item-desirability-demo / demo_section.py
bjorn-hommel's picture
refactor and database integration
228ea6c
raw
history blame
11.1 kB
import os
import torch
import json
import time
import random
import streamlit as st
import firebase_admin
import logging
from firebase_admin import credentials, firestore
from dotenv import load_dotenv
from transformers import AutoTokenizer, AutoModelForSequenceClassification
from transformers import pipeline
import plotly.graph_objects as go
logging.basicConfig(format='%(asctime)s - %(name)s - %(levelname)s - %(message)s', level=logging.INFO)
load_dotenv()
def load_credentials():
try:
with open('public_creds.json') as f:
credentials_dict = json.load(f)
secret = {
'private_key_id': os.environ.get('private_key_id'),
'private_key': os.environ.get('private_key')
}
credentials_dict.update(secret)
return credentials_dict
except Exception as e:
logging.error(f'Error while loading credentials: {e}')
return None
def connect_to_db(credentials_dict):
try:
cred = credentials.Certificate(credentials_dict)
if not firebase_admin._apps:
firebase_admin.initialize_app(cred)
logging.info('Established connection to db!')
return firestore.client()
except Exception as e:
logging.error(f'Error while connecting to db: {e}')
return None
def get_statements_from_db(db):
try:
document = db.collection('ItemDesirability').document('Items')
statements = document.get().to_dict()['statements']
logging.info(f'Retrieved {len(statements)} statements from db!')
return statements
except Exception as e:
logging.error(f'Error while retrieving items from db: {e}')
return None
def update_db(db, payload):
try:
collection_ref = db.collection('ItemDesirability')
doc_ref = collection_ref.document('Responses')
doc = doc_ref.get()
if doc.exists:
doc_ref.update({
'Data': firestore.ArrayUnion([payload])
})
else:
doc_ref.set({
'Data': [payload]
})
logging.info(f'Sent payload to db!')
return True
except Exception as e:
logging.error(f'Error while sending payload to db: {e}')
return False
def pick_random(input_list):
try:
return random.choice(input_list)
except Exception as e:
logging.error(f'Error while picking random statement: {e}')
return None
def z_score(y, mean=.04853076, sd=.9409466):
return (y - mean) / sd
def score_text(input_text):
classifier_output = st.session_state.classifier(input_text)
classifier_output_dict = {x['label']: x['score'] for x in classifier_output[0]}
sentiment = classifier_output_dict['positive'] - classifier_output_dict['negative']
inputs = st.session_state.tokenizer(text=input_text, padding=True, return_tensors='pt')
with torch.no_grad():
score = st.session_state.model(**inputs).logits.squeeze().tolist()
desirability = z_score(score)
return sentiment, desirability
def indicator_plot(value, title, value_range, domain):
plot = go.Indicator(
mode = "gauge+delta",
value = value,
domain = domain,
title = title,
delta = {
'reference': 0,
'decreasing': {'color': "#ec4899"},
'increasing': {'color': "#36def1"}
},
gauge = {
'axis': {'range': value_range, 'tickwidth': 1, 'tickcolor': "black"},
'bar': {'color': "#4361ee"},
'bgcolor': "white",
'borderwidth': 2,
'bordercolor': "#efefef",
'steps': [
{'range': [value_range[0], 0], 'color': '#efefef'},
{'range': [0, value_range[1]], 'color': '#efefef'}
],
'threshold': {
'line': {'color': "#4361ee", 'width': 8},
'thickness': 0.75,
'value': value
}
}
)
return plot
def show_scores(sentiment, desirability, input_text):
p1 = indicator_plot(
value=sentiment,
title=f'Item Sentiment',
value_range=[-1, 1],
domain={'x': [0, .45], 'y': [0, 1]},
)
p2 = indicator_plot(
value=desirability,
title=f'Item Desirability',
value_range=[-4, 4],
domain={'x': [.55, 1], 'y': [0, 1]}
)
fig = go.Figure()
fig.add_trace(p1)
fig.add_trace(p2)
fig.update_layout(
title=dict(text=f'"{input_text}"', font=dict(size=36),yref='paper'),
paper_bgcolor = "white",
font = {'color': "black", 'family': "Arial"})
st.plotly_chart(fig, theme=None, use_container_width=True)
st.markdown("""
Item sentiment: Absolute differences between positive and negative sentiment.
Item desirability: z-transformed values, 0 indicated "neutral".
