Update README.md

README.md

@@ -1,22 +1,57 @@
 ---
 language:
+
 - es
+
 license: apache-2.0
+
 tags:
+
 - "national library of spain"
+
 - "spanish"
+
 - "bne"
+
 - "capitel"
+
 - "ner"
+
 datasets:
+
 - "bne"
+
 - "capitel"  
+
+
 metrics:
+
 - "f1"
+
 inference:
   parameters:
     aggregation_strategy: "first"
 
+model-index:
+- name: roberta-base-bne-capiter-ner-plus
+  results:
+  - task: 
+      type: token-classification
+    dataset:
+      type:  ner
+      name: CAPITEL-NERC
+    metrics:
+      - name: F1
+        type: f1
+        value: 0.8960
+        
+
+widget:
+
+- "Me llamo francisco javier y vivo en madrid."
+
+- "Mi hermano ramón y su mejor amigo luis trabajan en el bsc."
+
 ---
 
 # Spanish RoBERTa-base trained on BNE finetuned for CAPITEL Named Entity Recognition (NER) dataset.
@@ -30,7 +65,13 @@ inference:
 - [How to use](#how-to-use)
 - [Limitations and bias](#limitations-and-bias)
 - [Training](#training)
+- [Training](#training)
+  - [Training data](#training-data)
+  - [Training procedure](#training-procedure)
 - [Evaluation](#evaluation)
+- [Evaluation](#evaluation)
+   - [Variable and metrics](#variable-and-metrics)
+   - [Evaluation results](#evaluation-results)
 - [Additional information](#additional-information)
   - [Author](#author)
   - [Contact information](#contact-information)
@@ -43,37 +84,55 @@ inference:
 </details>
 
 ## Model description
-RoBERTa-base-bne is a transformer-based masked language model for the Spanish language. It is based on the [RoBERTa](https://arxiv.org/abs/1907.11692) base model and has been pre-trained using the largest Spanish corpus known to date, with a total of 570GB of clean and deduplicated text processed for this work, compiled from the web crawlings performed by the  [National Library of Spain (Biblioteca Nacional de España)](http://www.bne.es/en/Inicio/index.html) from 2009 to 2019.
-
-Original pre-trained model can be found here: https://huggingface.co/PlanTL-GOB-ES/roberta-base-bne
-
+The **roberta-base-bne-capitel-ner-plus** is Named Entity Recognition (NER) model for the Spanish language fine-tuned from the [roberta-base-bne](https://huggingface.co/PlanTL-GOB-ES/roberta-base-bne) model, a [RoBERTa](https://arxiv.org/abs/1907.11692) base model pre-trained using the largest Spanish corpus known to date, with a total of 570GB of clean and deduplicated text, processed for this work, compiled from the web crawlings performed by the [National Library of Spain (Biblioteca Nacional de España)](http://www.bne.es/en/Inicio/index.html) from 2009 to 2019. This model is a more robust version of the [roberta-base-bne-capitel-ner](https://huggingface.co/PlanTL-GOB-ES/roberta-base-bne-capitel-ner) model that recognizes better lowercased Named Entities (NE).
 
 ## Intended uses and limitations
 
+**roberta-base-bne-capitel-ner-plus** model can be used to recognize Named Entities (NE). The model is limited by its training dataset and may not generalize well for all use cases.
+
 ## How to use
 
+```python
+from transformers import pipeline
+from pprint import pprint
+
+nlp = pipeline("ner", model="PlanTL-GOB-ES/roberta-base-bne-capitel-ner-plus")
+example = "Me llamo francisco javier y vivo en madrid."
+
+ner_results = nlp(example)
+pprint(ner_results)
+```
 ## Limitations and bias
 At the time of submission, no measures have been taken to estimate the bias embedded in the model. However, we are well aware that our models may be biased since the corpora have been collected using crawling techniques on multiple web sources. We intend to conduct research in these areas in the future, and if completed, this model card will be updated.
 
 ## Training
-The dataset used is the one from the [CAPITEL competition at IberLEF 2020](https://sites.google.com/view/capitel2020) (sub-task 1).
 
-**IMPORTANT ABOUT THIS MODEL:** We modified the dataset to make this model more robust to general Spanish input. In the Spanish language all the name entities are capitalized, as this dataset has been elaborated by experts, it is totally correct in terms of Spanish language. We randomly took some entities and we lower-cased some of them for the model to learn not only that the named entities are capitalized, but also the structure of a sentence that should contain a named entity. For instance: "My name is [placeholder]", this [placeholder] should be a named entity even though it is not written capitalized. The model trained on the original capitel dataset can be found here: https://huggingface.co/PlanTL-GOB-ES/roberta-base-bne-capitel-ner
+The dataset used for training and evaluation is the one from the [CAPITEL competition at IberLEF 2020](https://sites.google.com/view/capitel2020) (sub-task 1). We lowercased and uppercased the dataset, and added the additional sentences to the training.
+
+### Training procedure
+The model was trained with a batch size of 16 and a learning rate of 1e-5 for 5 epochs. We then selected the best checkpoint using the downstream task metric in the corresponding development set and then evaluated it on the test set.
 
-Examples:
+## Evaluation 
 
-This model:
-- "Me llamo asier y vivo en barcelona todo el año." → "Me llamo {asier:S-PER} y vivo en {barcelona:S-LOC} todo el año."
-- "Hoy voy a visitar el parc güell tras salir del barcelona supercomputing center." → "Hoy voy a visitar el {park:B-LOC} {güell:E-LOC} tras salir del {barcelona:B-ORG} {supercomputing:I-ORG} {center:E-ORG}."
 
-Model trained on original data:
-- "Me llamo asier y vivo en barcelona todo el año." → "Me llamo asier y vivo en barcelona todo el año." (nothing)
-- "Hoy voy a visitar el parc güell tras salir del barcelona supercomputing center." → "Hoy voy a visitar el parc güell tras salir del barcelona supercomputing center." (nothing)
+### Variable and metrics
+This model was finetuned maximizing F1 score.
 
-## Evaluation 
-F1 Score: 0.8867
+## Evaluation results
+We evaluated the *roberta-base-bne-capitel-ner-plus** on the CAPITEL-NERC test set against standard multilingual and monolingual baselines:
+
+
+| Model        | XNLI (Accuracy) | 
+| ------------|:----|
+| roberta-large-bne-capitel-ner | **90.51** |
+| roberta-base-bne-capitel-ner | 89.60|
+| roberta-base-bne-capitel-ner-plus | 89.60|
+| BETO       | 87.72 |
+| mBERT       | 88.10 |
+| BERTIN | 88.56 |
+| ELECTRA | 80.35 |
 
-For evaluation details visit our [GitHub repository](https://github.com/PlanTL-GOB-ES/lm-spanish).
+For more details, check the fine-tuning and evaluation scripts in the official [GitHub repository](https://github.com/PlanTL-GOB-ES/lm-spanish).
 
 
 ## Additional information