Upload folder using huggingface_hub

README.md

@@ -0,0 +1,76 @@
+---
+language: en
+license: apache-2.0
+base_model: roberta-base
+pipeline_tag: text-classification
+library_name: transformers
+tags:
+  - scientific-text
+  - abstract-sections
+  - citation-analysis
+  - sequential-sentence-classification
+widget:
+  - text: "We conclude that early intervention significantly improves patient outcomes."
+    example_title: "Conclusion sentence"
+  - text: "Participants were randomly assigned to treatment and control groups."
+    example_title: "Methods sentence"
+  - text: "The aim of this study was to evaluate the efficacy of the new vaccine."
+    example_title: "Objective sentence"
+---
+
+# RoBERTa Abstract-Section Classifier
+
+A `roberta-base` model fine-tuned to classify sentences from scientific abstracts into **five rhetorical sections**: `BACKGROUND`, `CONCLUSIONS`, `METHODS`, `OBJECTIVE`, `RESULTS` (the PubMed-RCT-style section scheme).
+
+It was built for a large-scale **citation-fidelity** study, where it selected each cited paper's *claim* sentences — sentences predicted as `CONCLUSIONS` or `RESULTS` (and longer than 50 characters) — from ~13M S2ORC abstracts. Those claims were then compared against citing sentences with the [SPICED](https://huggingface.co/copenlu/spiced) scientific-sentence similarity model to measure how faithfully papers describe the work they cite.
+
+## Labels
+
+| id | label | typical sentence |
+|----|-------------|--------------------------------------------------|
+| 0 | BACKGROUND | "Diabetes is a growing public health concern…" |
+| 1 | CONCLUSIONS | "We conclude that early intervention improves outcomes." |
+| 2 | METHODS | "Participants were randomly assigned to two groups…" |
+| 3 | OBJECTIVE | "The aim of this study was to evaluate…" |
+| 4 | RESULTS | "The treatment group showed a 40% reduction (p < 0.001)." |
+
+## How to use
+
+```python
+from transformers import pipeline
+
+clf = pipeline("text-classification", model="hongccccccc/roberta-abstract-section-classifier")
+clf("The treatment group showed a 40% reduction in mortality compared with placebo.")
+# [{'label': 'RESULTS', 'score': 0.99}]
+```
+
+In the original pipeline, inference used `truncation=True, max_length=50`; sentences are short, so this rarely truncates.
+
+## Training
+
+- **Base model:** [`roberta-base`](https://huggingface.co/roberta-base), fine-tuned with `RobertaForSequenceClassification` (single-label, 5 classes).
+- **Task/data:** abstract-sentence section classification following the PubMed-RCT label scheme.
+- **Trained:** January 2023, `transformers` 4.12.5. The original `training_args.bin` is included in this repo for provenance; the training script itself was not preserved, so exact hyperparameters and the precise training split are unknown.
+
+## Evaluation
+
+No held-out evaluation from the original training survives. As a release sanity check, the model correctly classified a small battery of unambiguous section sentences (see the examples above) with high confidence. Treat downstream metrics as unverified and evaluate on your own data before critical use.
+
+## Limitations
+
+- Trained on **abstract** sentences from scientific (largely biomedical-style) papers; full-text sentences or other domains may degrade accuracy.
+- Single-sentence input; it does not use surrounding-sentence context, which sequential models exploit for this task.
+- Label ids in `config.json` were reconstructed from the inference code (`sci_parser.py`) of the original project and verified on sample sentences.
+
+## Citation
+
+If you use this model, please cite the citation-fidelity paper (to appear — citation forthcoming). Related resources:
+
+```bibtex
+@inproceedings{wright2022modeling,
+  title={Modeling Information Change in Science Communication with Semantically Matched Paraphrases},
+  author={Wright, Dustin and Pei, Jiaxin and Jurgens, David and Augenstein, Isabelle},
+  booktitle={EMNLP},
+  year={2022}
+}
+```

config.json

@@ -0,0 +1,44 @@
+{
+  "add_cross_attention": false,
+  "architectures": [
+    "RobertaForSequenceClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "dtype": "float32",
+  "eos_token_id": 2,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "id2label": {
+    "0": "BACKGROUND",
+    "1": "CONCLUSIONS",
+    "2": "METHODS",
+    "3": "OBJECTIVE",
+    "4": "RESULTS"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "is_decoder": false,
+  "label2id": {
+    "BACKGROUND": 0,
+    "CONCLUSIONS": 1,
+    "METHODS": 2,
+    "OBJECTIVE": 3,
+    "RESULTS": 4
+  },
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 514,
+  "model_type": "roberta",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 1,
+  "position_embedding_type": "absolute",
+  "problem_type": "single_label_classification",
+  "tie_word_embeddings": true,
+  "transformers_version": "5.11.0",
+  "type_vocab_size": 1,
+  "use_cache": true,
+  "vocab_size": 50265
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1fc7207aada50f687c5ea1d087489c01c1ce9e69df1dccbf3c6511d4c0909246
+size 498622052

tokenizer_config.json

@@ -0,0 +1,17 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": "<s>",
+  "cls_token": "<s>",
+  "eos_token": "</s>",
+  "errors": "replace",
+  "is_local": false,
+  "local_files_only": false,
+  "mask_token": "<mask>",
+  "model_max_length": 512,
+  "pad_token": "<pad>",
+  "sep_token": "</s>",
+  "tokenizer_class": "RobertaTokenizer",
+  "trim_offsets": true,
+  "unk_token": "<unk>"
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e223b6115aa22b61c8a1d9ab03437de7b16d25c427b38c35ad75b32d7a9f54b
+size 2799