---
configs:
- config_name: uncased
  data_files:
  - path:
    - uncased.jsonl.zst
    split: train
  default: true
- config_name: cased
  data_files:
  - path:
    - cased.jsonl.zst
    split: train
language:
- en
tags:
- grammar
- linguistics
- NLP
license: odc-by
task_categories:
- feature-extraction
---
# Noun Phrases Dataset

This dataset contains noun phrases extracted from [allenai/c4](https://huggingface.co/datasets/allenai/c4).

It includes two configurations: **uncased** and **cased**, with 1 895 908 and 2 000 002 entries, respectively.

### JSONL Fields

Each entry contains:
- **noun_phrase:** The extracted noun phrase.
- **count:** Frequency of occurrence.

#### Example Rows
```json
{"noun_phrase": "ship invoices", "count": 1}
{"noun_phrase": "\"river", "count": 1}
{"noun_phrase": "no boiler system", "count": 1}
```

### Construction

- **Cased Data:** Original noun phrases with preserved casing.
- **Uncased Data:** Constructed by normalizing cased data to lowercase, aggregating counts for identical phrases, and retaining the original phrase with the highest count.

### Preprocessing and Extraction

- [**Preprocessing:**](https://textacy.readthedocs.io/en/latest/api_reference/preprocessing.html) Unicode normalization, bracket removal, bullet point standardization, quotation mark normalization, and whitespace normalization.
- [**SpaCy:**](https://spacy.io/) Noun phrase extraction using the `"en_core_web_sm"` model. Articles such as "a" or "the" were removed.

### Original Data Source

The C4 dataset is a cleaned version of the Common Crawl dataset, providing a comprehensive source of English text for linguistic analysis and NLP tasks.

### Applications

Suitable for grammar analysis, linguistic studies, and natural language processing (NLP) tasks.

### Dataset Details

- **Language:** English (`en`)
- **Licence:** Open Data Commons License