US 5941944

I'll search for authoritative information on this patent number, including any 2026 litigation activity.

Let me check for any litigation or docket references tied specifically to this patent number.

Summary of US Patent 5,941,944

Note on ID: Interpreting the number literally as US 5,941,944 B1 (US5941944A), application 08/828,813. No other number was substituted.

Bibliographic data

Abstract (as issued)

Weights are assigned to features of data objects and used to determine whether data objects are substantially identical. One application assigns weights to terms in web pages; the term weights are used to determine whether web pages are substantially identical. A set of "identicals" is generated for each indexed web page and used to repair broken hyperlinks — when a URL cannot be resolved (or cannot be resolved in a timely fashion), one of the identicals is returned to the requesting party instead of an error.

Plain-language overview of the independent claims

• Claim 1 (core method). Get weights for features of two data objects, where each weight quantifies how much that feature distinguishes the object within a collection of objects. Compare the weight of one selected feature in object #1 against the weight of the same feature in object #2 to decide whether the objects are substantially identical. If they are, serve/access object #2 when a request for object #1 cannot be satisfied in a predefined manner.
• Claim 10 (web-page variant of claim 1). Same weight-comparison logic, applied to terms of two web pages to judge substantial identity, and accessing the second page when the first is inaccessible.
• Claim 20 (hyperlink-repair method). A hypertext document has a hyperlink to a first web page with media content. When the user clicks and the access attempt fails (e.g., bad location info, or a timeout), a second web page is identified as holding substantially identical media content using the weight-comparison technique; the second page's media content is then accessed and returned to the user. Claims 24 emphasizes this happens transparently; claims 22–23 cover the "incorrect link info" and "timeout" failure modes.
• Claim 25 (weight-calculation method). Explicitly recites how each word weight is computed: (i) a collection frequency component (how often the term appears across the collection), (ii) a term frequency component (how many times the term appears in the document), and (iii) the product of the two, normalized. The normalized weights of the same term in two documents are compared to decide substantial identity, and the second document is accessed if the first cannot be accessed.
• Claim 28 — computer-readable medium version of claim 1.
• Claim 34 — computer-readable medium version of claim 10 (web pages).
• Claim 42 — computer-readable medium version of claim 20 (hyperlink repair / returning substitute media content).
• Claim 45 — computer-readable medium version of claim 25 (collection-frequency × term-frequency, normalized, then compared).

Dependent claims add details such as returning the substitute object to the requester, treating non-similar weights as a negative determination, comparing additional feature weights, requiring exact weight equality, URL-based requests, Internet/intranet placement, storing weights in an index, and determining the weights themselves.

Supporting disclosure highlights

• Weight formula: product of a collection-frequency component (e.g., log₁₀(N/n), where N = documents in collection, n = documents containing the term) and a document/term-frequency component, divided by a normalization term (the preferred normalization being cosine normalization, i.e., √ of the sum of squared term weights).
• Candidate search uses an inverted index (dictionary + occurrence streams + catalog) and scans the highest-weighted terms first, accumulating a dot-product similarity score compared against a threshold (example given: 0.8); a "best possible score" bound (1 − the term's contribution) is used for early rejection (an early-exit / branch-and-bound style pruning).
• Expressly extends to non-text features (image tokens, audio phonemes), OLE/CORBA object IDs, path names, and network OS resource identifiers — citing copending application Ser. No. 08/732,615 (filed 1996‑10‑16) for image classification.

2026 CAFC / USPTO docket check — no activity found

I ran targeted searches for litigation and Federal Circuit activity tied specifically to 5,941,944 (including "CAFC 2026 docket" and Microsoft Technology Licensing queries). Results:

• No CAFC 2026 appeal or docket entry referencing 5,941,944 was found. The 2026 Federal Circuit material that surfaced (e.g., Yechezkal Evan Spero v. Mercedes‑Benz USA, No. 2025‑1470, petition filed 2026‑06‑15) concerns U.S. Patent 10,894,503, an unrelated patent — not 5,941,944, and I am not treating it as related.
• The only hits for "5941944" outside the patent's own record were citations of it as prior art in later patents (e.g., US 11,688,145; US 10,108,612; US 10,249,300; US 6,321,228; US 8,943,428), not litigation.
• I also note that search results surfaced unrelated inventors named "Katie Jane Messerly" and "John Benjamin Messerly" and unrelated patents "to Messerly et al." (e.g., 6,076,051; application 08/886,814). These are not the inventor of 5,941,944 and were not conflated.

Uncertainty / caveats

• Litigation status is uncertain in the negative direction: my searches returned zero results rather than a definitive "no cases exist." I cannot rule out district-court, ITC, or PTAB proceedings that are not well indexed; PACER and the CAFC docket were not directly queried, so treat "no 2026 CAFC activity" as "none found," not as an affirmative clearance.
• The patent expired 2017‑03‑03 (anticipated expiration), so any 2026 appellate activity would most plausibly be a validity/damages tail from earlier litigation — none of which I located.
• Current-assignee and legal-status entries come from Google Patents' derived records and are explicitly assumptions, not legal conclusions, per that source's own disclaimer. The full authoritative text was used for the title, claims, dates, and inventor, so those are high-confidence.

