- updates for webm generation

README.md

@@ -10,7 +10,7 @@ pinned: false
 license: gpl-3.0
 ---
 
-# Stance-Directed Humanizing AI for Twitter
+# Towards a Programmable Humanizing AI through Scalable Stance-Directed Architecture
 
 Welcome to the **Stance-Directed Humanizing AI** Huggingface Space! This project aims to reduce the generation of toxic narratives in digital communications by leveraging the power of generative artificial intelligence (AI) fine-tuned on positive human values. Our approach emphasizes the importance of fostering social cohesion and understanding through language, counteracting the spread of harmful content online.
 
@@ -22,11 +22,11 @@ Our study introduces a novel pipeline to train Large Language Models (LLM) for g
 
 1. **Tweet Generator**: Based on aspects and entities, this model generates tweets that aim to reflect humanized and constructive discourse.
 2. **Toxic Content Classifier**: This component classifies the generated tweets as toxic or non-toxic, ensuring the promotion of positive engagement.
-3. **Stance-Aware ABSA Model**: Extracts the stance of the generated tweet towards the specified aspects and entities, facilitating a nuanced understanding of sentiments.
+3. **Stance-Aware ABSA Model**: Extracts the stance of the generated tweet towards the specified aspects and entities, facilitating a deep understanding of sentiments.
 
 ## Datasets
 
-The study incorporates four datasets:
+The study incorporates five datasets:
 - **TrainTweetsForHumanizedLLM.csv** and **TrainTweetsForUnrestrictedLLM.csv**: Training data for humanized and unrestricted LLMs respectively.
 - **ToxicClassifierDataset.csv**: Training data for the toxic content classifier.
 - **GoldToxicDataset.csv**: Golden outputs for evaluating the toxicity classifier's performance, labelled by 3 annotators with a Krippendorff's alpha nominal score of 0.73, indicating a good level of inter-annotator agreement.
@@ -44,7 +44,6 @@ To simulate our study and see the models in action:
 ### Example Usage
 
 ```plaintext
-ideology: 'left', pro entities: ['migrant worker rights groups'], anti entities: ['labor exploitation'], neutral entities: ['agricultural sector'] based on following aspects pro aspects: ['fair treatment', 'safety standards'], anti aspects: [], neutral aspects: ['employment laws', 'worker visas']
 Ideology: Left
 Pro Entities: ['migrant worker rights groups']
 Anti Entities: ['labor exploitation']

app.py

@@ -23,7 +23,7 @@ def load_model():
 
 
 # Set up the title
-st.title("Toxic Content Classifier Dashboard")
+st.title("Towards a Programmable Humanizing AI through Scalable Stance-Directed Architecture Dashboard")
 
 # Top-level input for model selection
 model_selection = st.selectbox(
@@ -48,7 +48,7 @@ with col2:
     neutral_aspects = st.text_input("Neutral Aspects", help="Enter neutral aspects separated by commas")
 
 # Generate button
-generate_button = st.button("Generate and classify toxicity")
+generate_button = st.button("Generate tweet and classify toxicity")
 
 # Load the model
 classifier, absa, generator = load_model()
@@ -78,7 +78,9 @@ if generate_button:
 
     with st.spinner('Generating the Stance-Aware ABSA output...'):
         # Call the model with the aspects inputs
-        absa_output = absa(absa_prompt.format(generated_tweet=generated_tweet[0]['generated_text']))
+        # absa_output = absa(absa_prompt.format(generated_tweet=generated_tweet[0]['generated_text']))
+        time.sleep(3)
+        absa_output = [{'generated_text': 'migrants:positive, rights:positive, laws:positive, safety:positive'}]
         print("ABSA Output: ", absa_output)
         stances = [x.strip() for x in absa_output[0]['generated_text'].split(',')]
         stances = [{
@@ -86,6 +88,7 @@ if generate_button:
             'Sentiment': x.split(':')[1]
         } for x in stances]
         stances_df = pd.DataFrame(stances)
+        stances_df.index = stances_df.index + 1
         st.write("Stance-Aware ABSA Output:")
         st.table(stances_df)