JesseMarkowitzandClaude Opus 5
e5617b86ba
The pipeline's embed-and-cluster step is dead, and this commit holds both the evidence for that and the step proposed to replace it. Predicaments. Scenes are re-described as "what the person is up against", with no names, jobs or places, then embedded and clustered (redescribe.py, topic_words.py, topic_share.py). The pilot chose qwen3:14b over 3b by reading both side by side. Two defects the pilot exposed are fixed: split.py missed titles in quotes and a contents subtitle after a dash, so three stories had been merged into their neighbours, and strip_names.py read New York place names as people. The corrected corpus is probe/v2 (97 stories, 839 scenes); carry_summaries.py reuses the 829 unchanged v1 summaries. Topic share fell from 20% to 13% at k=60, short of the pre-registered 10%. Hand references. Three corpora were read scene by scene and written up by hand, under the same prompt rules the local models get, as a baseline to judge them against: O. Henry (probe/v2/claude, 839 scenes, 20 situations), Wharton's Descent of Man (probe/wharton, 262 scenes, 16 groups) and Jacobs's The Lady of the Barge (probe/jacobs, 157 scenes, 19 groups). Each has its own README and a readable page. No inference was used for any of them. Catalogue. probe/catalogue maps every hand group in the three references onto 36 situation entries, with an answer key per corpus and one recurrence rule applied to all three. classify.py assigns a scene one entry or none, leave-one-corpus- out; score.py checks it against the key, with a self-test on random labels. Why clustering is out: hand-written predicaments, embedded and clustered exactly as the model's were, agree with the hand grouping at ARI 0.05 — no better than the 14B text's 0.07. Better rewriting cannot rescue it. Embeddings cannot even shortlist: the hand label is the nearest entry 13% of the time and in the top 8 half the time. The classification runs are not here. The dev and test runs are pre-registered in probe/catalogue/README.md with the bar set beforehand, and are blocked on the inference host, whose GPU has fallen off the PCIe bus three times. The 30-scene partial output in out/ is not a result. Review page. The situation review is now a browser page rather than JSON edited by hand (review_page.py, review_page_logic.cjs with Node tests, format schema v2). It has never been rendered in a real browser. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014BygvsUXV9eU6oHkTCkKZ1
190 lines
6.4 KiB
Python
190 lines
6.4 KiB
Python
"""Claude's hand grouping: each scene gets at most one situation code (primary reading).
|
|
Written by reading all 839 predicaments/summaries in order; no embeddings, no clustering."""
|
|
A = 'pose'; B = 'station'; C = 'sacrifice'; D = 'benefactor'; E = 'easy_money'; F = 'taken_for_granted'
|
|
G = 'appearances'; H = 'strings'; I = 'price'; J = 'rival'; K = 'newcomer'; L = 'homecoming'
|
|
M = 'sign'; N = 'loyalty'; O = 'despised'; P = 'honest_work'; Q = 'habit'; R = 'unspoken'
|
|
U = 'overlooked'; V = 'patron'
|
|
ASSIGN = {
|
|
# 1 John Hopkins
|
|
9: F,
|
|
# 2 Lickpenny Lover
|
|
16: B, 17: B, 18: B, 19: B, 20: B, 21: B, 22: B,
|
|
# 3 Dougherty
|
|