I'll search for litigation involving US patent 5941944 across patent litigation sources.

Let me search more targeted litigation databases and docket sources.

Let me check specific litigation-tracking databases directly.

Summary of Findings

I found no litigation involving U.S. Patent No. 5,941,944.

After searching patent-litigation-oriented sources (Google Patents litigation data, Justia; and targeted queries designed to surface district-court, ITC, PTAB, CAFC, and NPE-database records), I could not identify any lawsuit, ITC investigation, PTAB proceeding, or Federal Circuit appeal in which U.S. Patent 5,941,944 ("Method for providing a substitute for a requested inaccessible object by identifying substantially similar objects using weights corresponding to object features," inventor John J. Messerly, filed March 3, 1997, issued August 24, 1999, originally assigned to [Microsoft Corp.](/litigations/by-plaintiff/Microsoft%20Corp.)) was asserted or challenged as a patent-in-suit.

Important caveat on the searches

The searches returned many false positives that a reviewer might mistake for this patent. Every hit was a different patent bearing a "944" shorthand:

• All Cell Technologies / Workhorse — asserted a battery-cooling '944 patent (U.S. Patent No. 6,892,944-adjacent family; All Cell Techs. v. Workhorse, N.D. Ill. 1:18-cv-01419). Not this patent.
• Oyster Point Pharma v. Lupin (D.N.J. 2:25-cv-17843) — an ANDA case referencing a '944 pharmaceutical patent. Not this patent.
• Arista/Cisco "944 Investigation" — a Section 337 investigation number, unrelated.
• The DOJ/Microsoft (U.S. v. Microsoft) and PLUS search-result documents merely list 5,941,944 as a citation/reference number, not as an asserted patent.

None of these involve U.S. 5,941,944.

Why this is plausible

• The patent was originally assigned to Microsoft Corporation and now shows as held by Microsoft Technology Licensing, LLC. Microsoft historically did not assert its own patents offensively in infringement suits of this type, and the patent is a fairly narrow method claim directed to broken-link repair.
• The patent's term expired (anticipated expiration March 3, 2017; Google Patents legal status: "Expired – Lifetime"), removing it as a practical litigation target.
• Google Patents' "Cited By" data shows extensive citation activity (265+ citing references, mostly later search/assistant/software patents), but citations are not lawsuits.

Answers to your specific fields

Recommended verification steps (since I could not fully confirm)

Because I was unable to complete direct lookups, the following should be checked manually before relying on a "no litigation" conclusion:

1. Unified Patents Portal — https://portal.unifiedpatents.com/patents/patent/US-5941944 (verify whether a Litigation tab exists).
2. CourtListener / RECAP — search docket text for "5,941,944" or "5941944."
3. Docket Alarm / RPX Insight / Lex Machina — patent-level litigation profile.
4. PTAB (USPTO Patent Trial and Appeal Board) — confirm no IPR/PGR/CBM.
5. ITC EDIS — confirm no Section 337 investigation listing this patent.

Candid limitation

I did not obtain a definitive negative from a primary litigation database for this specific patent number. Based on the available search results, no litigation is known for U.S. Patent No. 5,941,944, but a final confirmation against Unified Patents' patent page or Lex Machina would be needed to state that with certainty. I have not fabricated any case; the absences above reflect what the sources returned.

PTAB proceedings overview

Zero AIA trial proceedings — the canonical USPTO Open Data Portal block returns no IPR, PGR, or CBM for US 5,941,944 (no institution denials, no settlements, no FWDs, no PTAB appeals), which means the patent's 47 claims were never tested at the Board and the "hardened by IPR" / "claims are dead" framings both do not apply; for a defendant today this is a claim set carried entirely on its 1997-03-03 priority date and its 1999-08-24 issue date, with validity exposure decided (if at all) in district court, not at the PTAB.

---

Proceedings on file

None — no AIA trial proceeding identified

• Type: n/a (no IPR / PGR / CBM)
• Filed: n/a
• Status: no proceedings returned by the structured source (USPTO ODP "PTAB proceedings on file"); cross-checked by web search, which surfaced only the patent's own record and later patents citing 5,941,944 as prior art (e.g., US 11,688,145; US 10,108,612; US 10,249,300; US 7,689,667; US 8,943,428) — no petition, no institution decision, no FWD.
• Judge panel: n/a
• Petition grounds: n/a
• Institution decision: n/a
• Final Written Decision: n/a
• Settlement / termination: n/a
• Appeal: n/a — and consistent with the prior section, no Federal Circuit appeal of any PTAB decision on this patent exists (there is no Board decision to appeal).
• Defensive value: You cannot rely on § 315(e)(2) estoppel, on a canceled claim, or on any PTAB-found invalidity. Any AIA-trial defense must be built from scratch, and the practical barriers below are substantial.

Verification note (not a substitute for a direct docket query). The prompt designates the ODP-derived block as canonical, and I treated it as such. My supplemental searches did not surface any proceeding, but I did not directly query PTAB E2E (https://ptacts.uspto.gov/) or the Board's decisions page (https://www.uspto.gov/patents/patent-trial-and-appeal-board), so read this as "none found," not as an affirmative negative-clearance certification. If you are counsel of record on a live matter, run the E2E patent-number search (https://ptacts.uspto.gov/ptacts/) before relying on it.

