README.md

---
pipeline_tag: question-answering
tags:
- feature-extraction
language:
- en
license: afl-3.0
---

[Extractive Question Answering application.](https://raphaelsty.github.io/blog/template/)

```python
import re

from transformers import pipeline

qa = pipeline(
    "question-answering", model="raphaelsty/carbonblog", tokenizer="raphaelsty/carbonblog"
)

def clean(document):
    """Pre-process the document."""
    document = re.sub("[^a-zA-Z0-9 \n\.]", " ", document)
    document = re.sub("\s\s+", " ", document)
    # [NONE] allows the model to handle missing components.
    document = f"[NONE] {document.lower()}"
    return document
```

```python
document = clean("body: 83% nylon 17% spandex; lace: 86% nylon 14% spandex")
```

```python
qa({"question": "What is the 1 component ?", "context": document})
```

```json
{"score": 0.9999976754188538, "start": 32, "end": 36, "answer": "lace"}
```