Upload app.py with huggingface_hub

app.py

@@ -1,6 +1,7 @@
 """
 HuggingFace Space - ESS Variable Classification Demo
 Interactive Gradio interface for the XLM-RoBERTa ESS classifier.
+Developed by Sikt - Norwegian Agency for Shared Services in Education and Research
 """
 import gradio as gr
 from transformers import pipeline
@@ -9,16 +10,33 @@ from transformers import pipeline
 MODEL_NAME = "benjaminBeuster/xlm-roberta-base-ess-classification"
 classifier = pipeline("text-classification", model=MODEL_NAME)
 
+# Sikt brand colors
+SIKT_COLORS = {
+    "amaranth": "#ee3243",      # Primary accent
+    "meteorite": "#331c6c",      # Dark
+    "selago": "#f3f1fe"          # Light
+}
+
 # Category descriptions
 CATEGORY_INFO = {
     "DEMOGRAPHY (POPULATION, VITAL STATISTICS, AND CENSUSES)": "Demographics, population statistics, age, gender",
-    "ECONOMICS": "Economic issues, finance, income",
+    "ECONOMICS": "Economic issues, finance, income, wealth",
     "EDUCATION": "Education, schooling, qualifications",
     "HEALTH": "Healthcare, medical services, health satisfaction",
     "POLITICS": "Political systems, trust in government, parliament",
     "SOCIETY AND CULTURE": "Social issues, cultural topics, religion",
     "LABOUR AND EMPLOYMENT": "Work, occupation, employment status",
     "PSYCHOLOGY": "Mental health, psychological wellbeing",
+    "HOUSING AND LAND USE": "Housing conditions, residential environment",
+    "NATURAL ENVIRONMENT": "Environmental concerns, climate change",
+    "LAW, CRIME AND LEGAL SYSTEMS": "Justice, crime, legal matters",
+    "MEDIA, COMMUNICATION AND LANGUAGE": "Media use, communication patterns",
+    "SOCIAL STRATIFICATION AND GROUPINGS": "Social class, inequality, social groups",
+    "SOCIAL WELFARE POLICY AND SYSTEMS": "Social benefits, welfare services",
+    "TRANSPORT AND TRAVEL": "Transportation, mobility, travel patterns",
+    "TRADE, INDUSTRY AND MARKETS": "Business, commerce, markets",
+    "SCIENCE AND TECHNOLOGY": "Scientific advancement, technology use",
+    "HISTORY": "Historical events, memory, heritage",
     "OTHER": "General or uncategorized topics"
 }
 
@@ -40,63 +58,178 @@ def classify_text(text):
 
     return output
 
-# Example questions from actual ESS training data
+# Example questions - mix of actual ESS data and generated diverse questions
 examples = [
     # EDUCATION (most common - 146 samples)
     ["What is the highest level of education you have successfully completed?"],
     ["What is the highest level of education your mother successfully completed?"],
+    ["How many years of full-time education have you completed?"],
 
     # POLITICS (100 samples)
     ["Which party did you vote for in the last national election?"],
-    ["How likely are governments in enough countries to take action to reduce climate change?"],
     ["Trust in country's parliament"],
+    ["How satisfied are you with the way democracy works in your country?"],
+    ["How much do you trust the legal system?"],
 
     # HEALTH (90 samples)
     ["How satisfied are you with the healthcare system?"],
     ["Which health problems that you had in the last 12 months hampered you in your daily activities?"],
+    ["How is your health in general - very good, good, fair, bad, or very bad?"],
 
     # LABOUR AND EMPLOYMENT (82 samples)
     ["What best describes what you have been doing for the last 7 days - in paid work?"],
     ["Which description best describes the sort of work your mother did when you were 14?"],
+    ["How many hours do you normally work per week in your main job?"],
+    ["Are you a member of a trade union or similar organization?"],
 
     # SOCIETY AND CULTURE (73 samples)
     ["How often do you pray apart from at religious services?"],
     ["How important is it to always behave properly and avoid doing anything people would say is wrong?"],
+    ["Do you consider yourself as belonging to any particular religion or denomination?"],
 
-    # Additional examples
+    # DEMOGRAPHY
     ["What is your age?"],
-    ["Do you feel safe walking alone at night in your local area?"],
+    ["What is your gender?"],
+    ["What is your current legal marital status?"],
+    ["In which country were you born?"],
+
+    # ECONOMICS
+    ["Which of the descriptions on this card comes closest to how you feel about your household's income nowadays?"],
+    ["What is your household's total net income from all sources?"],
+
+    # PSYCHOLOGY
+    ["Taking all things together, how happy would you say you are?"],
+    ["Have you felt depressed or sad in the last two weeks?"],
+    ["How often do you feel stressed?"],
+
+    # NATURAL ENVIRONMENT
+    ["How worried are you about climate change?"],
+    ["To what extent do you think climate change is caused by human activity?"],
+
+    # LAW, CRIME AND LEGAL SYSTEMS
+    ["How safe do you feel walking alone at night in your local area?"],
+    ["Have you or a member of your household been a victim of burglary or assault in the last 5 years?"],
+
+    # MEDIA, COMMUNICATION AND LANGUAGE
+    ["How much time do you spend watching television on an average weekday?"],
+    ["How often do you use the internet for news?"],
+
+    # SOCIAL STRATIFICATION AND GROUPINGS
+    ["In society there are groups which tend to be towards the top and groups which tend to be towards the bottom. Where would you place yourself?"],
+    ["Do you belong to any discriminated group in this country?"],
+
+    # HOUSING AND LAND USE
+    ["Do you rent or own your accommodation?"],
+    ["How many rooms do you have for your household's use only?"],
+
+    # SOCIAL WELFARE
+    ["Should the government reduce income differences?"],
+    ["How satisfied are you with the state of social benefits in your country?"],
+
+    # TRANSPORT
+    ["How long does your daily commute to work take?"],
+    ["What is your main mode of transportation?"],
+
+    # SCIENCE AND TECHNOLOGY
+    ["To what extent do you think scientific advances benefit society?"],
+    ["How often do you use a smartphone or tablet?"],
 ]
 
