Japan).
• Primary line of business: diversified consumer/industrial electronics; best known for the Panasonic and (in Japan) National brands. It is the company founded by Konosuke Matsushita in 1918.
• Product embodying the claims: Not identified. US 5,835,892 is a method/apparatus patent for expanding pronunciation-similar character strings and retrieving registered names (trademarks, compound names, place names, etc.) from a database. It is a text-retrieval / Japanese-language-processing technology of the type used internally in trademark-search and Japanese word-processor systems. No commercial product line is tied to it in any source consulted, so I cannot state that a shipped product embodies the claims.
• Current status: Operating. The entity was renamed Panasonic Corporation effective 2008-10-01 (approved at the June 2008 ordinary general meeting; see Panasonic's own press release). It was subsequently reorganized: effective 2022-04-01, Panasonic Corporation became Panasonic Holdings Corporation (a holding company) with operating businesses moved into a new Panasonic Corporation subsidiary. This matches the Google Patents "Current Assignee: Panasonic Holdings Corp." field. No insolvency, bankruptcy, or asset-sale event involving this patent was found.

Assignment timeline

The chain is remarkably short — one execution-and-recordation event, plus a housekeeping corporate name change. There is no arms-length sale anywhere in the record.

• 1996-04-03 (filing) / 1996-05-28 (recorded) — Reel/frame not surfaced in the sources I could reach (the Google Patents legal-events block reports the assignment but does not expose the reel/frame, and I could not open the Assignment Center result page)

  • Conveyance: Assignment of Assignors Interest ("ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS)")
  • Assignor: Yuji Kanno
  • Assignee: Matsushita Electric Industrial Co., Ltd.
  • Correspondent: not surfaced. Whether the recording correspondent recurs elsewhere on tracked patents could not be assessed from available data; a single appearance would not be a finding regardless.
  • Context: ordinary employee-invention assignment — the standard pre/post-filing transfer from an inventor to their employer; not a sale, not a securitization.

• 2008-10-01 (executed, per the recording) / recorded 2017-11-02 — USPTO Patent Assignment Reel 45312 / Frame 937

  • Conveyance: Change of Name
  • Assignor: Matsushita Electric Industrial Company, Ltd.
  • Assignee: Panasonic Corporation
  • Correspondent: not surfaced in the indexed rendering (PlainSite id=9585705).
  • Context: change of name only — a portfolio-level record reflecting the 2008-10-01 rename. ⚠️ Caveat: the indexed entry shows a bulk name-change covering numerous Matsushita/Panasonic patents (the specimen listings I could see were unrelated radio-communication cases). I could not confirm from available data that US 5,835,892 is specifically listed within Reel 45312/937 — the recording is real and the renaming is documented, but the specific inclusion of this patent number is unverified. Note the long gap between execution (2008) and recordation (2017), which is characteristic of these retroactive bulk name-change filings.

• 2022-04-01 (corporate reorganization; no separate assignment instrument surfaced) — Panasonic Corporation → Panasonic Holdings Corporation

  • Context: internal reorg / change of name reflected in Google Patents' "Current Assignee" field. I did not find a distinct recorded assignment instrument for this step tied to this patent.

No other assignment, security agreement, license, merger, or release appears in the record for this patent. After the 1996 employee assignment and the corporate renames, ownership stayed inside the same Japanese operating-company family throughout.

Bulk-recordation limitation: because the 2008 name change was filed as a mass recording, any per-patent reel/frame attribution (if one exists) should be confirmed against the Assignment Center's property-by-patent view.

Timeline diagram

timeline
    title Ownership of US 5835892
    1996 : Filed and assigned to Matsushita
    1998 : Patent issued as US 5835892
    2008 : Matsushita renamed Panasonic Corporation
    2016 : Patent expired for non-payment
    2022 : Panasonic Holdings becomes assignee of record

NPE / troll-pattern signals

Verdict

Insufficient data (per the rubric: only the original assignment, plus a corporate name change).