---

Strategic summary

Claim status: all 47 claims UNTESTED at the PTAB. No claim of 5,941,944 has been canceled, disclaimed, or held unpatentable in an AIA trial, and none have been upheld either. The independent set remains intact as issued: claim 1 (weight-comparison of features of two data objects), claim 10 (web-page term variant), claim 20 (hyperlink-repair / substitute media content, with claims 22–24 covering the "incorrect link information," "timeout," and "transparently to the party" flavors), claim 25 (collection-frequency × term-frequency product, normalized, then compared), plus the computer-readable-medium counterparts 28, 34, 42, 45. For a defendant, that is a clean slate procedurally but a fully loaded claim set substantively: the independent claims are drafted in genera ("data object," "media content," "features"), and the specification's express extensions — image tokens per copending Ser. No. 08/732,615, audio phonemes, OLE/CORBA OIDs, path names, network-OS resource identifiers — mean a § 112 enablement/definiteness attack is the more natural target than the bare claim language.

Estoppel landscape: nothing to inherit. Because no petitioner ever obtained an institution decision on this patent, no § 315(e)(2) estoppel attaches to anyone, and there is no "reasonably could have raised" bar to trip over. The corollary cuts the other way too: there is no petitioner-side work product, no Board-construed claim terms, and no Phillips constructions you can borrow from a FWD. That is a mixed blessing — you get a free hand on prior art, but you also get no free claim-construction wins. Note the two structural headwinds on new AIA petitions: (i) the patent expired on 2017-03-03 (anticipated expiration per Google Patents' derived record, consistent with the "Expired – Lifetime" status), so a 2026 filing is only rationally motivated by live or threatened back-damages exposure, and (ii) a § 315(b) one-year bar runs from service of a complaint alleging infringement, so the window is keyed to when you were sued, not to the patent's term.

Pattern signals: none of the usual IPR-ecosystem fingerprints. There is no repeat petitioner, no serial petitioning, no patent-owner appeal campaign (there being no appealable FWD), and no evidence of a defensive aggregator such as Unified Patents in the chain — the only assignee history on file is the internal Microsoft → Microsoft Technology Licensing, LLC reassignment recorded 2014-12-09. The strongest observable signal is the cited-by footprint (265 forward citations) and the fact that later broken-link-repair patents in this space (US 7,689,667, "Protocol to fix broken links on the world wide web") cite 5,941,944 as prior art — i.e., the art itself, not the Board, is where this patent's validity has historically been contested. Treat that cited-by cluster as your prior-art lead sheet.

Caveat on contradictions. I found none between the structured "no PTAB activity" block and the web record; the two are consistent. Any appearance of 5,941,944 in a PTAB exhibit list is as prior art or as a background citation in someone else's trial, not as the challenged patent — do not read those hits as proceedings against this patent.

---

Recommended next steps

1. Confirm the negative directly, then document it. Query PTAB E2E by patent number and save the null result (https://ptacts.uspto.gov/ptacts/). Because a live PTAB proceeding on this patent carries no estoppel for anyone, and its absence is unusual for a patent this heavily forward-cited, the absence is itself worth memorializing for your file.
2. Do not plan around an IPR. With the patent expired (2017-03-03) and no petition ever filed, an IPR-based strategy is a poor first move unless a complaint has been served and back-damages are genuinely at stake — and even then, § 315(b) timing plus the expired term make it a damages-defense tool, not a kill shot. There is no FWD to link to and no disposition to quote; say so plainly in any internal assessment rather than implying a PTAB record exists.
3. Shift the validity fight to where the art actually is. Build the § 102/§ 103 case from the 1995–1996 prior-art landscape — hyperlink-integrity and URL-error-handling systems, inverted-index/term-weighting IR art (the claimed log₁₀(N/n) × tf / cosine-normalization math in claim 25 sits squarely in classic IR practice), and the image/audio classification lineage the specification itself cites (Ser. No. 08/732,615, filed 1996-10-16). Where the claims' distinguishing hook is the comparison-then-substitute-service step, focus on 1995–96 link-rot/redirect and caching art.
4. Watch the mechanics if a suit is filed. Should a defendant receive a complaint, the strategic clock is: (a) 35 U.S.C. § 315(b) — one year from service to file any IPR; (b) § 314(a)/Fintiv-style discretionary denial remains a live risk given the patent's age and the likely forum; and (c) for any new petition, the § 325(d) / Advanced Bionics and § 102/103 posture analysis should be built on the '944 prosecution history, which I have not examined here.
5. CBM is not a path. 5,941,944 is not a financial-services patent, and it likely qualifies as a "technological invention" in any event; separately, the CBM transitional program no longer accepts new petitions (it sunset for new filings on 2020-09-16). Do not plan around it.

Key links: patent record — https://patents.google.com/patent/US5941944/en; PTAB E2E trial search — https://ptacts.uspto.gov/ptacts/; Board decisions — https://www.uspto.gov/patents/patent-trial-and-appeal-board. No FWD, institution decision, or CAFC opinion exists for this patent to link.

I'll research the assignment record for this patent through USPTO sources and cross-references.