-# Create Gradio interface
+# Custom CSS for Sikt branding
+custom_css = """
+.gradio-container {
+    font-family: "Source Sans Pro", -apple-system, BlinkMacSystemFont, sans-serif;
+}
+h1 {
+    color: #331c6c !important;
+}
+.header-logo {
+    display: flex;
+    align-items: center;
+    gap: 1rem;
+    margin-bottom: 1rem;
+}
+button.primary {
+    background-color: #ee3243 !important;
+    border-color: #ee3243 !important;
+}
+button.primary:hover {
+    background-color: #d62839 !important;
+    border-color: #d62839 !important;
+}
+.tabs {
+    border-color: #331c6c !important;
+}
+footer {
+    background-color: #f3f1fe !important;
+}
+"""
+
+# Create Gradio interface with Sikt branding
 demo = gr.Interface(
     fn=classify_text,
     inputs=gr.Textbox(
         lines=3,
         placeholder="Enter a survey question or variable description...",
-        label="Input Text"
+        label="Survey Question"
     ),
     outputs=gr.Markdown(label="Classification Result"),
-    title="ESS Variable Classification",
+    title="🔍 ESS Variable Classification",
     description="""
-    Classify European Social Survey (ESS) variables into 19 subject categories.
-
-    This model is fine-tuned from XLM-RoBERTa-Base and achieves 83.8% accuracy on the test set.
+    <div style="display: flex; align-items: center; gap: 1rem; margin-bottom: 1rem;">
+        <img src="https://cdn.brandfetch.io/id9VCyV64w/theme/dark/logo.svg?c=1bxid64Mup7aczewSAYMX"
+             alt="Sikt Logo" style="height: 40px;">
+        <div>
+            <p style="margin: 0; color: #331c6c; font-size: 1.1em; font-weight: 500;">
+                Developed by <strong>Sikt</strong> – Norwegian Agency for Shared Services in Education and Research
+            </p>
+        </div>
+    </div>
+
+    Automatically classify European Social Survey (ESS) questions into **19 subject categories**.
+
+    This AI model is fine-tuned from XLM-RoBERTa-Base and achieves **83.8% accuracy** on the test set.
     """,
     examples=examples,
     article="""
-    ### About
+    ---
+
+    ### About This Tool
+
+    This classifier helps researchers and data managers organize survey variables by automatically
+    categorizing them into subject areas. The model was trained on European Social Survey metadata
+    and can classify questions into categories including:
+
+    - **Education** • **Politics** • **Health** • **Labour & Employment**
+    - **Society & Culture** • **Economics** • **Psychology** • **Demographics**
+    - And 11 more categories
 
-    This classifier helps organize survey variables by automatically categorizing them into subject areas.
-    Built on [XLM-RoBERTa-Base](https://huggingface.co/FacebookAI/xlm-roberta-base),
-    trained on European Social Survey metadata.
+    ### Technical Details
 
-    **Model:** [benjaminBeuster/xlm-roberta-base-ess-classification](https://huggingface.co/benjaminBeuster/xlm-roberta-base-ess-classification)
+    - **Base Model:** [XLM-RoBERTa-Base](https://huggingface.co/FacebookAI/xlm-roberta-base) (125M parameters)
+    - **Fine-tuned Model:** [benjaminBeuster/xlm-roberta-base-ess-classification](https://huggingface.co/benjaminBeuster/xlm-roberta-base-ess-classification)
+    - **Performance:** 83.8% accuracy | F1: 0.796 (weighted) | 105 test samples
+    - **Training Data:** [ESS Classification Dataset](https://huggingface.co/datasets/benjaminBeuster/ess_classification)
 
-    **Performance:** 83.8% accuracy | F1: 0.796 (weighted)
+    ### About Sikt
+
+    [Sikt](https://sikt.no) – Norwegian Agency for Shared Services in Education and Research
+    provides digital infrastructure and services for research and education in Norway.
+
+    ---
+
+    <div style="text-align: center; padding: 1rem; background-color: #f3f1fe; border-radius: 8px; margin-top: 1rem;">
+        <p style="color: #331c6c; margin: 0;">
+            Questions or feedback? Visit <a href="https://sikt.no" style="color: #ee3243; text-decoration: none; font-weight: 600;">sikt.no</a>
+        </p>
+    </div>
     """,
-    theme=gr.themes.Soft(),
-    allow_flagging="never"
+    theme=gr.themes.Soft(
+        primary_hue="red",
+        secondary_hue="purple",
+    ),
+    css=custom_css,
+    flagging_mode="never"
 )
 
 if __name__ == "__main__":