nlpscu/Analyzing-Demographic-Biases

---
language:
- en
task_categories:
- text-classification
task_ids:
- text-scoring
pretty_name: PERSUADE 2.0 & ASAP 2.0 Essay Scoring Datasets
size_categories:
- 10K<n<100K
license: other
tags:
- essay-scoring
- automated-essay-scoring
- education
- nlp
---
# PERSUADE 2.0 & ASAP 2.0 — Essay Scoring Datasets

Datasets for our NLP class project on Automated Essay Scoring (AES), combining two widely used student essay corpora.

---

## PERSUADE 2.0

A large-scale corpus of argumentative and persuasive student essays (grades 6–12), annotated with discourse elements and holistic essay scores (1–6).

**Files:**
- `persuade_corpus_2.0_train.csv`
- `persuade_corpus_2.0_test.csv`

**Key columns:**

| Column | Description |
|---|---|
| `essay_id` | Unique essay identifier |
| `full_text` | Raw essay text |
| `holistic_essay_score` | Overall score (1–6) |
| `prompt_name` | Writing prompt |
| `grade_level` | Student grade (6–12) |

---

## ASAP 2.0

An updated version of the Automated Student Assessment Prize benchmark. Essays span 8 distinct prompts (argumentative, narrative, source-dependent) and are scored by human raters on prompt-specific rubrics.

**Files:**
- `ASAP_2_Final_github_train.csv`
- `ASAP_2_Final_github_test.csv`
- `ASAP2_train_sourcetexts.csv` — source texts referenced by source-dependent prompts

**Key columns:**

| Column | Description |
|---|---|
| `essay_id` | Unique essay identifier |
| `essay_set` | Prompt ID (1–8) |
| `full_text` | Raw essay text |
| `domain1_score` | Primary human rater score |
| `final_score` | Resolved score used for evaluation |