Update descriptions

app.py

@@ -19,31 +19,31 @@ metric = ["f1_macro", "accuracy", "weighted_f1"]
 
 
 def app():
-    with gr.Blocks(theme=gr.themes.Soft(primary_hue="pink", secondary_hue="emerald")) as demo:
-        gr.Markdown(open("pages/header.md", "r").read())
+    with gr.Blocks(theme=gr.themes.Soft(primary_hue="pink", secondary_hue="purple")) as demo:
+        gr.Markdown(open("pages/header.md", "r").read(), container=True)
 
         with gr.Tabs():
             with gr.Tab("🏆 Leaderboard", elem_classes='tab-item'):
                 with gr.Tab("Overall Results", elem_classes='tab-item'):
-                    gr.Markdown(open("pages/overall-results.md", "r").read(), elem_classes='tab-item')
-                    overall_table = gr.Dataframe()
+                    gr.Markdown(open("pages/overall-results.md", "r").read(), elem_classes='tab-item', container=True)
+                    overall_table = gr.Dataframe(show_row_numbers=True, pinned_columns=2)
 
                 with gr.Tab("Results per Language", elem_classes='tab-item'):
-                    gr.Markdown(open("pages/results-per-language.md", "r").read(), elem_classes='tab-item')
+                    gr.Markdown(open("pages/results-per-language.md", "r").read(), elem_classes='tab-item', container=True)
                     languages_filter = gr.CheckboxGroup(choices=languages, label="Select columns", value=languages)
                     select_lang_metric = gr.Radio(metric, value='f1_macro', label="Select metric")
-                    lang_table = gr.Dataframe()
+                    lang_table = gr.Dataframe(show_row_numbers=True, pinned_columns=2)
 
                 with gr.Tab("Results per Dataset", elem_classes='tab-item'):
-                    gr.Markdown(open("pages/results-per-dataset.md", "r").read(), elem_classes='tab-item')
+                    gr.Markdown(open("pages/results-per-dataset.md", "r").read(), elem_classes='tab-item', container=True)
                     dataset_filter = gr.CheckboxGroup(choices=datasets, label="Select columns", value=datasets)
                     select_ds_metric = gr.Radio(metric, value='f1_macro', label="Select metric")
-                    dataset_table = gr.Dataframe()
+                    dataset_table = gr.Dataframe(show_row_numbers=True, pinned_columns=2)
 
                 with gr.Tab("Results per Emotion", elem_classes='tab-item'):
-                    gr.Markdown(open("pages/results-per-emotion.md", "r").read(), elem_classes='tab-item')
+                    gr.Markdown(open("pages/results-per-emotion.md", "r").read(), elem_classes='tab-item', container=True)
                     emo_filter = gr.CheckboxGroup(choices=emotions, label="Select columns", value=emotions)
-                    emotion_table = gr.Dataframe()
+                    emotion_table = gr.Dataframe(show_row_numbers=True, pinned_columns=2)
 
                 df_state = gr.State()
 
@@ -88,7 +88,7 @@ def app():
             with gr.Tab("🔢 Evaluate your model", elem_classes='tab-item'):
                 gr.Markdown(open("pages/evaluate.md", "r").read(), elem_classes='tab-item')
             
-            with gr.Tab("🚀 Submit here!", elem_classes='tab-item'):
+            with gr.Tab("📬 Submit here!", elem_classes='tab-item'):
                 gr.Markdown(open("pages/submit.md", "r").read(), elem_classes='tab-item')
         
         with gr.Column():

pages/about.md

@@ -1,14 +1,25 @@
 # 📝 About
 
-The **CAMEO Leaderboard** is part of the **Speech Emotion Recognition (SER)** benchmark, which is performed on the **CAMEO** dataset.  
-<br>
+**CAMEO** (**C**ollection of Multilingu**a**l E**m**otional Sp**e**ech C**o**rpora) is a benchmark 
+dataset designed to support research in **Speech Emotion Recognition (SER)** — especially in 
+multilingual and cross-lingual settings.
 
-## Evaluation
-As an input, the model accepts an audio sample. On the output, return an emotion label.  After the response given by the model, post processing takes place to find the nearest label from the dataset. 
+The collection brings together **13 emotional speech datasets** covering **8 languages**, including 
+English, German, Spanish, French, Serbian, and more. In total, it contains **41,265 audio samples**, 
+with each sample annotated for **emotion**, and in most cases, also for **speaker ID**, **gender**, 
+and **age**.
 
-After receiving the label, `f1_macro`, `accuracy` and `weighted_f1` are counted.   
-<br>
+Here are a few quick facts about the dataset:
+- Over **33%** of the samples are in English.
+- **17 distinct emotional states** are represented across datasets.
+- **93.5%** of samples fall under the **seven primary emotions**: *neutral*, *anger*, *sadness*, 
+*surprise*, *happiness*, *disgust*, and *fear*.
+- **Gender annotations** are available for over **92%** of samples.
 
-## Software Library
+All datasets included in **CAMEO** are openly available. We've made the full collection accessible 
+on Hugging Face, along with metadata, tools, and a leaderboard for evaluation.
 