Justification: The entire recorded chain is (i) the inventor-to-employer assignment recorded 1996-05-28 (assignor Yuji Kanno → Matsushita Electric Industrial Co., Ltd.) and (ii) a portfolio-level Change of Name from Matsushita Electric Industrial to Panasonic Corporation (USPTO Patent Assignment Reel 45312/937, executed 2008-10-01, recorded 2017-11-02), with the 2022 Panasonic Holdings reorg reflected only in the "Current Assignee" field. There is no arms-length sale, no NPE, no security interest, and no defensive-aggregator termination — the patent stayed with the original operating-company family until it expired for non-payment around 2016-04-03. On the evidence available, the practical conclusion is that this is an operating-company-held, never-asserted patent with zero NPE signals; because the record contains no post-issuance third-party transfers to analyze, I cannot classify it as an NPE chain even in the negative with the highest confidence.

Verification links:

• USPTO Assignment Center (search by patent number 5835892): https://assignmentcenter.uspto.gov/
• USPTO Patent Assignment Search (legacy index, same underlying records): https://assignment.uspto.gov/patent/index.html
• Google Patents legal events for US 5,835,892: https://patents.google.com/patent/US5835892/en
• Indexed rendering of the name-change record (PlainSite id=9585705): https://www.plainsite.org/patents/assignment.html?id=9585705

Items to confirm directly at the Assignment Center: (a) the reel/frame and execution date of the 1996 Kanno→Matsushita assignment, and (b) whether US 5,835,892 is enumerated within Reel 45312/937 or whether a separate per-patent name-change frame exists.

I'll research the USPTO record and the cited prior art for US 5,835,892.

The cited-references section wasn't returned. Let me query dedicated patent records directly.

Let me try the patent PDF front page and the EP family search report.

I've hit the research-tool step limit, so I need to stop and report what I could and could not verify. I'm going to be explicit about a significant gap rather than fill it with invented citations.

Prior-Art Analysis — U.S. Patent No. 5,835,892

1. Confirmation of the target patent (strict ID check)

I found no USPTO/PTAB, district court, ITC or Federal Circuit proceeding keyed to this number (negative result, consistent with the earlier litigation section).

2. The critical limitation on this task — I could not retrieve the "References Cited" front page

This is the honest headline: the authoritative full text supplied to me is truncated before the claims AND before the front-page "References Cited" block. I am therefore unable to enumerate the actual U.S. Patent Documents / Foreign Patent Documents cited on the face of 5,835,892, and I will not fabricate a citation list.

My retrieval attempts all failed to surface the citation block:

• Google Patents results returned only the description text (FIGS. 1–4 narrative), not the citation table.
• Espacenet / EPO searches returned other patents' search reports (EP 1,705,413; EP 990,709; EP 1,440,657; WO 2004/026128; WO 2013/101581) — none of these relate to '892 and must not be conflated with it.
• The EPO family search-report route (via EP 0749076, the European sibling) was not reached before the tool step limit.

Because you asked me to interpret IDs literally and not invent them, I am reporting a verified gap instead of a plausible-looking list. Recommend direct retrieval from: USPTO Patent Public Search / PatentCenter for 5,835,892 (full document images include the "References Cited" block), or the Google Patents "Patent Citations" API endpoint.

3. The only references actually named in the authoritative text I do have

The specification cites exactly two non-patent-literature items (in Section describing the regular expression, immediately after FIG. 14):

Neither is a § 102 reference to the claims as I understand them; both are admitted background for terminology.

4. Prior art described within the specification (the "conventional" art)

The specification itself characterizes the following as well-known conventional systems. These are the figures that matter for any § 102/§ 103 mapping, even though they are described rather than formally listed as cited documents:

5. Related-family subject matter (context, not face-citations)

6. How the § 102 analysis should map once the real citations are in hand

Because I can't see the actual citation list, here is the element-by-element framework keyed to the reconstructed claim scope from the earlier summary — so that when you pull the real face-citations, each can be tested against these elements:

Independent claim 1 — expanding method requires, as a combination: (a) storing similar-sound tables mapping key sounds → phonetically similar sounds; (b) storing regular-expression + applicable-range pairs derived from similarity criteria; (c) receiving a question pronunciation string; (d) producing derivation elements (derived sound + sound position + derivation type); (e) producing regular-expression elements with an applicable sound number computed from the question string's sound count; (f) producing a finite state automaton over the applicable strings of derivation types; (g) outputting the element set + automaton as the expansion result.
→ To anticipate, a reference must disclose all of (a)–(g) in one document, including the declarative regular-expression encoding with a numeric applicability range. FIG. 1's per-rule procedural units fail element (b)/(f).