I hit the tool-call limit before I could pull every field directly from the USPTO Assignment Center, so I'm reporting what I could verify, clearly separating verified facts from inferences and gaps. No assignment was fabricated below.

---

Inventors

Basis: The Google Patents legal-events record for US 5,941,944 contains the row "1997‑03‑03 — Assigned to MICROSOFT CORPORATION … Assignors: MESSERLY, JOHN J.," i.e., an assignment from the sole inventor to Microsoft on the filing date. A same‑day inventor→Microsoft conveyance is strong circumstantial evidence of employment, but the assignment instrument itself (which would state the "employee/inventor" capacity and any employment-agreement recital) was not retrieved, so I mark the employer as inferred.

Unusual patterns — none found.

• Single inventor; no co-inventor departures to track.
• I found no evidence of the inventor departing Microsoft within 12 months of filing (no corroborating record either way — this is a data gap, not a clean negative).
• De-confliction flag (carried forward from the summary section): search results surfaced unrelated individuals named Katie Jane Messerly and John Benjamin Messerly, and unrelated patents "to Messerly et al." These are not the inventor of 5,941,944 and were not conflated.

---

Original assignee

Microsoft Corporation (Redmond, Washington) — the entity named on the issued patent.

• Line of business: Software/platforms — operating‑system, productivity, and web‑browser software. The claimed subject matter (broken‑link repair by substituting a near‑identical indexed web page) sits squarely in the browser/search lane that was core to Microsoft's business in 1997–2000.
• Did they ship a product embodying the claims? Unclear / likely not productized as a discrete feature. The specification describes a pre‑built "identicals" index consulted to substitute a page when a URL fails. I found no evidence in the record of a shipping Microsoft product implementing that specific substitute‑page mechanism. I am not asserting product embodiment — this is a genuine unknown, not a negative finding.
• Current status: Operating. Microsoft Corporation remains a going concern; the patent itself is now held by Microsoft Technology Licensing, LLC, Microsoft's wholly‑owned IP holding/licensing subsidiary created in the 2014 reorganization. Microsoft has never been in bankruptcy.

---

Assignment timeline

Recorded assignments found: 2 (inventor→Microsoft in 1997; Microsoft→Microsoft Technology Licensing, LLC in 2014). Google Patents' legal‑events block for this patent lists only these two ownership events, which corroborates that no third‑party transfers were recorded.

1. Executed 1997‑03‑03 (same date as filing) / recorded on or about 1997‑03‑03 — Reel/Frame NOT RETRIEVED

• Conveyance: Assignment ("ASSIGNMENT OF ASSIGNORS INTEREST")
• Assignor: John J. Messerly
• Assignee: MICROSOFT CORPORATION
• Correspondent: Not retrieved. The fetched patent record does not expose it, and I did not reach the Assignment Center record directly.
• Context: Standard inventor‑to‑employer invention assignment contemporaneous with filing. Caveat: the "1997‑03‑03" date as surfaced by Google Patents is the event date (the row appears because of the filing/priority event); it is likely but not confirmed to be the instrument's execution date.

2. Executed/effective 2014‑10‑14 / recorded 2014‑12‑09 — Reel 034543 / Frame 0001

• Conveyance: Assignment ("ASSIGNMENT OF ASSIGNORS INTEREST")
• Assignor: MICROSOFT CORPORATION
• Assignee: MICROSOFT TECHNOLOGY LICENSING, LLC
• Correspondent: Not retrieved for this patent's entry.
• Context: Internal corporate reorganization — Microsoft's 2014 carve‑out of its patent licensing and IP‑enforcement function into MTL (announced after Satya Nadella became CEO). This was a portfolio‑wide bulk conveyance, not a sale to a third party.

⚠️ Inference flag on Reel 034543/0001. I did not read that reel/frame off US 5,941,944's own record. It is the reel/frame that Espacenet INPADOC and Dimensions show for other Microsoft patents bearing the identical assignor (Microsoft Corporation), assignee (Microsoft Technology Licensing, LLC), effective date 2014‑10‑14, and recording date 2014‑12‑09. Because Microsoft recorded this portfolio transfer as a single master instrument, the same reel/frame almost certainly covers this patent — but treat the reel/frame as unverified for 5,941,944 until read directly at the Assignment Center.

End of chain: No further assignments. Patent expired 2017‑03‑03 (anticipated expiration; full term) while still held by MTL. There is no post‑expiry transfer.

---

Timeline diagram

timeline
    title Ownership of US 5941944
    1997 : Filed by Messerly
         : Assigned to Microsoft Corp
    1999 : Patent issued
    2014 : Microsoft internal reorg
         : Assigned to Microsoft Tech Licensing
    2017 : Patent term expired

---

NPE / troll-pattern signals

---

Verdict

Operating-company assertion — with an explicit qualification: no assertion of this patent was located, so the practical status is operating‑company‑held and non‑asserted.

