---
license: other
license_name: arxiv-abstracts-various
license_link: https://info.arxiv.org/help/api/tou.html
task_categories:
- token-classification
language:
- en
tags:
- accessibility
- rhetorical-structure
- scientific-papers
- weak-supervision
pretty_name: Semantic Highlighting Rhetorical Roles
---

# Semantic Highlighting: Rhetorical Roles in arXiv Abstracts

Rhetorical-role span annotations for ML/NLP arXiv abstracts, built for a
semantic highlighting accessibility tool (dyslexia/ADHD reading support) --
see the [project repo](https://github.com/Ryan-Hancock/semantic-highlighting-improve-reading-accessibility)
for the full research writeup (`docs/findings.md`).

## Labels

Each span is tagged with one of 7 roles: **Contribution**, **Method**,
**Result**, **Evidence**, **Limitation**, **FutureWork**, **Safety**. Full
definitions, cue phrases, and boundary rules in
[`docs/annotation_guideline.md`](https://github.com/Ryan-Hancock/semantic-highlighting-improve-reading-accessibility/blob/main/docs/annotation_guideline.md).

## Schema

```
{
  "arxiv_id": str,
  "title": str,
  "text": str,               # the abstract
  "spans": [
    {"start": int, "end": int, "label": str, "text": str},
    ...
  ]
}
```
`start`/`end` are character offsets into `text`. Spans within a record do
not overlap.

## Splits

- **train**: silver-labeled abstracts (LLM weak-labeled, see Methodology),
  excluding every abstract that appears in `test`.
- **validation**: a small held-out slice of the same silver pool, used only
  to watch for overfitting during training -- also silver-labeled, so treat
  it as noisy, not as a second gold set.
- **test**: hand-corrected gold examples. The only split with human review.

## Methodology

1. Abstracts fetched from the arXiv API (`cat:cs.LG`, `cat:cs.CL`, plus
   `cat:cs.RO AND abs:safety` and `abs:"future work"`-biased queries added
   specifically because Safety/FutureWork are rare under random sampling).
2. Weak-labeled by a local LLM (Ollama, `ministral-3:latest`) prompted with
   the annotation guideline, constrained to valid JSON matching the label
   schema. Every proposed span is checked against the source text
   character-for-character; anything the model didn't quote verbatim is
   dropped rather than guessed at (~9% drop rate).
3. A subset hand-corrected against the guideline to build `test`.
   Systematic error patterns found during correction (Evidence tagged
   without any number, FutureWork tagged on plain findings with no
   forward-looking language, overlapping/duplicate spans) were then
   propagated across the *entire* silver corpus via automated rules in
   `scripts/clean_silver.py` -- not just the hand-corrected subset.

Full detail, including exact error rates and what didn't work, in the
project's [findings log](https://github.com/Ryan-Hancock/semantic-highlighting-improve-reading-accessibility/blob/main/docs/findings.md).

## Known limitations

- Silver labels are LLM-generated and imperfect even after automated
  cleaning; only `test` has human review.
- Skewed label distribution: Method/Result/Contribution/Evidence are common,
  Safety/FutureWork/Limitation are rare even after deliberately biasing
  collection toward them.
- `test` is small (21 abstracts) -- any model evaluated against it should be
  treated as a directional signal, not a precise number.

## Licensing

Abstract text is sourced from arXiv under arXiv's
[API Terms of Use](https://info.arxiv.org/help/api/tou.html); individual
papers retain their own licenses (arXiv itself does not claim copyright over
submitted abstracts). The span annotations (labels, offsets) added by this
project are original work.