Update README.md (#3)

- Update README.md (6602198dcbb24d8bba04b4a60f058ba6f7600e9d)

README.md

@@ -3,131 +3,182 @@ license: apache-2.0
 base_model: openai-community/gpt2
 pipeline_tag: token-classification
 tags:
-- nlp
+- gpt2
 - token-classification
-- named-entity-recognition
+- nlp
 - event-extraction
+- eventx
+- sequence-labeling
+datasets:
+- mdg-nlp/eventx-recognition-original
 metrics:
 - f1
 ---
 
-# gpt-2-eventx-sentence-recognition
+# GPT-2 Fine-Tuned for Event Expression Extraction (EVENTX)
+
+## Overview
 
-## 🧠 Overview
+This model is a fully fine-tuned version of `openai-community/gpt2` for **Event Expression Extraction (EVENTX)**.
 
-This model is a **fully fine-tuned version** of `openai-community/gpt2` for **Event Expression Extraction**.
+The model identifies spans in natural language text that correspond to events, including actions, occurrences, processes, and states expressed in a sentence.
 
-It identifies event expressions in natural language text using BIO sequence labeling.
+The task is formulated as token-level sequence labeling using BIO tags.
+
+### Label Map
 
-### Labels:
-- **B-E**
-- **I-E**
-- **O**
+| ID | Label | Description |
+|---|---|---|
+| 0 | O | Token is outside an event expression |
+| 1 | B-E | Beginning of an event expression |
+| 2 | I-E | Inside an event expression |
 
 ---
-## 📊 Task
 
-**Event Expression Extraction**
+## Task
 
-Given an input sequence, the model identifies event expression spans using BIO token labels.
+**Event Expression Extraction (EVENTX)**
+
+Given an input sentence, the model identifies spans corresponding to event expressions.
+
+For example:
+
+```text id="2nxbq8"
+Input:
+The company announced the acquisition on Monday.
+
+Event expressions:
+announced
+acquisition
+```
+
+Each token is assigned one of three BIO labels:
+
+- **O**: outside an event expression
+- **B-E**: beginning of an event expression
+- **I-E**: inside an event expression
+
+Event expression extraction provides structured information about events mentioned in text and can support downstream temporal reasoning, event ordering, relation extraction, and timeline construction.
 
 ---
-## 📚 Dataset
 
-- **Name:**
-- **Format:** CoNLL (token + label)
-- **Labels:** B-E, I-E, O
-- **Distribution:**
+## Model
 
-### Label Map
+GPT-2 is a decoder-only transformer architecture. This model uses GPT-2 with a token-classification head and is fully fine-tuned for sentence-level event expression extraction.
 
-- **0**: B-E
-- **1**: I-E
-- **2**: O
+The model is one of the fine-tuned pretrained language model baselines used to study multi-dimensional generalization in temporal extraction tasks.
 
 ---
-## ⚙️ Training Details
-
-- **Model:** openai-community/gpt2
-- **Fine-tuning:** Full fine-tuning (all parameters updated)
-- **Epochs:** 3
-- **Learning Rate:** 2e-5
-- **Train Batch Size:** 16
-- **Evaluation Batch Size:** 32
-- **Optimizer:** AdamW
-- **Weight Decay:** 0.01
-- **Warmup Steps:** 500
-- **Max Sequence Length:** 128
-- **Evaluation:** Every 250 steps
-- **Early Stopping:** Patience of 3 evaluations
-- **Model Selection:** Entity-level F1
+
+## Dataset
+
+**Dataset:** `mdg-nlp/eventx-recognition-original`
+
+### Labels
+
+```text id="i9llpo"
+0 -> O
+1 -> B-E
+2 -> I-E
+```
+
+---
+
+## Training Details
+
+| Configuration | Value |
+|---|---|
+| Base Model | `openai-community/gpt2` |
+| Task | Event Expression Extraction |
+| Formulation | Token Classification |
+| Label Scheme | BIO |
+| Fine-tuning | Full fine-tuning |
+| Epochs | 3 |
+| Learning Rate | 2e-5 |
+| Optimizer | AdamW |
+| Weight Decay | 0.01 |
+| Warmup Steps | 500 |
+| Train Batch Size | 16 |
+| Evaluation Batch Size | 32 |
+| Max Sequence Length | 128 |
 
 ---
-## 📈 Evaluation
 
-Scores are reported using **F1**.
+## Evaluation
+
+All reported scores use **micro-F1**.
 