Justification. The entire recorded ownership chain runs inventor John J. Messerly → Microsoft Corporation (1997) → Microsoft Technology Licensing, LLC (Reel 034543/0001, effective 2014‑10‑14, recorded 2014‑12‑09), and terminates there at the 2017‑03‑03 expiry. There is no transfer to any third‑party LLC, no listed NPE in the chain, no cascading transfers, and no bankruptcy event — so none of the eight NPE/troll signals is present (signal 3 remains unclear only because I could not read the correspondent field). Because the terminal owner is a captive licensing subsidiary of a large, product‑shipping operating company rather than a defensive aggregator, "Operating‑company assertion" is the closest rubric label; the honest caveat is that Microsoft/MTL never sued anyone over this patent, and the patent's usefulness was consumed entirely within the Microsoft family before expiring.

Verify at: https://assignmentcenter.uspto.gov/ (patent‑number search) — and the legacy mirror https://assignment.uspto.gov/patent/index.html

---

Data gaps a reviewer should close before relying on this

1. Reel/frame for the 1997 Messerly→Microsoft assignment — not retrieved; only the 1997‑03‑03 event date is confirmed.
2. Reel/frame for the 2014 transfer as applied to this patent — Reel 034543/0001 is inferred from sibling Microsoft recordings with matching dates; confirm it directly.
3. Correspondent of record on both entries — the single most diagnostic field in the requested brief, and the one I could not obtain. This is the reason signal 3 is "unclear" rather than "not present."
4. No contradiction detected with the previously generated sections: the expiry date (2017‑03‑03), the assignee history (Microsoft → MTL), and the "no litigation found" conclusion are all consistent with the assignment chain reconstructed above.

I'll search for the authoritative USPTO/Google Patents record for US 5,941,944 and, critically, its cited references (front-page "References Cited" / patent citations), since the previously-fetched text contained the Cited By list but not the citations made by the patent itself.

The general search didn't surface the front-page "References Cited" section. Let me pull the patent's citation list from Justia and other patent-data mirrors directly.

The results keep surfacing other patents' citation lists. I need US 5,941,944's own front-page References Cited section. Let me target the patent's PDF and USPTO full-text record directly.

Prior-Art Analysis — US 5,941,944 (literal number, no substitution)

0. What I searched, and the critical retrieval failure you need to know about

Identity confirmed (literal): US 5,941,944 B1 / US5941944A, application 08/828,813, "Method for providing a substitute for a requested inaccessible object by identifying substantially similar objects using weights corresponding to object features," John J. Messerly, [Microsoft Corp.](/litigations/by-plaintiff/Microsoft%20Corp.), filed 1997‑03‑03, issued 1999‑08‑24. I confined queries to the literal string "5941944"/"5,941,944" and discarded near-miss numbers.

What I could NOT retrieve: the front-page "References Cited" / "Patent Citations" block of US 5,941,944 (the ~backward citations listed on the patent). Every targeted query returned one of three things instead:

1. The patent's own Cited By (forward) list — the 265+ later patents that cite 5,941,944;
2. Other patents' citation lists that happen to contain "5,941,944" as one of their references (e.g., US 10,509,862, US 9,986,419, US 8,315,976, US 10,416,644, US 8,812,507, US 11,688,145);
3. False positives on the "944" shorthand (already flagged in your Litigation section — All Cell Technologies, Oyster Point, the Arista §337 investigation).

I also hit the tool-call ceiling before I could reach patentimages / the ppubs.uspto.gov full-text record, which is where the backward-citation block lives. I will not reconstruct that list from memory, because assigning invented numbers and dates to a citation block is exactly the fabrication risk to avoid. So the honest state of play: the examiner/applicant-cited prior-art list for 5,941,944 is not in my retrieved evidence base. Below I give (a) the one reference I can ground in the authoritative full text, (b) a rigorous §102 framework your team can populate once the block is pulled, and (c) an explicit correction on the backward-vs-forward citation confusion that this task hinges on.

---

1. Disambiguation: "each patent citation for 5941944" is ambiguous — and the two readings lead to opposite conclusions

The 265‑entry "Cited By" list in the fetched text is thematically rich but legally irrelevant as prior art. It is useful only as evidence of the field's later development. Examples I saw (all post-date the filing, so not §102 art):

• US 6,687,734 — AOL, "System and method for determining if one web site has the same information as another web site"
• US 6,601,066 — General Electric, "Method and system for verifying hyperlinks"
• US 2003/0158953 — Lal, "Protocol to fix broken links on the world wide web"
• US 8,307,073 / US 8,458,227 — Amazon, "URL rescue by correction of encoding errors" / "…by identifying information related to an item referenced in an invalid URL"

Flag as a potential contradiction to watch: if an earlier draft or a downstream summary treats any of the "Cited By" entries above as §102 art, that is an error and should be corrected. They are §102-ineligible by date.

---

2. The one prior-art-type reference I can ground in the authoritative text

The full specification (authoritative, supplied) contains exactly one express reference to another document:

Reference R1 (as cited in the patent's own ______ background passage):

Copending application titled "An Image Classification and Retrieval System Using a Query‑By‑Example Paradigm," application Ser. No. 08/732,615, filed 1996‑10‑16, cited as "copending with the present application."

