README.md

---
task_categories:
- token-classification
tags:
- ner
- span-extraction
- data-mentions
- economics
license: cc-by-4.0
configs:
- config_name: bio
  data_files:
  - split: train
    path: bio_train.jsonl
  - split: val
    path: bio_val.jsonl
  - split: holdout
    path: bio_holdout.jsonl
- config_name: gliner
  data_files:
  - split: train
    path: gliner_train.jsonl
  - split: val
    path: gliner_val.jsonl
  - split: holdout
    path: gliner_holdout.jsonl
- config_name: gliner2
  data_files:
  - split: train
    path: gliner2_train.jsonl
  - split: val
    path: gliner2_val.jsonl
  - split: holdout
    path: gliner2_holdout.jsonl
dataset_info:
  config_name: gliner2
  features:
  - name: input
    dtype: string
  - name: output
    struct:
    - name: entities
      struct:
      - name: named_data
        list: string
      - name: descriptive_data
        list: string
      - name: vague_data
        list: string
    - name: entity_descriptions
      struct:
      - name: named_data
        dtype: string
      - name: descriptive_data
        dtype: string
      - name: vague_data
        dtype: string
  - name: corpus
    dtype: string
  - name: origin
    dtype: string
  - name: corpus_id
    dtype: string
  - name: page
    dtype: int64
  - name: chunk
    dtype: int64
  - name: title
    dtype: string
  - name: pdf_url
    dtype: string
  - name: extractor
    dtype: string
  - name: footnote_link
    dtype: bool
  - name: dedupe_overlap
    dtype: bool
  - name: has_data_score
    dtype: float64
  - name: split
    dtype: string
  - name: spans
    list:
    - name: text
      dtype: string
    - name: label
      dtype: string
    - name: score
      dtype: float64
    - name: start
      dtype: int64
    - name: end
      dtype: int64
  - name: singlepass_model
    dtype: string
  - name: singlepass_keep_thr
    dtype: float64
  - name: singlepass_entities
    list:
    - name: text
      dtype: string
    - name: start
      dtype: int64
    - name: end
      dtype: int64
    - name: extractor_score
      dtype: float64
    - name: probe_score
      dtype: float64
    - name: keep
      dtype: bool
  splits:
  - name: train
    num_bytes: 434822580
    num_examples: 201025
  - name: val
    num_bytes: 188660358
    num_examples: 42414
  - name: holdout
    num_bytes: 196872595
    num_examples: 44174
  download_size: 497942440
  dataset_size: 820355533
---

# Data-use mentions (NER / span extraction)

Data mentions extracted from World Bank Policy Research Working Papers and FCV documents, predicted by a span-extraction model with no human or LLM-judge
validation, and formatted for span-extraction (GLiNER / GLiNER2) and
token-classification (LFM2.5-encoder) fine-tuning.

## Labels

Three entity types:

- `NAMED_DATA` — a proper name, title, or acronym of a specific data source
- `DESCRIPTIVE_DATA` — a source described in words but not named
- `VAGUE_DATA` — generic data wording with no identifiable source

`O` (BIO) = background, including hard negatives.

## Negative strategy

- **No hard negatives** — every predicted span is kept as a positive.
- **Negative-only** chunks (no predicted spans) are sampled at a controlled ratio per split.

## Re-chunking (why windows are <= 384 tokens)

The upstream `input_text` is already an extractor chunk of at most `max_tokens` (default 384) whitespace tokens, so it fits GLiNER/GLiNER2's context window as-is; a sliding-window re-chunk pass runs only as a safety net.

## Provenance

Every row contains `corpus` (`prwp` or `fcv`) and `origin` (the extraction config, e.g. `general_prwp`, `fcv_pads_east_africa`, `jdc_operational`, `refugee_pads`, `reliefweb`). See `split_stats.json` for per-split provenance counts.

## Configs

- `gliner` — `{"tokenized_text": [...], "ner": [[start, end, label], ...]}` (word-level spans)
- `bio` — `{"tokens": [...], "ner_tags": ["O", "B-NAMED_DATA", ...]}`
- `gliner2` — `{"input": "...", "output": {"entities": {...}, "entity_descriptions": {...}}}`
  (GLiNER2 flat-NER, span strings + descriptions)


## Source (raw extraction, no judge)

Unlike the judged build, every span here is a raw prediction of `rafmacalaba/gliner_datause` (labels `NAMED_DATA, DESCRIPTIVE_DATA, VAGUE_DATA`, footnote_link=True, dedupe_overlap=True). There are no hard negatives; negatives are entity-less chunks sampled at ~0.25 per positive per split. Treat spans as noisy positives, not gold labels.

## Traceability columns

Every config carries the same flat provenance columns alongside its training keys, so any row joins back to the parsed `.md` (`<origin>/<corpus_id with ':' -> '_'>.md`), the source PDF, and the extraction run:

- `corpus_id` — document id (`<origin>:<NNNNNN>`); `page`, `chunk` — position within the document
- `title`, `pdf_url` — document title and source PDF URL
- `extractor` — model id that produced the spans; `footnote_link`, `dedupe_overlap` — pipeline flags baked into `input`/`text`
- `has_data_score` — chunk-level data score (1.0 iff spans present)
- `split` — which file the row came from (`train` / `val` / `holdout`)
- `spans` — full span detail the training keys collapse: `[{text, label, score, start, end}]` with char offsets into `input`/`text`

## Splits

Document-disjoint `train` / `val` / `holdout` (70% / 15% / 15%). See `split_stats.json`.

## Usage

```python
from datasets import load_dataset
gliner  = load_dataset("rafmacalaba/datause-extracted", "gliner")
bio     = load_dataset("rafmacalaba/datause-extracted", "bio")
gliner2 = load_dataset("rafmacalaba/datause-extracted", "gliner2")
```

## Files

- `gliner_{train,val,holdout}.jsonl`, `bio_{train,val,holdout}.jsonl`, `gliner2_{train,val,holdout}.jsonl`
- `labels.json` (label lists), `split_stats.json` (per-split counts)