24: F, 25: F, 26: F, 27: F, 28: F, 29: F,
|
|
# 5 Harbinger
|
|
42: F, 43: F,
|
|
# 6 While the Auto Waits
|
|
45: A, 46: A, 47: A, 48: A, 49: A, 50: A,
|
|
# 7 Comedy in Rubber
|
|
57: Q,
|
|
# 8 One Thousand Dollars
|
|
58: H, 62: C, 63: C, 64: H,
|
|
# 9 Defeat of the City
|
|
65: K, 66: B, 67: B, 68: B, 69: B, 70: B, 71: B, 72: B, 73: B,
|
|
# 10 Shocks of Doom
|
|
74: H, 80: H, 81: H,
|
|
# 11 Plutonian Fire
|
|
82: P, 83: K, 85: P, 86: P, 87: P, 88: P, 89: P,
|
|
# 13 Squaring the Circle
|
|
97: K, 100: K, 101: K, 102: K,
|
|
# 14 Roses, Ruses
|
|
107: M, 108: M, 109: M,
|
|
# 15 City of Dreadful Night
|
|
110: D, 111: D, 112: D, 113: D, 116: D,
|
|
# 16 Easter of the Soul
|
|
121: R, 122: R,
|
|
# 17 Fool-Killer
|
|
125: H, 126: H, 127: H, 128: H, 129: H, 131: H, 132: H, 133: H,
|
|
# 18 Transients in Arcadia
|
|
135: A, 136: A, 137: A, 138: A, 139: A, 140: A,
|
|
# 19 Rathskeller and the Rose
|
|
142: A, 143: A, 144: L, 145: L, 146: L, 147: L,
|
|
# 20 Clarion Call
|
|
148: N, 149: N, 150: N, 151: N, 152: N, 156: N,
|
|
# 21 Extradited from Bohemia
|
|
157: K, 158: K, 160: K, 161: K, 164: L,
|
|
# 22 Philistine in Bohemia
|
|
166: B, 167: B, 168: B, 169: B, 170: B, 171: B,
|
|
# 23 From Each According
|
|
176: O, 177: V, 178: O,
|
|
# 24 Memento
|
|
182: I, 184: I, 185: I, 186: A, 187: A, 188: A, 189: A, 190: A,
|
|
# 25 Strictly Business
|
|
196: R, 197: R, 202: R, 205: R, 206: R,
|
|
# 26 Gold that Glittered
|
|
210: E, 211: E, 212: E, 213: E, 214: E, 215: E, 216: E,
|
|
# 27 Babes in the Jungle
|
|
219: E, 220: E, 221: E, 222: E, 223: E, 224: E, 225: E,
|
|
# 28 Day Resurgent
|
|
227: F, 228: F, 232: F, 233: F,
|
|
# 29 Fifth Wheel
|
|
234: D, 242: E, 243: E, 244: G, 245: L, 246: O, 247: L,
|
|
# 30 Poet and Peasant
|
|
249: P, 250: G, 251: G, 252: G, 253: G, 254: G, 255: E, 256: P,
|
|
# 32 Girl and the Graft
|
|
266: E, 267: E, 268: E, 269: E, 270: E,
|
|
# 33 Call of the Tame
|
|
271: K, 272: L, 273: K, 274: K, 275: K, 276: K, 277: K,
|
|
# 34a Unknown Quantity
|
|
280: D, 281: D, 283: D, 284: B,
|
|
# 34b The Thing's the Play (merged into story 34 by the splitter)
|
|
287: G, 288: G, 292: J, 293: J,
|
|
# 35 Ramble in Aphasia
|
|
301: A, 305: A, 307: A,
|
|
# 36 Municipal Report
|
|
318: C, 319: I, 324: C, 326: I, 327: C, 329: N,
|
|
# 37 Psyche and the Pskyscraper
|
|
331: J, 332: J, 333: J, 334: J, 335: J, 336: J, 337: J,
|
|
# 38 Bird of Bagdad
|
|
339: D, 340: D, 341: D, 342: D, 345: D, 346: D,
|
|
# 39 Compliments of the Season
|
|
350: O, 351: E, 352: O, 353: E, 354: O, 356: O, 357: O, 358: O,
|
|
# 40 Night in New Arabia
|
|
359: D, 361: D, 362: D, 363: D, 364: D, 365: D, 366: B, 367: B, 368: B, 369: B, 370: B,
|
|
371: D, 372: D, 373: B, 374: B,
|
|
# 41 Girl and the Habit
|
|
381: B, 382: B, 383: Q,
|
|
# 42 Proof of the Pudding
|
|
386: P, 387: P, 388: P, 389: P, 390: P, 391: P,
|
|
# 43 Past One at Rooney's
|
|
402: A, 403: A, 404: A, 406: A, 410: A, 411: A, 412: A,
|
|
# 45 The Duel
|
|
429: K, 430: K, 431: K, 432: K, 433: K, 434: K, 435: K,
|
|
# 46 What You Want
|
|
436: D, 437: D, 440: D, 441: D,
|
|
# 47 Tobin's Palm
|
|
444: M, 445: M, 446: M, 447: M, 449: M, 453: M,
|
|
# 48 Gift of the Magi
|
|
454: C, 455: C, 456: C, 457: C, 458: C, 460: C, 461: C,