§102 analysis of R1 (pre‑AIA, discussed below):

• §102(e): R1 is not per se disqualified by being a copending application — under pre‑AIA §102(e) a U.S. application "by another" becomes prior art as of its filing date once it is published under §122(b) or issues as a patent. Filed 1996‑10‑16, R1 predates the 1997‑03‑03 filing, so if it names a different inventive entity and if it later issued/published, it is §102(e) art as of 1996‑10‑16.
• Anticipation reach: R1 on its face addresses image feature tokenization / query‑by‑example retrieval — one embodiment the '944 specification expressly extends to (image files). It does not, from the description in the '944 text, disclose the core novelty: weight comparison of a shared feature across two objects to detect substantial identity, nor the broken‑reference repair use. So R1 is realistically §103 (obviousness) material for the image‑feature dependent‑claim/embodiment scope, and a weak anticipation candidate for independent claims 1/10/20/25 (which require weight comparison and substitution/access behavior R1 is not described as having). I cannot assert §102 anticipation of any specific claim by R1 — the text I hold describes it only in one sentence, and no claim chart is possible from that.
• Caveat: I did not retrieve R1's own issue/publication record (e.g., whether it became a numbered patent). Treat the §102(e) point as conditional.

---

3. Why the "backward" list matters — and the §102 framework that must be applied to it

US 5,941,944 has a 1997‑03‑03 filing date, so pre‑AIA 35 U.S.C. §102 governs (the AIA first‑inventor‑to‑file §102 applies only to applications filed on/after 2013‑03‑16). The relevant sub‑sections your team should map each cited reference into:

Independent claims to chart (from the issued claim set): 1, 10, 20, 25, and the medium claims 28, 34, 42, 45. For a §102 anticipation rejection to be viable, a single reference must disclose every element, including:

The four‑element structure of claim 1 is the gating test. Term‑weighting art (see §4) alone cannot anticipate claim 1 because it lacks the substitution‑on‑inaccessibility element; broken‑link art alone cannot anticipate it because it lacks the "weight quantifies distinguishing amount within a collection" element. Anticipation in this patent will almost certainly turn on a reference that teaches both, or on §103 combinations.

---

4. Candidate prior-art fields to evaluate (NOT asserted as the patent's cited references)

Because I could not retrieve the actual citation block, I identify below the technical areas the examiner would have searched — labeled expressly as my own field analysis, not as the patent's "References Cited." Populate the real citations from the sources in §5 and run the §102/§103 tests above.

1. Term‑weighting / vector‑space retrieval (drives claim 25). Pre‑1997: the SMART/vector‑space literature (Salton's tf‑idf weighting and cosine normalization, which the '944 specification explicitly adopts as the "preferred" normalization and even names as the "cosine normalization function"). This is the strongest natural §102/§103 candidate against claim 25's weighting formula and against the "quantifies how much the feature distinguishes the object" limitation. I am confident this body of work predates 1997 and uses collection‑frequency × term‑frequency with cosine normalization.
2. Document near‑duplicate / similarity detection. Pre‑1997 work on detecting duplicate or near‑identical documents by comparing term weights — bears on claims 1/10/4/5/6.
3. Broken‑link detection & URL repair / substitute‑page serving. Bears on claim 20 and dependent claims 22–24; note the earliest broken‑link work largely postdates this filing, which is likely why this patent was allowed.
4. Inverted‑index architectures (dictionary + occurrence streams + catalog). Bears on claim 18 and the §112 support, not on novelty.
5. Image/audio feature tokenization. Bears on the non‑textual embodiments; this is where R1 (§2 above) lands, and where the patent itself cites copending art.

---

5. What to do to complete this task properly (I could not finish it)

To produce the per‑reference table you asked for (full citation · pub/filing date · description · claim(s) potentially anticipated under §102), the backward‑citation block must be pulled from a primary source, since my web results did not surface it:

1. USPTO Patent Public Search — https://ppubs.uspto.gov — open US 5,941,944; the "References Cited" and "Referenced By" tabs are separate. Use the "References Cited" tab.
2. Official PDF front page — the (56) "References Cited" block on the patent's own PDF (via patentimages.storage.googleapis.com/.../US5941944.pdf) — this is the version of record.
3. Google Patents "Citations" tab — https://patents.google.com/patent/US5941944A/en — verify you are on Citations (backward), not Cited By (forward).
4. Espacenet — "View list of cited documents" for the US publication.

Once retrieved, apply the §102 map in §3. Expect the examiner‑cited art to be categorized as §102(b) patents (issue dates before 1996‑03‑03) and §103 combinations; expect the allowance to have rested on the combination of a retrieval/weighting reference with a link‑resolution reference rather than a single anticipatory reference.

---

6. Candid limitations (do not over‑read this answer)

• I did not obtain US 5,941,944's own cited‑references list. This answer therefore does not provide the requested per‑reference §102 chart, because doing so would require inventing citations. That gap is a retrieval failure, not a finding that no such art exists.
• The only reference I can ground in the authoritative text is copending application Ser. No. 08/732,615 (filed 1996‑10‑16) — itself an application, cited for the image‑feature embodiment.
• The "Cited By" entries discussed in §1 (US 6,687,734; US 6,601,066; US 2003/0158953; US 8,307,073; US 8,458,227) are forward citations and are ineligible as §102 art to this patent.
• The sibling Messerly filings that appeared in PTAB petition excerpts — application 08/886,814 and US 6,076,051 ("Information retrieval utilizing semantic representation of text," filed 1997‑03‑07) — were cited as prior art against a different patent, and their 1997‑03‑07 filing postdates the 1997‑03‑03 filing here, so they are not §102 art to US 5,941,944. Do not confuse them with this patent's citation block.
• The patent is expired (anticipated expiration 2017‑03‑03), so §102 analysis is relevant only to historical validity challenges, not to current enforcement.

