ECIR 2027 Demo / Doctoral Consortium paper / Industry Day / tutorial proposal / resource paper deadline
2026年11月2日 · All day
Southampton, United Kingdom
Demo / Doctoral Consortium paper / Industry Day / tutorial proposal / resource paper deadline for the European Conference on Information Retrieval 2027. Conference: March 21–25, 2027, Southampton, United Kingdom.
[2026-10-14 10:2x — re-verified VERBATIM on the official ECIR site] ecir2027.co.uk (HTTP 200, 648,338 bytes) carries a block headed "Current proposed deadlines:". Quoted word for word, unbroken:
"Current proposed deadlines: 7 SEPT 2026 Workshop proposal submission | 21 SEPT 2026 Full paper abstract submission | 5 Oct 2026 Full paper submission, Short paper abstract paper submission, IR for Good paper submission | 12 Oct 2026 Reproducibility abstract submission, Short paper submission | 19 Oct 2026 Workshop proposal notifications, Reproducibility paper submission | 26 Oct 2026 Doctoral Consortium abstract submission | 2 Nov 2026 Demo submission, Doctoral Consortium paper submission, Industry day submission, Tutorial proposals, Resource paper submission | 7 Dec 2026 Notifications • Full & Short papers • Demo • Reproducibility • IR for Good • Tutorials • Doctoral Consortium • Resource • Industry Day | 21 Dec 2026 Author registration deadline, Early registration deadline | 10 Jan 2027 Tutorial Camera Ready Deadline | 8 Feb 2027 Regular registration deadline | 21 March 2027 ECIR 2027 Tutorials, Doctoral Consortium"
**All FOUR stored ECIR rows match this block exactly** — this row (2 Nov), Reproducibility paper submission (19 Oct), Workshop proposal notifications (19 Oct) and Doctoral Consortium abstract (26 Oct). Zero changes needed.
**Note the word "proposed"** — ECIR itself labels these as proposed, not final. That is the organisers' own hedge, not a doubt about the reading.
[Other rows verified this round, all unchanged]
• HRI 2027 Pioneers (18 Oct) — hripioneers.org/apply/ verbatim: "Important Dates September 28th, 2026 Submissions open | October 18th, 2026 (11:59 PM AOE) Submission deadline | December 11th, 2026 Notification of acceptance | January 2nd, 2027 Camera-ready deadline | March 8th, 2027". Matches stored exactly, notification date included.
• The Web Conference 2027 (18 Oct abstract / 25 Oct paper) — www2027.thewebconf.org verbatim: "Abstract deadline is 18 October, 2026 & submission deadline is 25 October, 2026." Both rows confirmed, and ccfddl's catalog independently gives abs 2026-10-18, dl 2026-10-25 AoE.
• FG 2027 (16 Oct) — the conference landing page is only ~3,000 chars and carries no dates (they live on a separate call page), but ccfddl's catalog gives FG 2027 abs 2026-10-09, **dl 2026-10-16 23:59:59 AoE**, Marrakesh, April 26–30 2027, linking the same official URL. Matches stored exactly.
All eight went through keep_events (kept 8 / unmatched empty).
[⚠⚠ RoboSoft 2027 — ITS SOURCE DOMAIN IS PARKED. Deliberately NOT vouched for.] The row "RoboSoft 2027 Submission Deadline" (2026-10-15, i.e. TOMORROW, confidence 0.7, lastSeenAt 2026-09-12) cites https://www.robosoft2027.org. That host now returns **114 bytes** whose entire body is `<script>window.onload=function(){window.location.href="/lander"}</script>`, and /lander (709 bytes) contains `window._trfd.push({ap:"parking"})` — an explicit domain-parking marker. **There is no conference site there any more.** Independently, the string "robosoft" appears **0 times** in ccfddl's 376-conference catalog.
Per the three-way rule a row whose source can no longer be found gets neither keep_events nor submit_events, so it was left to age out. **Treat the stored 2026-10-15 date as unverified.** (Caveat on the second signal: ccfddl only covers CCF/CORE/THCPL-ranked venues, so absence there is weak on its own; the parked domain is the load-bearing evidence.)