-| Evaluation Setting | F1 |
+| Evaluation Setting | Micro-F1 |
 |---|---:|
-| Base | 0.84 |
+| Base | **0.84** |
 | T2C | 0.18 |
-| Adv | 0.78 |
+| Adversarial (Adv) | 0.78 |
 | Length | 0.85 |
-| Voc | 0.84 |
-| Adv_P | 0.50 |
-
----
-## 🔬 Analysis
+| Vocabulary (Voc) | 0.84 |
+| Adversarial Perturbation (Adv_P) | 0.50 |
 
-The model achieves an F1 score of **0.84** on the base evaluation setting.
-Performance across T2C, adversarial, length, vocabulary, and adversarial perturbation settings provides a broader view of model generalization.
+The evaluation settings measure generalization across the original task, cross-domain data, adversarial examples, structural changes, vocabulary variation, and adversarial perturbations.
 
 ---
-## 🚀 Usage
 
-```python
-from transformers import AutoTokenizer, AutoModelForTokenClassification, pipeline
+## Usage
+
+```python id="fjbd21"
+from transformers import AutoTokenizer, AutoModelForTokenClassification
 
 model_id = "mdg-nlp/gpt-2-eventx-sentence-recognition"
 
 tokenizer = AutoTokenizer.from_pretrained(model_id)
 model = AutoModelForTokenClassification.from_pretrained(model_id)
 
-classifier = pipeline("token-classification", model=model, tokenizer=tokenizer)
-text = "The patient arrived on Monday and underwent surgery the next day."
+text = "The company announced the acquisition on Monday."
+
+inputs = tokenizer(text, return_tensors="pt", truncation=True)
+outputs = model(**inputs)
+
+predictions = outputs.logits.argmax(dim=-1)[0].tolist()
+
+id2label = {
+    0: "O",
+    1: "B-E",
+    2: "I-E"
+}
 
-print(classifier(text))
+tokens = tokenizer.convert_ids_to_tokens(inputs["input_ids"][0])
+
+for token, prediction in zip(tokens, predictions):
+    print(token, id2label[prediction])
 ```
 
 ---
-## 📊 Training Logs
 
-**Weights & Biases run:**
+## Limitations
 
+- Performance may decrease under distribution shifts.
+- Cross-domain event expressions may be more difficult to identify.
+- Adversarial perturbations can affect extraction performance.
+- Tokenization can affect predicted event boundaries.
+- Longer or structurally different inputs may produce different behavior.
+- Ambiguous or implicit events may not be identified correctly.
+- The model extracts event expressions and does not directly perform event relation or temporal relation classification.
 
 ---
-## 🧩 Limitations
 
-- Performance depends on dataset distribution
-- May not generalize well to unseen domains
-- Sequence truncation may affect long inputs
-- Performance can vary across different generalization settings
+## Research Context
 
----
-## 🔬 Research Context
+This model was developed as part of research on **multi-dimensional generalization in temporal extraction tasks**.
 
-This model was developed as part of research on:
-- Temporal reasoning in NLP
-- Temporal and event expression extraction
-- Generalization of language models
+The broader work studies temporal and event expression extraction under different forms of distribution shift, including domain shift, adversarial perturbations, vocabulary variation, compositional changes, and structural changes.
 
 ---
-## 📬 Contact
 
-**Author:** Arnav Verma
-**Affiliation:** New York University
+## Contact
+
+  **Author:** Arnav Verma  <br>
+  **Affiliation:** New York University <br>
 
 ---
-## ⭐ Citation
 
-If you use this model in your research, please cite appropriately.
+## Citation
+
+If you use this model in your research, please cite the associated work when available.