I'll ground this analysis in the actual prior-art record. First, let me try to retrieve the "Prior Art"/"References Cited" data for this patent, since the fetched page text I was given contains only the "Prior art keywords" line and a forward-citation ("Cited By") list.

Let me make one more targeted attempt to retrieve the examiner-cited prior-art list and non-patent literature for this patent.

Obviousness Analysis — U.S. Patent No. 5,941,944 (pre-AIA § 103)

0. Threshold note on the prior-art record (read first)

The instruction "use the results from the Prior Art section of this page" cannot be fully satisfied as written, and I want to flag that rather than paper over it. The authoritative page text I was given for US 5,941,944 contains:

• a "Prior art keywords" line only — web page; weights; computer; media content; substantially identical; and
• a "Cited By (265)" list — which is a list of later patents citing '944, i.e. forward citations covering material dated 1998 onward.

Neither is an examiner/applicant "References Cited" list. My follow-up searches (Google Patents, Justia, patentimages, PTAB exhibits) did not surface the front-page references-cited table for this patent. I therefore cannot responsibly recite a specific examiner-cited U.S.-patent list, and I will not fabricate numbers.

What I can ground the analysis on is (a) the references the patent itself identifies, (b) admitted prior art inside the specification, (c) the identified copending application, and (d) the indisputable state of the art in information retrieval as of 1997-03-03, which the search results independently corroborate. Everything below is labelled with its source and prior-art status. These are the combinations a challenger would actually run.

---

1. Governing framework and critical date

• Effective filing/priority date: 1997-03-03 (app. 08/828,813; issued 1999-08-24). Pre-AIA § 103 governs. Prior art = § 102(a)/(b)/(e) art as of that date.
• Status: "Expired – Lifetime," anticipated expiration 2017-03-03. Obviousness is therefore relevant today chiefly as (i) a validity/defense question in any tail litigation, (ii) an IPR/PGR-style challenge to the extent available, and (iii) a licensing/valuation question.
• Standard: Graham v. John Deere factors, as refined by KSR Int'l v. Teleflex (2007): scope/content of the prior art, differences, PHOSITA level, and secondary considerations. Post-KSR, a combination is obvious where it is "the product of ordinary creativity," a "predictable variation," or the pursuit of a "known option" with a "reasonable expectation of success." Note the asymmetry: KSR (2007) post-dates issuance, but it applies to the validity inquiry now.

---

2. Person having ordinary skill in the art (PHOSITA)

A team practitioner (or an individual) with a bachelor's degree in CS plus ~2 years, or equivalent, working on text-retrieval / index-engine and web-client engineering. By 1997 that person was presumed familiar with vector-space retrieval, TF-IDF weighting, cosine normalization, inverted-index construction, HTTP/HTML link resolution, and web crawling — all of which the specification itself concedes are "well known within the art" or "conventional."

---

3. Pivotal claim terms (construed for the § 103 inquiry)

The construction point matters: claim 1 is not limited to exact duplicate detection, to text, to the Internet, or to any particular weighting formula (all those are optional dependents). That breadth is what makes the independent claims vulnerable.

---

4. Reference inventory and prior-art status

Grounded / verifiable in this session:

Not prior art (field evidence only): every entry in the "Cited By" list (all post-3/1997 priority, e.g. US 6,112,203 (1998-04-09); US 6,185,614 (1998-05-26); US 6,631,492 (1998-06-26); US 6,687,734 (2000-03-21)), and the later US 7,689,667 / US 2003/0158953 "Protocol to fix broken links on the world wide web" family. These prove the problem was well known and the field was crowded, but they cannot be combined as § 103 art against the 1997 date. (This is the main reason the "Prior Art section" instruction matters: a naïve reading of the Cited-By list would produce an anachronistic analysis.)

I recall, but could not verify in this session, contemporaneous near-duplicate work such as Brin/Davis/Garcia-Molina, "Copy detection mechanisms for digital documents," SIGMOD 1995. I flag it as unverified rather than cite it as if confirmed.

---

5. Claim-by-claim § 103 mapping

Claim 1 (and its CRM twin, claim 28)

Combination [RA-2 or RA-1] + [RA-5/RFC 1945]. Every element is present. The only "extra" is using the similarity result to choose a substitute response when resolution fails — a routing/error-handling step, not a new computation.

Claim 10 (and claim 34) — web-page variant

Same as above with "web page" for "data object." RA-1 expressly contemplates "large hypermedia collections," and RA-2's pairwise-similarity machinery is document-agnostic. Even closer to the art than claim 1, because the field was already web-centric by 1994–96.

