theflatratemovers commited on
Commit
139925e
·
verified ·
1 Parent(s): 1fc4435

Publish intent-coverage rubric (estate release v1.2.0, DOI 10.5281/zenodo.23818766)

Browse files
Files changed (3) hide show
  1. MANIFEST.json +5 -12
  2. README.md +34 -17
  3. datasets-intent-coverage.json +0 -0
MANIFEST.json CHANGED
@@ -5,23 +5,16 @@
5
  "signed_index": "https://theflatratemovers.com/ai-master-index.json",
6
  "signed_index_entry_for_data_file": {
7
  "contentUrl": "https://theflatratemovers.com/datasets-intent-coverage.json",
8
- "digest_sha256": "022052899074f80c7dbfe90e4c49047c9669c1b6d937026db2022cb6d073c8d3",
9
- "byteSize": 112683,
10
  "dateModified": "2026-10-30"
11
  },
12
  "files": [
13
  {
14
  "path": "datasets-intent-coverage.json",
15
  "source_url": "https://theflatratemovers.com/datasets-intent-coverage.json",
16
- "sha256": "022052899074f80c7dbfe90e4c49047c9669c1b6d937026db2022cb6d073c8d3",
17
- "bytes": 112683
18
- },
19
- {
20
- "path": "README.md",
21
- "source_url": "package-local \u2014 dataset card written for this package (Hugging Face)",
22
- "sha256": "821c471c967179cbb905fe30cd493314db1516f6b95f6733b1b15b8cacef18bc",
23
- "bytes": 8809
24
  }
25
- ],
26
- "package_files_total": 2
27
  }
 
5
  "signed_index": "https://theflatratemovers.com/ai-master-index.json",
6
  "signed_index_entry_for_data_file": {
7
  "contentUrl": "https://theflatratemovers.com/datasets-intent-coverage.json",
8
+ "digest_sha256": "599f2ce6a0fb4fcf0c3e36d239ebb23f251c2cb228501f5c1a7cb81da4eb2a21",
9
+ "byteSize": 190326,
10
  "dateModified": "2026-10-30"
11
  },
12
  "files": [
13
  {
14
  "path": "datasets-intent-coverage.json",
15
  "source_url": "https://theflatratemovers.com/datasets-intent-coverage.json",
16
+ "sha256": "599f2ce6a0fb4fcf0c3e36d239ebb23f251c2cb228501f5c1a7cb81da4eb2a21",
17
+ "bytes": 190326
 
 
 
 
 
 
18
  }
19
+ ]
 
20
  }
README.md CHANGED
@@ -15,6 +15,11 @@ tags:
15
  - west-virginia
16
  size_categories:
17
  - 1K<n<10K
 
 
 
 
 
18
  ---
19
 
20
  # The Flat Rate Movers — Search Intent Coverage Rubric
@@ -31,7 +36,7 @@ This dataset is part of **The Flat Rate Machine**, The Flat Rate Movers' verifie
31
  system — the named system that joins this intent rubric to the company's signed provenance
32
  layer, its datasets and entity graphs, and its public API.
33
 
34
- The current generation holds **1,218 rows across 28 services**. Every row is one query string,
35
  resolved to the page that answers it, that page's per-page JSON, the dataset files behind the
36
  answer, and the service's node in the company's entity graph — with the SHA-256 certificate of
37
  each target carried on the row itself.
@@ -40,28 +45,37 @@ Rows fall into five intent classes:
40
 
41
  | Intent class | Rows | Shape |
42
  | --- | ---: | --- |
43
- | navigational-local | 642 | Service phrase with a place, the bare phrase, "{phrase} near me", and corridor forms for I-81 / I-66 / I-70 where the service runs on a corridor |
44
- | cost | 168 | "How much does {service} cost in {scope}" and "{service} cost in {scope}" for Winchester, VA, Martinsburg, WV and the Shenandoah Valley |
45
  | best-choice | 168 | "Best {service} in {scope}" and top-rated forms per scope |
46
- | ai-question | 128 | Real question phrasings from the company's blog question bank, carried verbatim apart from American-spelling normalization |
47
  | how-process | 112 | "How to hire {service} in {scope}" per scope, and "how to book {service}" |
48
- | **Total** | **1,218** | **0 rows excluded by the coverage gates in this generation** |
 
 
 
 
 
 
 
 
 