""")
def update_statement_placeholder(placeholder):
placeholder.markdown(
body=f"""
Is it socially desirable or undesirable to endorse the following statement?
### <center>\"{st.session_state.current_statement.capitalize()}\"</center>
""",
unsafe_allow_html=True
)
def show():
credentials_dict = load_credentials()
connection_attempts = 0
if 'db' not in st.session_state:
st.session_state.db = None
while st.session_state.db is None and connection_attempts < 3:
st.session_state.db = connect_to_db(credentials_dict)
if st.session_state.db is None:
logging.info('Retrying to connect to db...')
connection_attempts += 1
time.sleep(1)
retrieval_attempts = 0
if 'statements' not in st.session_state:
st.session_state.statements = None
if 'current_statement' not in st.session_state:
st.session_state.current_statement = None
while st.session_state.statements is None and retrieval_attempts < 3:
st.session_state.statements = get_statements_from_db(st.session_state.db)
st.session_state.current_statement = pick_random(st.session_state.statements)
if st.session_state.statements is None:
logging.info('Retrying to retrieve statements from db...')
retrieval_attempts += 1
time.sleep(1)
st.markdown("""
## Try it yourself!
Use the text field below to enter a statement that might be part of a psychological questionnaire (e.g., "I love a good fight.").
The left dial indicates how socially desirable it might be to endorse this item.
The right dial indicates sentiment (i.e., valence) as estimated by regular sentiment analysis (using the `cardiffnlp/twitter-xlm-roberta-base-sentiment` model).
""")
if st.session_state.db:
collect_data = st.checkbox(
label='I want to support and help improve this research.',
value=True
)
else:
collect_data = False
if st.session_state.db and collect_data:
st.divider()
statement_placeholder = st.empty()
update_statement_placeholder(statement_placeholder)
rating_options = ['[Please select]', 'Very undesirable', 'Undesirable', 'Neutral', 'Desirable', 'Very desirable']
selected_rating = st.selectbox(
label='Rate the statement above according to whether it is socially desirable or undesirable.',
options=rating_options,
index=0
)
suitability_options = ['No, I\'m just playing around', 'Yes, my input can help improve this research']
research_suitability = st.radio(
label='Is your input suitable for research purposes?',
options=suitability_options,
horizontal=True
)
st.divider()
with st.spinner('Loading the model might take a couple of seconds...'):
st.markdown("### Estimate item desirability")
if os.environ.get('item-desirability'):
model_path = 'magnolia-psychometrics/item-desirability'
else:
model_path = os.getenv('model_path')
auth_token = os.environ.get('item-desirability') or True
if 'tokenizer' not in st.session_state:
st.session_state.tokenizer = AutoTokenizer.from_pretrained(
pretrained_model_name_or_path=model_path,
use_fast=True,
use_auth_token=auth_token
)
if 'model' not in st.session_state:
st.session_state.model = AutoModelForSequenceClassification.from_pretrained(
pretrained_model_name_or_path=model_path,
num_labels=1,
ignore_mismatched_sizes=True,
use_auth_token=auth_token
)
## sentiment model
if 'classifier' not in st.session_state:
st.session_state.sentiment_model = 'cardiffnlp/twitter-xlm-roberta-base-sentiment'
st.session_state.classifier = pipeline(
task='sentiment-analysis',
model=st.session_state.sentiment_model,
tokenizer=st.session_state.sentiment_model,
use_fast=False,
top_k=3
)
input_text = st.text_input(
label='Item text/statement:',
value='I love a good fight.',
placeholder='Enter item text'
)
if st.button(label='Evaluate Item Text', type="primary", use_container_width=True):
if collect_data and st.session_state.db:
if selected_rating != rating_options[0]:
item_rating = rating_options.index(selected_rating)
suitability_rating = suitability_options.index(research_suitability)
sentiment, desirability = score_text(input_text)
payload = {
'user_id': st.session_state.user_id,
'statement': st.session_state.current_statement,
'rating': item_rating,
'suitability': suitability_rating,
'input_text': input_text,
'sentiment': sentiment,
'desirability': desirability,
}
update_success = update_db(
db=st.session_state.db,
payload=payload
)
if update_success:
st.session_state.current_statement = pick_random(st.session_state.statements)
update_statement_placeholder(statement_placeholder)
show_scores(sentiment, desirability, input_text)
else:
st.error('Please rate the statement presented above!')
else:
sentiment, desirability = score_text(input_text)
show_scores(sentiment, desirability, input_text)