Datasets:
Tasks:
Text Classification
Modalities:
Text
Formats:
parquet
Languages:
German
Size:
10K - 100K
Tags:
legal
DOI:
License:
Update README.md
Browse files
README.md
CHANGED
|
@@ -8,11 +8,11 @@ configs:
|
|
| 8 |
path: "eval.parquet"
|
| 9 |
- split: test
|
| 10 |
path: "test.parquet"
|
| 11 |
-
|
| 12 |
-
|
|
|
|
| 13 |
- split: metadata
|
| 14 |
path: "project_summaries.parquet"
|
| 15 |
-
default: true
|
| 16 |
|
| 17 |
license: cc-by-4.0
|
| 18 |
task_categories:
|
|
@@ -61,6 +61,7 @@ These annotations are based on a specific snapshot of the sustainability procure
|
|
| 61 |
- `text_cleaned` (string): Normalized and cleaned text
|
| 62 |
- `catalog` (string): Sector-specific guideline (e.g., "Road_Transport", "Print_Services", "ICT", "Food", etc.)
|
| 63 |
- `area_of_action_id` (string): Specific criterion ID within the catalog (e.g., "HF-001", or "not present" for negative examples)
|
|
|
|
| 64 |
- `procurement_criterion` (string): Legal context type (e.g., "technical_specification", "award_criterion", "selection_criterion", "terms_and_conditions", "evidence_requirement", or "not present")
|
| 65 |
- `self_declaration` (string): Indicates if it's a self-declaration ("True"/"False" as string, or "not present")
|
| 66 |
- `evidence` (string): Evidence requirements (or "not present")
|
|
@@ -68,16 +69,32 @@ These annotations are based on a specific snapshot of the sustainability procure
|
|
| 68 |
|
| 69 |
### Data Splits
|
| 70 |
|
| 71 |
-
The dataset provides a **
|
| 72 |
- **50% Training**
|
| 73 |
-
- **25% Evaluation**
|
| 74 |
- **25% Test**
|
| 75 |
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 81 |
|
| 82 |
### Sector-Specific Guidelines (Catalogs)
|
| 83 |
|
|
@@ -92,6 +109,42 @@ The dataset includes sustainability criteria from multiple sectors:
|
|
| 92 |
- Product Independent Criteria
|
| 93 |
- And others
|
| 94 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 95 |
## Dataset Creation
|
| 96 |
|
| 97 |
### Source Data
|
|
@@ -110,7 +163,7 @@ The generation pipeline includes:
|
|
| 110 |
2. **Negative Sampling:**
|
| 111 |
- **Hard Negatives:** Sentences with high semantic similarity (cosine similarity) but no sustainability criteria
|
| 112 |
- **Random Negatives:** Random sentences for standard administrative language
|
| 113 |
-
3. **Contextual Windowing:** Creates context windows (1 or 2 surrounding sentences)
|
| 114 |
4. **Text Cleaning:** Normalization of line breaks, removal of excess whitespace (available in `text_cleaned`)
|
| 115 |
5. **Deduplication:** Based on `project_id`, `simap_version`, `text_cleaned`, `filename`, `catalog`, and `area_of_action_id`
|
| 116 |
|
|
@@ -132,6 +185,7 @@ Annotation categories (`procurement_criterion`) include:
|
|
| 132 |
|
| 133 |
- **NLP Engine:** spaCy (model: de_dep_news_trf) for German sentence segmentation
|
| 134 |
- **Embeddings:** SentenceTransformers (all-MiniLM-L6-v2) for semantic search and hard negative mining
|
|
|
|
| 135 |
- **Environment:** Python 3.10 via condavenv
|
| 136 |
|
| 137 |
## Considerations
|
|
@@ -145,10 +199,11 @@ This dataset supports:
|
|
| 145 |
|
| 146 |
### Limitations
|
| 147 |
|
| 148 |
-
- **Language:** Dataset is German-only, specific to Swiss German administrative language
|
| 149 |
- **Domain Specificity:** Trained on Swiss public procurement documents; may not generalize to other jurisdictions
|
| 150 |
- **Context Dependency:** Some criteria require understanding of broader document context
|
| 151 |
- **Class Imbalance:** Distribution between positive and negative examples should be monitored during training (mapped via `area_of_action_id` != "not present")
|
|
|
|
| 152 |
|
| 153 |
## Citation
|
| 154 |
|
|
|
|
| 8 |
path: "eval.parquet"
|
| 9 |
- split: test
|
| 10 |
path: "test.parquet"
|
| 11 |
+
default: true
|
| 12 |
+
- config_name: metadata
|
| 13 |
+
data_files:
|
| 14 |
- split: metadata
|
| 15 |
path: "project_summaries.parquet"
|
|
|
|
| 16 |
|
| 17 |
license: cc-by-4.0
|
| 18 |
task_categories:
|
|
|
|
| 61 |
- `text_cleaned` (string): Normalized and cleaned text
|
| 62 |
- `catalog` (string): Sector-specific guideline (e.g., "Road_Transport", "Print_Services", "ICT", "Food", etc.)