Independent apparatus claim — expanding apparatus: same elements in means-plus-function form; must disclose the five means including regular-expression storing means and FSA producing means.

Independent claim — retrieving method/apparatus requires: derivation elements encoding both derivation and positional difference from the remarked sound; producing possible strings of derivation types in sound-position order; testing against the FSA to yield "probable" strings; testing those against the applicable sound number to yield "definite" strings; assembling derived sounds into definite character strings; and retrieving matching registered strings from a database.
→ FIG. 11's admitted art (expand-everything-then-index-each) fails the "derive strings via FSA over derivation types without enumerating similar strings" limitation, but any ordinary TRIE/database-lookup reference could combine with it under § 103.

Claim 32 (combined expanding-and-retrieving apparatus, with selectivity): adds selecting particular regular expressions whose applicable ranges encompass the question string's sound count, and determining an applicable sound number per selected expression.
→ This selectivity feature is the narrowest and most likely to survive; a § 102 reference must disclose the range-based expression selection step, not merely range-based applicability.

7. Bottom line

• Verified: the two NPL books (Hopcroft/Ullman 1969; Salomaa 1973), cited only as background for regular expressions; and the internally-described conventional systems of FIGS. 1, 5 and 11 as admitted prior art.
• Not verified / could not retrieve: the actual U.S. Patent Documents and Foreign Patent Documents listed under "References Cited" on the face of 5,835,892 — the source text I was given is truncated before that section, and my searches returned unrelated documents.
• No fabricated citations were produced. To complete a defensible § 102 chart, retrieve the front page of 5,835,892 from USPTO Patent Public Search or PatentCenter (document images include the citation block), then test each cited reference against elements (a)–(g) above.
• Note the patent has been expired since 2016-04-03, and no litigation was identified, so this prior-art analysis is of historical/validity-record value rather than live-exposure value.

I'll first try to recover the actual prior-art citations (front-page references, "Cited By," and family) for US 5,835,892, since the supplied page text was truncated before the patent's Citations/Prior Art section.

Let me pull the full backward-citation list for US 5,835,892 and related family/same-inventor art.

Let me try to nail down the complete front-page reference list and the Japanese priority/family art.

Obviousness Analysis — U.S. Patent No. 5,835,892 under 35 U.S.C. § 103

0. Scope, ground rules, and a flagged inconsistency

Date inconsistency (flagged per operating rules). The task header states "Current Date: April 26, 2026," while my operating context states today is 2026‑09‑29. I cannot reconcile these two values; I note it and proceed. The discrepancy does not change the legal analysis, because the patent's pre‑AIA § 103 status is fixed by its 1996‑04‑03 filing date.

Contradiction check against prior sections. The previously generated summary said the supplied patent text is "truncated before the claims." The full text supplied for this task is likewise truncated — it ends mid‑sentence in the Description (at "An example of t"). So the previous caveat stands and is not contradicted. Exact claim wording for claims 1–32 remains unverified. My element decomposition below is derived from (a) the "Summary of the Invention" section of the specification, which the drafter wrote to track the claims, and (b) the claim 31/32 fragment previously retrieved from Justia. Any element‑by‑element conclusion is therefore provisional pending full claim text.

Prior art source. The task directs me to "the results from the Prior Art section of this page." The page text as supplied was truncated before Google Patents' "Citations / References Cited / Cited By" sections, so I attempted to reconstruct that section by search. I hit a tool step limit before completing the reconstruction. What I could verify is set out below, with confidence flags. Where I could not verify a title or number, I say so rather than guessing.