-The dataset is available on the HuggingFace ([here](https://huggingface.co/datasets/amu-cai/CAMEO)).
+🔗 [View the **CAMEO** Dataset on Hugging Face](https://huggingface.co/datasets/amu-cai/CAMEO)
+
+Whether you're building SER models or exploring emotion understanding across languages, **CAMEO** 
+is here to support your research.

pages/header.md

@@ -1 +1,9 @@
-# Open Speech Emotion Recognition Leaderboard
+# Open Speech Emotion Recognition Leaderboard
+
+Welcome to the **Open SER Leaderboard** — part of the [**CAMEO**](https://huggingface.co/datasets/amu-cai/CAMEO) project!
+
+This leaderboard tracks how well different models recognize emotions in speech across multiple languages.
+
+Everything is open, transparent, and reproducible — you're invited to explore, evaluate, and contribute.
+
+<br><br>

pages/overall-results.md

@@ -1 +1,5 @@
-The results show the values of the `f1_macro`, `weighted_f1` and `accuracy` metrics between the expected and generated emotion labels by each model with the given `temperature`.
+This tab shows how different models perform across the entire **CAMEO** collection. You’ll find 
+**macro F1**, **weighted F1**, and **accuracy** scores for each model, tested at different 
+temperature settings.
+
+It's a great place to get a quick overview of how models compare on the full dataset.

pages/results-per-dataset.md

@@ -1 +1,5 @@
-The results show the values of the `f1_macro`, `weighted_f1` or `accuracy` metrics between the expected and generated emotion labels by each model with a given `temperature` for each `dataset` in the CAMEO dataset.
+This tab breaks down results by individual datasets included in the **CAMEO** collection. You can 
+choose which datasets to view and which metric to focus on.
+
+It’s helpful for spotting differences in performance, potential data overlap, or just understanding 
+how models behave on different kinds of emotional speech.

pages/results-per-emotion.md

@@ -1 +1,5 @@
-The results show the values of the `f1_macro` metrics between the expected and generated emotion labels by each model with a given `temperature` for each `emotion` in the CAMEO dataset.
+Which emotions are easier for models to recognize — and which ones still trip them up? This view 
+shows how models perform on specific emotional states.
+
+Pick the emotions and metric you’re interested in, and see which models handle them best. It's a 
+great tool for digging deeper into model behavior.

pages/results-per-language.md

@@ -1 +1,7 @@
-The results show the values of the `f1_macro`, `weighted_f1` or `accuracy` metrics between the expected and generated emotion labels by each model with a given `temperature` for each `language` in the CAMEO dataset.
+Curious how models handle different languages? This view lets you compare performance across 
+languages like English, French, German, and more. Use the checkboxes to pick which languages you 
+want to see, and switch between metrics like **macro F1**, **weighted F1**, or **accuracy** using 
+the radio buttons.
+
+This is especially useful if you’re working on multilingual models or looking to improve 
+performance in a specific language.

pages/submit.md

@@ -1,15 +1,17 @@
-# ✉️✨ Submit Your Model Here! ✨✉️
+# 📬 Submit Here!
 
-Help us improve the leaderboard by submitting your model.  
+Want your model to appear on the leaderboard?
 
-<br>
+Send us an email at [[email protected]](mailto:[email protected]?subject=CAMEO%20Leaderboard%20Submission) with the subject line **"CAMEO Leaderboard Submission"**.
 
-## 📌 How to Submit Your Model:  
+Please include:
+- Your model's name and a short description.
+- The temperature setting you used.
+- A JSONL file with your predictions.
+- Any other details you'd like to share.
 
-✉️ **Step 1:** Send an email to [`[email protected]`](mailto:[email protected]).  
+If you **don’t have access to the resources** needed to run the evaluation yourself, no 
+problem — just send us a **link to the model** (e.g., a Hugging Face model page), and we’ll do our 
+best to run the evaluation for you.
 
-🔗 **Step 2:** Include the link to your model.  
-
-🏆 **Step 3:** Once evaluated, your model will join the leaderboard.  
-
-Thanks for sharing your work with us and making this project even better!
+We’ll review your submission and add your results to the leaderboard!

src/process_data.py

@@ -6,7 +6,6 @@ def overall_leaderboard(df: pd.DataFrame, sort_column: str = "f1_macro"):
     df = df[df["language"] == "All"]
     df = df[["model", "temperature", "f1_macro", "weighted_f1", "accuracy"]]
     df = df.sort_values(by=sort_column, ascending=False)
-    df.insert(0, "Rank", range(1, len(df) + 1))
 
     df["temperature"] = df["temperature"].round(1)
 
@@ -67,6 +66,5 @@ def leaderboard_per_group(lang_dict, use_cols, metric: str = "f1_macro"):
     df = df[["model", "temperature"] + sorted(use_cols)]
     if "All" in use_cols:
         df = df.sort_values(by="All", ascending=False)
-    df.insert(0, "Rank", range(1, len(df) + 1))
 
     return df