|
| 63 |
- `area_of_action_id` (string): Specific criterion ID within the catalog (e.g., "HF-001", or "not present" for negative examples)
|
| 64 |
+
- `sustainability_dimensions` (string): Sustainability dimensions of the annotated area of action as a bracketed list, e.g. "[ecologic, social]" ("[]" for negative examples or when no dimension applies)
|
| 65 |
- `procurement_criterion` (string): Legal context type (e.g., "technical_specification", "award_criterion", "selection_criterion", "terms_and_conditions", "evidence_requirement", or "not present")
|
| 66 |
- `self_declaration` (string): Indicates if it's a self-declaration ("True"/"False" as string, or "not present")
|
| 67 |
- `evidence` (string): Evidence requirements (or "not present")
|
|
|
|
| 69 |
|
| 70 |
### Data Splits
|
| 71 |
|
| 72 |
+
The dataset provides a **coverage-aware, leakage-safe split**:
|
| 73 |
- **50% Training**
|
| 74 |
+
- **25% Evaluation**
|
| 75 |
- **25% Test**
|
| 76 |
|
| 77 |
+
Splitting happens at the *document-snippet* level: a snippet and all of its
|
| 78 |
+
annotation rows always land in the same split. Two seeded (and therefore
|
| 79 |
+
reproducible) phases then allocate the snippet groups:
|
| 80 |
+
1. **Coverage** — every `(catalog, area_of_action_id)` pair of the positively
|
| 81 |
+
annotated snippets and every real `procurement_criterion` value is
|
| 82 |
+
processed rarest-first and placed into splits that do not contain it yet.
|
| 83 |
+
Pairs held by 3 or more snippets are guaranteed to appear in all three
|
| 84 |
+
splits; rarer pairs cover as many splits as possible and are flagged in the
|
| 85 |
+
`Split_Coverage` sheet of `statistics.xlsx`.
|
| 86 |
+
2. **Balancing** — the remaining snippets go to the split where their labels
|
| 87 |
+
lag furthest behind the ideal 50/25/25 quota, so `procurement_criterion`
|
| 88 |
+
shares typically deviate by less than 1 % from the ideal proportions.
|
| 89 |
+
|
| 90 |
+
`statistics.xlsx` additionally reports how the annotations are distributed
|
| 91 |
+
across the splits at row level: the `Split_Areas_Of_Action`, `Split_Catalogs`
|
| 92 |
+
and `Split_Criteria` sheets each list every label with its total annotation
|
| 93 |
+
count, its count per split and the share of the label's annotations that each
|
| 94 |
+
split receives.
|
| 95 |
+
|
| 96 |
+
The sentinels "not present"/"not specified" never count as labels; negative
|
| 97 |
+
snippets carry no labels and are distributed purely by remaining split capacity.