I. The prior art of record (as reconstructed)

Not prior art / must not be used: the [PA] PAJ abstract above; the JP patent JP 5835892 B2 (Canon, charged‑particle beam drawing) confusion item raised in the prior litigation section — it remains irrelevant here; and any post‑1996 references (e.g., the Amazon/Microsoft/Nuance forward citations from the "Cited By" list), which are only useful as evidence of the ordinary skill level or of the art's later trajectory, not as § 102/§ 103 art.

II. Claim scope to be met (element decomposition)

Because the claims are not in the record, I use the Summary of the Invention. Pre‑AIA § 103 applies (filing 1996‑04‑03). The independent claims reduce to five functional clusters:

The single point of novelty over the applicant's own admitted art is structural, not functional: the same similarity criteria that FIG. 4 expressed as procedures are, in the claims, expressed as regular expressions over derivation‑type tokens plus numeric applicability ranges, compiled into an FSA, and the enumerated expansions are replaced by the (derivation‑element table, state‑transition table) pair. Elements A, C, and F's end result are all squarely in the admitted prior art.

III. Legal framework

1. Graham v. John Deere Co., 383 U.S. 1 (1966) — scope/content of the prior art; differences; PHOSITA level; secondary considerations.
2. KSR Int'l Co. v. Teleflex Inc., 550 U.S. 398 (2007) — a combination of familiar elements according to known methods is obvious when it yields no more than predictable results; the "motivation" need not be in the references themselves and may come from ordinary creativity, the nature of the problem, or the market.
3. MPEP 2129 — statements in the applicant's own specification describing the prior art are admissions usable as prior art. This is the single most powerful lever here: the '892 specification concedes the entire FIGS. 1–11 pipeline.
4. Pre‑AIA § 103(a); § 103(c) may disqualify the same‑assignee US 5,495,409 as art if commonly owned at the time of invention (likely, both Matsushita) — so I treat it as corroboration of the state of the art rather than as a necessary reference.

IV. The obviousness case

Combination I (primary — the strong one): Admitted prior art (FIGS. 1–11) + Hopcroft & Ullman / Salomaa

Differences to be overcome: (i) representing expanding rules declaratively as regex‑over‑derivation‑types with numeric ranges instead of as procedures; (ii) compiling those regexes into an FSA; (iii) emitting a derivation‑element table + state‑transition table as the stand‑in for the enumerated list of similar strings.

Why the combination is obvious:

• Every substantive function is admitted. The conventional apparatus 101 already (a) applies plural pronunciation similarity criteria (FIG. 3), (b) produces, from a question pronunciation character string, the set of similar‑pronunciation strings (FIG. 4 rules 1–5, units 102–106), (c) removes duplicates (unit 108), and (d) retrieves against a TRIE index (FIGS. 10–11). Substituting a different internal encoding for the very same criteria does not change what the system does to the outside world.
• The similarity‑table concept (element A) is inherent in the admitted art. FIG. 2's Japanese syllabary and FIG. 3's "weak sounds na/nu/ru" and "same‑line sounds" criteria are equivalence classes over key sounds — precisely a similar‑sound table in tabular form. Nothing more than a data‑structure formalization of FIG. 2/FIG. 3 is required.
• Regular expressions and regex→automaton conversion were textbook knowledge at the priority date. Both Hopcroft & Ullman (1969) and Salomaa (1973) are cited in the specification itself as the source of the formalism. Kleene's theorem and the subset construction are decades old. Using a regular expression to denote a pattern of symbol sequences and mechanically compiling it into a recognizer is the paradigm "known technique used for its known purpose."
• Motivation, on the applicant's own stated problem. The specification states the drawback of the prior art as: (1) rules must be "manually transformed into one pronunciation expanding rule programmed in a procedure form," so alteration of a criterion requires reprogramming; and (2) enumerating millions of expansions is impractical for memory and time. The invention's motivation is therefore supplied by the specification itself, and the solution — declarative rule tables + compiled automaton + lazy generation — is the standard remedy a POSITA would reach for. Under KSR this is "a predictable use of prior art elements according to their established functions."
• Reasonable expectation of success. Automaton construction from a regex is a mechanical, deterministic algorithm; determinization is well‑defined. There is no experimental uncertainty of the kind that defeats obviousness.
• Even the indexing side already used an automaton. The admitted TRIE index (FIG. 10) traverses nodes on sounds in sequence — functionally a deterministic recognizer. A POSITA who already used an automaton to look up strings would naturally use an automaton to constrain generation of strings, especially when the generation step was the bottleneck.