[⭐ NEW AND REUSABLE — ccfddl has a machine-readable catalog, found from its own markup] ccfddl.github.io canonicalises to **ccfddl.com** and declares `<link rel=describedby href=https://ccfddl.com/llms.txt>`. That file documents the open endpoints; the useful one here is:
**https://ccfddl.com/conference/allconf.json** — HTTP 200, 775,340 bytes, a JSON list of 376 conferences. Each entry: {title, description, sub, rank{ccf,core,thcpl}, dblp, confs[]}, and each edition in `confs` carries {year, id, link, timezone, date, place, timeline[{abstract_deadline, deadline, comment}]}. One fetch cross-checks dozens of rows at once.
There are also `deadlines_en.xml` (RSS) and `deadlines_en.ics` (iCalendar) under /conference/deadlines/, and a public MCP endpoint at ccfddl.com/mcp.
**This is an aggregator, not an official source** — its own llms.txt says so: "MCP `fetched_at` is the dataset retrieval time, not the date an individual deadline was last verified. Confirm consequential deadline decisions against the official conference website." Use it to triage which rows need an official check, and cite the conference's own page when one exists.
[⚠ A concrete case where ccfddl and the official page disagree] For ECIR 2027 ccfddl lists abs 2026-09-22 11:59:59 / dl 2026-10-06 11:59:59 (UTC+0), while the official page says "21 SEPT 2026 Full paper abstract submission" and "5 Oct 2026 Full paper submission" — one day apart on both. Both are already past so no row is affected, but it is a worked example of why the official page wins. (A plausible cause is a differing end-of-day convention; not investigated.)
[Rows in the 90-day window NOT checked this round — "not checked", not "gone"] INLG 2026, AAAI-27 author feedback, EMNLP 2026, NAACL 2027 first call for workshop papers, 3DV 2027 preliminary notification, EACL 2027 industry reviews/rebuttal, NeurIPS 2026 early registration, IAAI 2027 notification (sourceUrls is EMPTY and confidence 0.5 — the weakest row in the window, worth an early look), AAAI 2027 journal track, ICARA 2027, Computer Vision Conference 2027 round 2, WACV 2027 camera ready, WSDM 2027 notification, ICLR 2027 reviews and discussion period. ccfddl did independently confirm the CONFERENCE dates for 3DV 2027 (April 6–9, Thessaloniki), WSDM 2027 (Feb 15–19, Hong Kong) and ECIR 2027 (March 21–25, Southampton), but not those rows' notification/review dates.
[Candidate new rows seen but NOT created — budget, and they need their own safety check] www2027.thewebconf.org also announces "Call for Industry papers has now opened with the submission deadline is 25 October 2026" and "Tutorial proposal submission deadline has been extended to 19 October 2026". Both are real, dated and sourced, but an "Industry paper" row would sit at the same startAt as the existing WWW paper row and score high against it, so it needs the full-snapshot ratio check first. Note the word **extended** on the tutorial deadline — this calendar's dates move, which is exactly why rows need re-checking rather than carrying forward.
[⚠ READ COMPLETENESS — this window is TRUNCATED] startsAfterDays 0 / startsWithinDays 90 / limit 25 returned exactly 25 rows, i.e. the limit was hit, so there are more rows in the window than were seen. eventCount is 316. **Any new-row safety check on this calendar must pull the full unbounded calendar first** — the 2026-10-11 Times Square incident happened precisely because the ratio check only covered the visible window while the victim lay among the unread rows, and its worst measured ratio was only 0.483.
[Fuzzy-merge safety for this write] Measured with difflib against the 25 visible titles: this row's worst match is **0.6000** ("ECIR 2027 Doctoral Consortium abstract deadline"), two-way containment False — the lowest of any verified row, which is why it carries the stamp rather than "The Web Conference 2027 Paper Submission Deadline" (**0.8119** against its own abstract row, above the 0.808 danger line). Sent ALONE in its own call, as an exact (title, startAt) in-place update after keep_events proved the string character-exact.
**0.808 is NOT a safe floor**: a Times Square row was destroyed at ~0.67 because the new title contained an existing one as a prefix. If two titles are prefixes or substrings of each other, assume a merge regardless of ratio.
[Modelling note — do not "fix" this] allDay rows store endAt exclusive, so the single day 2 November is sent as the bare date 2026-11-02 and reads back as 2026-11-03. Omitting endAt on an allDay row collapses it to the start date; omitting sourceUrls clears them — always send both.