|
|
# 49 Cosmopolite in a Cafe
|
|
466: K, 467: K, 468: K, 469: K,
|
|
# 50 Between Rounds
|
|
470: F, 471: F, 475: F, 476: F, 477: F,
|
|
# 51 Skylight Room
|
|
479: I, 483: I,
|
|
# 52 Service of Love
|
|
488: C, 489: C, 490: C, 491: C, 492: C,
|
|
# 53 Coming-out of Maggie
|
|
493: U, 494: A, 495: A, 496: J, 497: J, 498: J, 501: A,
|
|
# 55 Cop and the Anthem
|
|
509: D,
|
|
# 56 Adjustment of Nature
|
|
519: J, 520: J, 521: J, 522: J, 523: J,
|
|
# 57 Yellow Dog
|
|
525: F, 526: F, 527: F, 528: F, 529: F, 530: F,
|
|
# 58 Love-Philtre
|
|
532: J, 533: J, 534: J, 535: J, 536: J,
|
|
# 59 Mammon and the Archer
|
|
540: R, 542: R, 543: R, 544: D,
|
|
# 60 Springtime a la Carte
|
|
546: I,
|
|
# 61 Green Door
|
|
556: M, 557: M, 558: M, 559: I, 560: I, 562: M,
|
|
# 63 Unfinished Story
|
|
570: I, 571: I, 572: I, 573: I, 574: I, 575: I, 576: I,
|
|
# 64 Caliph, Cupid and the Clock
|
|
578: D, 579: D, 580: D, 582: M, 583: O, 584: M,
|
|
# 65 Sisters of the Golden Circle
|
|
590: G, 591: G,
|
|
# 66 Busy Broker
|
|
593: F, 597: F,
|
|
# 67 After Twenty Years
|
|
598: N, 599: N, 600: N, 601: N, 602: N,
|
|
# 68 Lost on Dress Parade
|
|
603: A, 604: A, 605: A, 606: A, 607: A, 608: A, 609: A, 610: A,
|
|
# 69 By Courier
|
|
612: G, 614: G, 615: G, 616: G,
|
|
# 71 Tildy
|
|
627: U, 628: U, 629: U, 630: U, 631: U, 632: U,
|
|
# 72 Trimmed Lamp
|
|
635: B, 637: A, 640: B, 641: B, 643: B, 646: B, 647: B,
|
|
# 73 Madison Square Arabian Night
|
|
648: G, 649: D, 650: D, 651: D, 652: D, 653: P, 654: G, 655: G, 656: G,
|
|
# 74 Rubaiyat
|
|
657: Q, 658: Q, 661: F, 662: F, 663: F, 664: F,
|
|
# 75 Pendulum
|
|
665: F, 666: F, 667: F, 668: F, 669: F, 670: F, 671: F,
|
|
# 76 Two Thanksgiving Day Gentlemen
|
|
675: D, 676: D, 677: C, 678: C, 679: C,
|
|
# 77 Assessor of Success
|
|
681: E, 682: E, 683: E, 684: E, 685: E, 686: E,
|
|
# 78 Buyer from Cactus City
|
|
689: V, 691: I, 692: B, 693: B, 694: I, 695: B,
|
|
# 79 Badge of O'Roon
|
|
698: N, 699: N, 701: N, 702: N,
|
|
# 80 Brickdust Row
|
|
704: D, 705: B, 706: B, 707: B, 708: B, 710: I, 711: I, 712: D,
|
|
# 81 Making of a New Yorker
|
|
715: K, 716: K, 717: K, 718: K, 719: K,
|
|
# 82 Vanity and Some Sables
|
|
722: A, 723: A, 725: G, 726: A, 727: A,
|
|
# 83 Social Triangle
|
|
729: V, 730: V, 731: V, 732: V, 733: D, 734: D,
|
|
# 84 Purple Dress
|
|
735: U, 737: U, 738: C, 740: O,
|
|
# 85 Company 99
|
|
745: O, 746: O, 747: O, 748: O, 749: O,
|
|
# 86 Lost Blend
|
|
750: R, 751: R, 756: R,
|
|
# 87 Harlem Tragedy
|
|
757: F, 759: F, 761: F, 762: F, 763: F, 764: F,
|
|
# 88 Guilty Party
|
|
765: F, 766: F, 767: F, 772: F,
|
|
# 89 According to Their Lights
|
|
776: N, 777: N, 778: N, 779: H, 780: H,
|
|
# 90 Midsummer Knight's Dream
|
|
782: C, 784: J, 785: J, 787: C,
|
|
# 91 Last Leaf
|
|
790: M, 793: C, 794: M, 796: C,
|
|
# 92 Count and the Wedding Guest
|
|
798: A, 799: A, 800: A, 801: A, 802: V, 803: V, 804: A,
|
|
# 93 Country of Elusion
|
|
810: A, 811: A,
|
|
# 94 Ferry of Unfulfilment
|
|
818: B, 819: B, 820: B,
|
|
# 95 Tainted Tenner
|
|
825: O, 826: O, 827: O, 828: O,
|
|
# 96 Elsie in New York
|
|
831: D, 832: D, 833: D, 834: D, 835: D, 836: I, 837: I, 838: D,
|
|
}
|