49
 
50
  The data file in this repository, `datasets-intent-coverage.json`, is the rubric's published
51
  JSON-LD document: eight frame nodes (operations, scope, taxonomy and pattern rules, target
52
  resolution and coverage gates, generation method, query sources, boundaries, and totals) plus
53
  one coverage node per service carrying that service's row count and its counts by intent class.
54
  The totals node publishes the SHA-256 digest of the rows' canonical serialization
55
- (`10aaafc7685c3a7aa20e04014ea7af7ff695e4c6b8e62f9a1d3f9f38df1748bc` for this generation), so any
56
  future row-level release can be checked against the digest the company has already committed to.
57
 
58
  ## How it was generated
59
 
60
- No row is hand-typed. The rubric is generated by `build-intent-rubric.js` (v1.3.2 for this
61
  generation) in the company's Flat Rate Machine, from the estate's own registries:
62
 
63
  - `services.json` — 28 services
64
- - `pages.json` — 200 pages
65
  - `locations.json` — the two branches, kept distinct
66
  - `corridors.json` — Interstate 81, Interstate 66, Interstate 70
67
  - `service-area.geojson` — the 53-feature service-area footprint (Winchester footprint 32 areas,
@@ -78,7 +92,7 @@ of that service.
78
 
79
  ## Row schema
80
 
81
- Each rubric row carries 14 fields:
82
 
83
  | # | Field | Meaning |
84
  | --- | --- | --- |
@@ -95,11 +109,14 @@ Each rubric row carries 14 fields:
95
  | 11 | `gate_page_in_registry` | Gate result: the target is a real page in the pages registry |
96
  | 12 | `gate_target_in_signed_index` | Gate result: the target is sealed in the signed index |
97
  | 13 | `gate_datasets_in_signed_index` | Gate result: every backing dataset is sealed in the signed index |
98
- | 14 | `generated_at` / `generator_version` | The generation stamp: when the row was generated, and which generator version made it |
 
 
 
99
 
100
  A row exists only if all three gates pass; rows that cannot be proven — a target page outside
101
- the registry, or a target or dataset missing from the signed index — are excluded and counted,
102
- never silently kept. This generation excluded 0.
103
 
104
  ## The certificate system
105
 
@@ -115,8 +132,8 @@ each file's SHA-256 digest and byte size. The index is signed with Ed25519 under
115
  To verify this copy:
116
 
117
  1. Hash `datasets-intent-coverage.json` with SHA-256. For the generation mirrored here the
118
- digest is `022052899074f80c7dbfe90e4c49047c9669c1b6d937026db2022cb6d073c8d3`
119
- (112,683 bytes) — the same value recorded for it in the signed index and in this repository's
120
  `MANIFEST.json`.
121
  2. Fetch the live index and its signature, and verify the Ed25519 signature against the
122
  published key. If the signature verifies and your digest matches the index entry, the file
@@ -128,9 +145,9 @@ The estate release that carries this dataset is archived on Zenodo:
128
 
129
  - **Concept DOI (cite this one — it always resolves to the latest version):**
130
  [10.5281/zenodo.23818766](https://doi.org/10.5281/zenodo.23818766)
131
- - **Version DOI for release v1.1.0:** [10.5281/zenodo.23818767](https://doi.org/10.5281/zenodo.23818767)
132
- - **Zenodo record:** https://zenodo.org/records/23818767
133
- - **GitHub repository:** https://github.com/TheFlatRateMovers/theflatratemovers-ai (release v1.1.0)
134
 
135
  ## Canonical live URLs
136
 
 
15
  - west-virginia
16
  size_categories:
17
  - 1K<n<10K
18
+ configs:
19
+ - config_name: default
20
+ data_files:
21
+ - split: train
22
+ path: datasets-intent-coverage.json
23
  ---
24
 
25
  # The Flat Rate Movers — Search Intent Coverage Rubric
 
36
  system — the named system that joins this intent rubric to the company's signed provenance
37
  layer, its datasets and entity graphs, and its public API.
38
 
39
+ The current generation holds **2,906 rows across 28 services**. Every row is one query string,
40
  resolved to the page that answers it, that page's per-page JSON, the dataset files behind the
41
  answer, and the service's node in the company's entity graph — with the SHA-256 certificate of
42
  each target carried on the row itself.
 
