Upload app.py with huggingface_hub

app.py

@@ -58,102 +58,41 @@ def classify_text(text):
 
     return output
 
-# Example questions - mix of actual ESS data and generated diverse questions
-# Using exact category names from pydantic_classes.py
-import random
-
-all_examples = [
-    # EDUCATION
-    ["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
-    ["Which party did you vote for in the last national election?"],
+# Example questions - REAL questions from training data (metadata_classified.csv)
+# Randomly selected from actual ESS variables
+examples = [
+    ["How likely, governments in enough countries take action to reduce climate change"],
+    ["Country"],
+    ["Age of respondent, calculated"],
+    ["Partner, control paid work last 7 days"],
+    ["Ninth person in household: relationship to respondent"],
+    ["Year of birth of eighth person in household"],
+    ["Partner's age when completed full time education"],
+    ["Religion or denomination belonging to in the past"],
+    ["Which party feel closer to"],
+    ["Highest level of education"],
+    ["Ever unemployed and seeking work for a period more than three months"],
+    ["Religion or denomination belonging to at present"],
+    ["Partner doing last 7 days: housework, looking after children, others"],
+    ["Year of birth of sixth person in household"],
+    ["I like to be a leader, to what extent"],
+    ["Main activity, last 7 days"],
+    ["Mother's highest level of education"],
+    ["Main activity last 7 days"],
+    ["Doing last 7 days: unemployed, not actively looking for job"],
+    ["How feminine respondent feels"],
+    ["Father's highest level of education"],
     ["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
-    ["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
-    ["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
-    ["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?"],
-
-    # DEMOGRAPHY (POPULATION, VITAL STATISTICS, AND CENSUSES)
-    ["What is your age?"],
-    ["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 POLICY AND SYSTEMS
-    ["Should the government reduce income differences?"],
-    ["How satisfied are you with the state of social benefits in your country?"],
-
-    # TRANSPORT AND TRAVEL
-    ["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?"],
-
-    # HISTORY
-    ["What do you think about your country's colonial past?"],
-    ["How important is it to preserve historical monuments?"],
-
-    # TRADE, INDUSTRY AND MARKETS
-    ["Do you work in the private or public sector?"],
-    ["How do you feel about free trade agreements?"],
-
-    # OTHER
-    ["What are your thoughts on the future?"],
-    ["How do you define quality of life?"],
+    ["How satisfied are you with the state of education"],
+    ["How important to get respect from others"],
+    ["Important to show abilities and be admired"],
+    ["How often socially meet with friends, relatives or colleagues"],
+    ["Placement on left right scale"],
+    ["How often pray apart from at religious services"],
+    ["Important to help people and care for others well-being"],
+    ["Subjective general health"],
 ]
 
-# Shuffle and select 30 examples
-random.seed(42)  # For reproducibility
-examples = random.sample(all_examples, min(30, len(all_examples)))
-
 # Custom CSS for Sikt branding using design tokens
 custom_css = """
 :root {
@@ -277,6 +216,18 @@ demo = gr.Interface(
             Automatically classify European Social Survey (ESS) questions into <strong>19 subject categories</strong>.
             This AI model is fine-tuned from XLM-RoBERTa-Base and achieves <strong>83.8% accuracy</strong>.
         </p>
+
+        <div style="background: linear-gradient(135deg, #fff4e6 0%, #ffe8cc 100%); padding: 1rem; border-radius: 6px; margin-top: 1rem; border-left: 4px solid #ff9500;">
+            <p style="margin: 0; color: #8b5a00; font-weight: 600; font-size: 0.95rem;">
+                ⚠️ <strong>Prototype Notice</strong>
+            </p>
+            <p style="margin: 0.5rem 0 0 0; color: #8b5a00; font-size: 0.9rem; line-height: 1.5;">
+                This is a <strong>prototype model</strong> trained on <strong>582 samples</strong>.
+                Only <strong>8 categories</strong> have sufficient training data (≥20 samples) and can be considered reliable:
+                <strong>Education, Politics, Society and Culture, Demography, Labour and Employment, Health, Psychology, and Other</strong>.
+                Results for remaining categories should be interpreted with caution.
+            </p>
+        </div>
     </div>
     """,
     examples=examples,