Bottom line on Combination I: claims 1 and 2, and the broad retrieval claims 3 and 4, are, on this record, likely obvious. The named invention is essentially (admitted similarity expansion) + (textbook regex/FSA compilation) with no asserted unexpected result.

Combination II (for the domain‑limited/trademark claims): Combination I + US 3,952,184 (Sodema)

US 3,952,184 discloses apparatus for automatically classifying and finding groupings of distinguishing signs according to the risk of conflict — i.e., automated trademark‑conflict screening. For any claim limited to trademark registrations/trademark numbers, Sodema supplies the field‑of‑use motivation (an automated clearance search has commercial value, and phonetic confusion is the classic statutory likelihood‑of‑confusion factor). Combining Sodema's automated conflict‑classification purpose with the admitted string‑similarity pipeline, implemented via the Hopcroft & Ullman formalism, is an application of a known technique to a known field with predictable results. See KSR; In re Kollman‑type reasoning on analogous‑art/field‑of‑use motivation.

Combination III (for the retrieval claims 3/4 and 31/32): Combination I + US 5,495,409 (Kanno)

If the Examiner needs an express reference for "produce a finite state automaton … according to a well‑known subset constructing method," US 5,495,409 describes constructing a finite‑state machine — including for a "pattern‑matching machine … utilized for a retrieving apparatus" — from regular expressions and transforming transitions. That is a direct teaching of elements E and F's pruning mechanics. Two cautions: (a) its 1996‑02‑27 grant postdates the 1995‑06‑12 priority, so it is prior art only if it qualifies under pre‑AIA § 102(e) via its earlier (≈1994) underlying US filing or via an earlier JP publication; (b) § 103(c) may disqualify it entirely as commonly owned by Matsushita. Use it as corroboration of the skill level, not as a load‑bearing reference, unless its 102(e) status is confirmed.

Combination IV (Japanese‑language evidence): Combination I + JP 06‑259804 A / JP 08‑055363 A / the EP 0 749 076 [A] PAJ abstracts

The EP search report for the family member EP 0 749 076 A1 cited three PAJ abstracts, two of them [A] (background) documents from 1993‑07‑28 and 1994‑08‑26. These establish Japanese‑language retrieval art of the same vintage. They are useful to (i) corroborate the PHOSITA's ordinary skill in Japanese phonetic string handling at the priority date, and (ii) show that the problem space was crowded. Caution: the third item is marked [PA] (published 1995‑12‑26), postdating the priority date, and the underlying JP publication numbers were not retrieved — so I would not rely on any of these as a primary reference without first pulling the PAJ records and the corresponding JP publications.

V. Why a PHOSITA would have been motivated — consolidated

Secondary considerations (Graham factor 4): I found no evidence of unexpected results, long‑felt unmet need, failure of others, copying, industry praise, licensing, or commercial success attributable to the claimed encoding. The specification's asserted benefits (short computation time, small memory) are exactly the results one predicts from compiling a regex and generating strings lazily — a strong indication that they are not the "unexpected results" that rebut a prima facie case. The patent's expiry in 2016 and the absence of any identified assertion history (per the prior litigation section) also mean there is no litigation record of objective indicia to weigh.

VI. Where the obviousness case is weak — and what would need checking

I want to be explicit about the limits, because an honest § 103 opinion should not overstate.