Claim 20 (and claim 42) — the hyperlink-repair method

Claim 25 (and claim 45) — the weight-calculation claim — weakest claim / strongest obviousness case

This claim recites the TF-IDF formula: "(i) collection frequency component… (ii) term frequency component… (iii) products of [the two]… and normalizing the products." RA-2 (Salton/Buckley 1988) and RA-1 (MG: "similarity heuristic such as the cosine rule… based on term frequency, document length, and co-occurrence") disclose this directly. Independently, the specification concedes cosine normalization is what "those skilled in the art will recognize." Claim 25 is a serious § 102(b) anticipation candidate, and a fortiori obvious.

Dependent claims

Claims 2/13/27/29/37/47 (return the substitute), 5/31 (exact equality), 4/15/30/38 (compare additional weights), 6/14/19/33/41 (negative determination on dissimilar weights), 11/12/35/36 (URL-based request/invalid URL), 22–23/43–44 (bad link info / timeout), 24 (transparent), 7/9/16 (Internet/intranet, network-OS resources), 8/17/32/39 & 40 (determine / index-store the weights) — each is either an express RA-1/RA-2 teaching or a predictable implementation choice with no criticality. None supplies an inventive contribution.

---

6. Motivation to combine (the KSR analysis a challenger would run)

1. Same field, same problem. Both RA-1 and RA-2 are information-retrieval references addressing document comparison; the patent's own field is "identification of substantially identical data objects by comparing scoring of features." KSR permits combination of references "designed to address the same problem."
2. The problem was notorious and the solution direction obvious. "Link rot" was a well-documented web-authorship problem by 1996 (evidenced by the dense post-1997 Cited-By activity around broken-link repair). When a broken link occurs, the obvious options are: error page, redirect, or substitute the same content elsewhere — a "finite number of identified, predictable solutions."
3. The building blocks were pre-assembled. RA-1 literally states that its document identifiers enable "an efficient translation process to convert symbolic hypertext links to the address of their target." Extending that to near-duplicate targets is a one-step generalization using machinery RA-1/RA-2 already provide (inverted index + tf/idf + cosine).
4. Reasonable expectation of success. Weighted-vector similarity was a mature, well-validated method for judging document likeliness; applying it to choose a substitute was expected to work.
5. Design incentives. Reducing user-visible failures, preserving content availability, and reusing an index the crawler already builds (spec's own stated aim: "without requiring large computational overhead") are classical efficiency/market motivations.
6. No teaching away, no criticality. Nothing in RA-1/RA-2 disparages the use; the 0.8 threshold is expressly exemplary ("different threshold values may be utilized"), which negates any argument that the claimed numbers are critical.

---

7. The best non-obviousness arguments (and why they likely fail)

The only claim with a plausible (not strong) defense is claim 1's specific combination of (weight-equality → substantial identity → substitute access) as a single claim, and even that is a textbook KSR obvious combination of RA-1/RA-2 with conventional link-failure handling. Claim 25 has essentially no defense — it claims TF-IDF.

---

8. Secondary considerations

No evidence of record supports objective indicia. There is no demonstrated unexpected result, no industry praise, no copying, no licensing-demand evidence, and no long-felt-but-unmet need documented in the materials provided. The "high degree of certainty without large computational overhead" language is an efficiency recitation tied to a predictable ordering of comparisons — a classic "results of ordinary design" argument that does not rebut a prima facie case.

---

9. Bottom line and confidence

• Claims 25 and 45: Likely anticipated under § 102(b) by Salton/Buckley and/or Witten-Moffat-Bell, and obvious in any event. High confidence.
• Claims 1, 10, 28, 34: Likely obvious over RA-1/RA-2 in view of the admitted link-failure art (RA-5/RFC 1945). Moderate-to-high confidence.
• Claims 20, 42 (+ 21–24, 43–44): Likely obvious over the same combination plus conventional hyperlink/substitute-service practice. Moderate confidence.
• All dependent claims: obvious as predictable implementations. High confidence, absent some unretrieved reference showing unexpected results.
• Confidence limiter: because the examiner-cited "References Cited" list is not present in the page text and I could not retrieve it, I have not tested the claims against the specific references the examiner actually considered. My confidence in the analysis is limited by that gap, not by any weakness in RA-1/RA-2, which are independently sufficient for claims 25/45 and highly probative for the rest.

10. What should be verified next

1. Front page "References Cited" of US 5,941,944 (USPTO PatentCenter / patentimages PDF) — to run the claim charts against the actual examiner-cited art rather than the reconstructed set above.
2. USPTO file wrapper for app. 08/828,813 — noting any § 102/103 rejections and how the applicant distinguished over cited art (the prosecution history may itself show the applicant's own view of the closest art).
3. Confirm RA-4 status: whether Ser. No. 08/732,615 issued, and its publication/date, to fix its § 102(e) treatment.
4. Unified Patents / PTAB / Lex Machina for any IPR or validity ruling that already addressed these combinations.

Contradiction/limitation flag for the previously generated sections: both the earlier Patent summary and Litigation summary already noted the thin prior-art/litigation record. This section does not contradict them; it sharpens the finding: the "Prior Art" content on the fetched page is limited to keyword tags and forward citations, so any § 103 opinion must be built on the references the specification itself identifies plus the field's admitted state of the art — which is what I did above.

This is a technical analysis, not a legal opinion; validity/invalidity determinations are context-specific and carry a presumption of validity.