Update README.md

README.md

@@ -2,8 +2,6 @@
 license: cc-by-nc-4.0
 tags:
 - generated_from_trainer
-language: 
-- es
 metrics:
 - precision
 - recall
@@ -13,9 +11,8 @@ model-index:
 - name: roberta-es-clinical-trials-neg-spec
   results: []
 widget:
-- text: "Pacientes sanos, sin ninguna enfermedad, que no tomen medicamentos"
+- text: "Pacientes sanos, sin ninguna enfermedad, que no tomen ningún medicamento"
 - text: "Sujetos adultos con cáncer de próstata asintomáticos y no tratados previamente"
-- text: "Enfermedades con posibles síntomas de urticaria o angioedema"
 ---
 
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
@@ -25,21 +22,21 @@ should probably proofread and complete it, then remove this comment. -->
 
 This named entity recognition model detects negation and speculation entities, and negated and speculated concepts:
 - Neg_cue: negation cue (e.g. *no*, *sin*)
-- Negated: negated entity or event (e.g. *sin **dolor***)
+- Negated: negated entity or event (e.g. *sin* **dolor**)
 - Spec_cue: speculation cue (e.g. *posiblemente*)
-- Speculated: speculated entity or event (e.g. *posiblemente **sobreviva***)
+- Speculated: speculated entity or event (e.g. *posiblemente* **sobreviva**)
 
 The model achieves the following results on the test set (when trained with the training and development set; results are averaged over 5 evaluation rounds):
-- Precision: 0.838 (±0.003)
+- Precision: 0.840 (±0.003)
 - Recall: 0.866 (±0.005)
-- F1: 0.852 (±0.003)
-- Accuracy: 0.986 (±0.001)
+- F1: 0.853 (±0.004)
+- Accuracy: 0.985 (±0.001)
 
 ## Model description
 
 This model adapts the pre-trained model [bsc-bio-ehr-es](https://huggingface.co/PlanTL-GOB-ES/bsc-bio-ehr-es), presented in [Pio Carriño et al. (2022)](https://aclanthology.org/2022.bionlp-1.19/). 
 It is fine-tuned to conduct medical named entity recognition on Spanish texts about clinical trials. 
-The model is fine-tuned on the [NUBEs corpus (Lima et al. 2020)](https://aclanthology.org/2020.lrec-1.708/) and on the [CT-EBM-SP corpus (Campillos-Llanos et al. 2021)](https://bmcmedinformdecismak.biomedcentral.com/articles/10.1186/s12911-021-01395-z).
+The model is fine-tuned on the [NUBEs corpus (Lima et al. 2020)](https://aclanthology.org/2020.lrec-1.708/) and on the [CT-EBM-ES corpus (Campillos-Llanos et al. 2021)](https://bmcmedinformdecismak.biomedcentral.com/articles/10.1186/s12911-021-01395-z).
 
 ## Intended uses & limitations
 
@@ -64,15 +61,15 @@ El propietario o creador de los modelos de ningún modo será responsable de los
 
 The data used for fine-tuning are:
 
-1) The [Negation and Uncertainty in Spanish Corpus (NUBes)](https://github.com/Vicomtech/NUBes-negation-uncertainty-biomedical-corpus):
+1) The [Negation and Uncertainty in Spanish Corpus (NUBes)](https://github.com/Vicomtech/NUBes-negation-uncertainty-biomedical-corpus)
 It is a collection of 29 682 sentences (518 068 tokens) from anonymised health records in Spanish, annotated with negation and uncertainty cues and their scopes.
 
-2) The [Clinical Trials for Evidence-Based-Medicine in Spanish corpus](http://www.lllf.uam.es/ESP/nlpdata/wp2/):
+2) The [Clinical Trials for Evidence-Based-Medicine in Spanish corpus](http://www.lllf.uam.es/ESP/nlpdata/wp2/).
 It is a collection of 1200 texts about clinical trials studies and clinical trials announcements:
 - 500 abstracts from journals published under a Creative Commons license, e.g. available in PubMed or the Scientific Electronic Library Online (SciELO)
 - 700 clinical trials announcements published in the European Clinical Trials Register and Repositorio Español de Estudios Clínicos
 
-If you use the CT-EBM-SP resource, please, cite as follows:
+If you use the CT-EBM-ES resource, please, cite as follows:
 
 ```
 @article{campillosetal-midm2021,
@@ -100,24 +97,24 @@ The following hyperparameters were used during training:
 - seed: we used different seeds for 5 evaluation rounds, and uploaded the model with the best results
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: linear
-- num_epochs: 8
+- num_epochs: average 10.5 epochs (±1.9); trained with early stopping if no improvement after 5 epochs (early stopping patience: 5)
 
 
 ### Training results (test set; average and standard deviation of 5 rounds with different seeds)
 
 |   Precision    |     Recall     |       F1       |    Accuracy    |
 |:--------------:|:--------------:|:--------------:|:--------------:|
-| 0.838 (±0.003) | 0.866 (±0.005) | 0.852 (±0.003) | 0.986 (±0.001) |
+| 0.840 (±0.003) | 0.866 (±0.005) | 0.853 (±0.004) | 0.985 (±0.001) |
 
 
 **Results per class (test set; average and standard deviation of 5 rounds with different seeds)**  
 
 | Class       |   Precision    |     Recall     |       F1       |  Support  |
 |:-----------:|:--------------:|:--------------:|:--------------:|:---------:|
-| Neg_cue     | 0.945 (±0.004) | 0.961 (±0.002) | 0.953 (±0.003) |    2416   |
-| Negated     | 0.815 (±0.003) | 0.838 (±0.005) | 0.826 (±0.003) |    3064   |
-| Spec_cue    | 0.811 (±0.005) | 0.868 (±0.009) | 0.839 (±0.005) |     746   |
-| Speculated  | 0.685 (±0.009) | 0.719 (±0.016) | 0.701 (±0.008) |     993   |
+| Neg_cue     | 0.938 (±0.004) | 0.963 (±0.003) | 0.950 (±0.002) |    2436   |
+| Negated     | 0.799 (±0.018) | 0.843 (±0.008) | 0.820 (±0.010) |    3086   |
+| Spec_cue    | 0.821 (±0.021) | 0.852 (±0.015) | 0.836 (±0.008) |     749   |
+| Speculated  | 0.710 (±0.002) | 0.721 (±0.010) | 0.715 (±0.005) |     996   |
 
 
 ### Framework versions