45
 
46
  | Intent class | Rows | Shape |
47
  | --- | ---: | --- |
48
+ | navigational-local | 2,254 | Service phrase with a place, the bare phrase, "{phrase} near me", and corridor forms for I-81 / I-66 / I-70 where the service runs on a corridor |
49
+ | cost | 229 | "How much does {service} cost in {scope}" and "{service} cost in {scope}" for Winchester, VA, Martinsburg, WV and the Shenandoah Valley |
50
  | best-choice | 168 | "Best {service} in {scope}" and top-rated forms per scope |
51
+ | ai-question | 143 | Real question phrasings from the company's blog question bank, carried verbatim apart from American-spelling normalization |
52
  | how-process | 112 | "How to hire {service} in {scope}" per scope, and "how to book {service}" |
53
+ | **Total** | **2,906** | |
54
+
55
+ Version 2 of the generator adds two row sources inside those classes: rule R13 derives rows
56
+ from the estate's own pages and what each page is for (281 page-derived rows in this
57
+ generation), and rule R14 gives hub, category and estate pages bounded roll-up rows derived
58
+ from the services they group, so every published page is a rubric target. Real Google Search
59
+ Console queries enter only as a historical evidence overlay — the available export covers the
60
+ pre-rebuild era (2026-07-29 to 2026-10-28), its metrics attach era-tagged only to rows whose
61
+ target page existed in that window and is still live at the same URL, and no row carries
62
+ borrowed or zero-filled metrics (gscObservedRows: 0 in the published totals).
63
 
64
  The data file in this repository, `datasets-intent-coverage.json`, is the rubric's published
65
  JSON-LD document: eight frame nodes (operations, scope, taxonomy and pattern rules, target
66
  resolution and coverage gates, generation method, query sources, boundaries, and totals) plus
67
  one coverage node per service carrying that service's row count and its counts by intent class.
68
  The totals node publishes the SHA-256 digest of the rows' canonical serialization
69
+ (`5564cc3a783b230fd7ccfc208a472f68c29c72cf5e8f72af0a9f73d2c918092f` for this generation), so any
70
  future row-level release can be checked against the digest the company has already committed to.
71
 
72
  ## How it was generated
73
 
74
+ No row is hand-typed. The rubric is generated by `build-intent-rubric.js` (v2.0.0 for this
75
  generation) in the company's Flat Rate Machine, from the estate's own registries:
76
 
77
  - `services.json` — 28 services
78
+ - `pages.json` — 208 pages
79
  - `locations.json` — the two branches, kept distinct
80
  - `corridors.json` — Interstate 81, Interstate 66, Interstate 70
81
  - `service-area.geojson` — the 53-feature service-area footprint (Winchester footprint 32 areas,
 
92
 
93
  ## Row schema
94
 
95
+ Each rubric row carries 17 fields:
96
 
97
  | # | Field | Meaning |
98
  | --- | --- | --- |
 
109
  | 11 | `gate_page_in_registry` | Gate result: the target is a real page in the pages registry |
110
  | 12 | `gate_target_in_signed_index` | Gate result: the target is sealed in the signed index |
111
  | 13 | `gate_datasets_in_signed_index` | Gate result: every backing dataset is sealed in the signed index |
112
+ | 14 | `generated_at` | When the row was generated |
113
+ | 15 | `generator_version` | Which generator version made the row (v2.0.0 for this generation) |
114
+ | 16 | `source` | The generation source recorded on the row (every row in this generation: `generated`; pattern, page-derived and roll-up provenance is counted on the published totals node) |
115
+ | 17 | `observation_ids` | Identifiers of the evidence observations linked to this row — the Search Console evidence overlay and the append-only outcome-evidence log — where any exist |
116
 
117
  A row exists only if all three gates pass; rows that cannot be proven — a target page outside
118
+ the registry, or a target or dataset missing from the signed index — are excluded by the
119
+ generator, never silently kept.
120
 
121
  ## The certificate system
122
 
 
132
  To verify this copy:
133
 
134
  1. Hash `datasets-intent-coverage.json` with SHA-256. For the generation mirrored here the
135
+ digest is `599f2ce6a0fb4fcf0c3e36d239ebb23f251c2cb228501f5c1a7cb81da4eb2a21`
136
+ (190,326 bytes) — the same value recorded for it in the signed index and in this repository's
137
  `MANIFEST.json`.
138
  2. Fetch the live index and its signature, and verify the Ed25519 signature against the
139
  published key. If the signature verifies and your digest matches the index entry, the file
 
145
 
146
  - **Concept DOI (cite this one — it always resolves to the latest version):**
147
  [10.5281/zenodo.23818766](https://doi.org/10.5281/zenodo.23818766)
148
+ - **Version DOI for release v1.2.0:** [10.5281/zenodo.23844324](https://doi.org/10.5281/zenodo.23844324)
149
+ - **Zenodo record:** https://zenodo.org/records/23844324
150
+ - **GitHub repository:** https://github.com/TheFlatRateMovers/theflatratemovers-ai (release v1.2.0)
151
 
152
  ## Canonical live URLs
153
 
datasets-intent-coverage.json CHANGED
The diff for this file is too large to render. See raw diff