|
| 98 |
|
| 99 |
### Sector-Specific Guidelines (Catalogs)
|
| 100 |
|
|
|
|
| 109 |
- Product Independent Criteria
|
| 110 |
- And others
|
| 111 |
|
| 112 |
+
### Metadata Subset (`project_summaries`)
|
| 113 |
+
|
| 114 |
+
Besides the annotated text data, the repository publishes a per-project
|
| 115 |
+
metadata subset — `project_summaries.json` / `project_summaries.parquet`, one
|
| 116 |
+
row per procurement project (~56 rows). It documents *who annotated what* and
|
| 117 |
+
*which sector catalogs came into question*, so the text data can be put into
|
| 118 |
+
project context without exposing the annotation workflow's internal
|
| 119 |
+
bookkeeping. On Hugging Face the subset is published as its own config/split
|
| 120 |
+
pair — config `metadata`, split `metadata` (see the `configs:` block above) —
|
| 121 |
+
so the Dataset Viewer shows it in the split selector next to the `default`
|
| 122 |
+
config's `train`/`eval`/`test` splits, and
|
| 123 |
+
`load_dataset("IntelliProcure/SwissSPARK", "metadata", split="metadata")`
|
| 124 |
+
loads it directly:
|
| 125 |
+
|
| 126 |
+
| Field | Type | Description |
|
| 127 |
+
|---|---|---|
|
| 128 |
+
| `project_id` | int | SIMAP project ID |
|
| 129 |
+
| `simap_version` | string | SIMAP platform version (e.g. "simap") |
|
| 130 |
+
| `division` / `cpv_codes` | string | CPV division text / CPV codes of the project |
|
| 131 |
+
| `annotators` | list of strings | **Pseudonymized** annotator IDs (see below) |
|
| 132 |
+
| `candidate_catalogs` | list of strings | Catalogs whose workbook column is marked `match` — they came into question for this project |
|
| 133 |
+
| `annotated_catalogs` | list of strings | Catalogs marked `annotated` — they were actually annotated for this project |
|
| 134 |
+
| `documents` | list of strings | The project's documents (dataset filenames, joinable to the main subset's `filename`) |
|
| 135 |
+
|
| 136 |
+
Catalog names equal the `catalog` vocabulary of the default config (e.g.
|
| 137 |
+
`Food`, `Road_Transport`, `Product_Independent_Criteria`).
|
| 138 |
+
|
| 139 |
+
**Annotator pseudonymization:** annotator names are never published raw. For
|
| 140 |
+
each release, all annotators recorded in the source workbook are sorted
|
| 141 |
+
alphabetically and mapped onto `annotator_1` … `annotator_N`; the mapping is
|
| 142 |
+
identical in every row. The IDs are **release-scoped**: they are regenerated
|
| 143 |
+
on every rebuild and the numbering can shift between releases when the
|
| 144 |
+
annotator roster changes. Columns that may contain personal data ("Recommended
|
| 145 |
+
by", internal comments) are never exported, and generation fails hard if a
|
| 146 |
+
real annotator name is detected anywhere in the subset.
|
| 147 |
+
|
| 148 |
## Dataset Creation
|
| 149 |
|
| 150 |
### Source Data
|
|
|
|
| 163 |
2. **Negative Sampling:**
|
| 164 |
- **Hard Negatives:** Sentences with high semantic similarity (cosine similarity) but no sustainability criteria
|
| 165 |
- **Random Negatives:** Random sentences for standard administrative language
|
| 166 |
+
3. **Contextual Windowing:** Creates context windows (1 or 2 surrounding sentences) around each excerpt; negative candidates overlapping annotated excerpts are skipped during mining instead of being published with an in-context warning
|
| 167 |
4. **Text Cleaning:** Normalization of line breaks, removal of excess whitespace (available in `text_cleaned`)
|
| 168 |
5. **Deduplication:** Based on `project_id`, `simap_version`, `text_cleaned`, `filename`, `catalog`, and `area_of_action_id`
|
| 169 |
|
|
|
|
| 185 |
|
| 186 |
- **NLP Engine:** spaCy (model: de_dep_news_trf) for German sentence segmentation
|
| 187 |
- **Embeddings:** SentenceTransformers (all-MiniLM-L6-v2) for semantic search and hard negative mining
|
| 188 |
+
- **Language filter:** lingua (restricted to German/French/Italian/English) excludes non-German sentences from mined negative examples
|
| 189 |
- **Environment:** Python 3.10 via condavenv
|
| 190 |
|
| 191 |
## Considerations
|
|
|
|
| 199 |
|
| 200 |
### Limitations
|
| 201 |
|
| 202 |
+
- **Language:** Dataset is German-only, specific to Swiss German administrative language (mined negative examples are filtered to German-only sentences via lingua; human-annotated positives are kept as annotated)
|
| 203 |
- **Domain Specificity:** Trained on Swiss public procurement documents; may not generalize to other jurisdictions
|
| 204 |
- **Context Dependency:** Some criteria require understanding of broader document context
|
| 205 |
- **Class Imbalance:** Distribution between positive and negative examples should be monitored during training (mapped via `area_of_action_id` != "not present")
|
| 206 |
+
- **Release-Scoped Annotator Pseudonyms:** the `annotator_*` IDs of the `project_summaries` config are recomputed from the source workbook on every release and may change between releases; they identify a person only within one published version
|
| 207 |
|
| 208 |
## Citation
|
| 209 |
|