1. Full claim language is unavailable. Elements D, E and G (derivation element fields, applicable sound number, the incremental/backtracking generation with state stack, claim 31's specific decrement‑and‑extend mechanics) are data‑structure‑specific. If the granted claims recite a particular field‑level encoding — e.g., the derivation type distinguished by name of the similar‑sound table plus a signed positional shift, with list numbers substituted for derivation types — the obviousness showing must come from the references' combined teachings, not merely from "regexes and FSAs were known." My conclusion is stated at the level of the Summary; the claim‑level chart must be redone once the text is in hand.
2. Once‑truncated citations list. I verified US 3,952,184 and (partially) US 5,138,669 only. The complete "References Cited" list was not obtainable in this session, so there may be additional, closer art (especially Japanese Kokai). This materially affects the strength — but not the direction — of the analysis.
3. Same‑inventor art and § 103(c). US 5,495,409's usefulness depends on ownership and on its § 102(e) date.
4. Analogue‑art/field‑of‑use. Sodema is a 1976 mechanical/electrical classification apparatus; whether it is analogous art (reasonably pertinent to the problem) should be argued via In re Bigio / In re Clay, and it is best used only for the trademark‑domain motivation, not as the teaching of the algorithmic elements.
5. A genuine "nonobvious representation" counter‑argument exists and should be addressed head‑on: the applicant could argue that the prior art taught only "expand everything, then look up," and that recognizing that the whole enumeration can be replaced by (i) a per‑position derivation‑element table and (ii) an automaton over derivation‑type strings is a non‑trivial insight — a change in the kind of thing computed, not just a substitution of components. The rebuttal (which I find more persuasive at the level of the Summary) is that automaton‑guided, pruned generation of a regular language is itself textbook: the automaton that recognizes the language is the generator constraint, and Hopcroft & Ullman teach precisely the regex↔automaton↔language equivalence that makes the substitution available. Under KSR, re‑characterizing an eager enumeration as a lazy traversal of an equivalent automaton is a predictable improvement, not an inventive leap.

VII. Conclusion

On the record available, the independent claims of US 5,835,892 are most likely obvious under pre‑AIA § 103, on the primary combination of:

the applicant's own admitted prior art (FIGS. 1–11: expanding apparatus 101, indexing apparatus 111, retrieving apparatus 121) in view of Hopcroft & Ullman (1969) and/or Salomaa (1973) — both cited in the specification — and, for the trademark‑limited claims, further in view of US 3,952,184 (Sodema), with US 5,495,409 (Kanno/Matsushita) available as corroboration (subject to § 102(e)/§ 103(c)) for the finite‑state‑automaton construction and pattern‑matching‑for‑retrieval elements.

Motivation: the specification itself identifies the problem (procedurally coded rules that must be reprogrammed when criteria change; infeasible enumeration of millions of strings), and the claimed solution — declarative regex‑plus‑range rule tables compiled into an FSA, with the derivation‑element table + state‑transition table standing in for the enumerated expansions — is the predictable application of known formal‑language techniques to a known string‑similarity pipeline, with a reasonable expectation of success and no evidence of unexpected results.

Residual uncertainty: my conclusion is pitched at the level of the "Summary of the Invention" because the grant claims were not part of the supplied text. If the granted claims add specific structural limitations beyond what the Summary recites, the analysis must be re‑run against the verbatim claim language.

Recommended verification steps (to close the gaps)

1. Retrieve the full grant text including all 32 claims (USPTO PatentCenter / Google Patents PDF) — the single most important gap.
2. Pull the complete "References Cited" / "Cited By" block from the US 5,835,892 front page and the EP 0 749 076 A1 search report, including the underlying JP publication numbers for the three PAJ abstracts (two [A], one [PA]).
3. Confirm the title and teaching of US 5,138,669 A.
4. Confirm US 5,495,409's US filing date and the § 103(c) common‑ownership status at the 1995‑06‑12 priority date.
5. Confirm the JP 06‑259804 A and JP 08‑055363 A subject matter and dates, and whether either has a pre‑1995‑06‑12 effective date.
6. Because the prior-art section of the page was truncated in the source text, re‑fetch the page and confirm nothing material was omitted between the Description and the Citations block.