app.py

import streamlit as st
from st_annotated_text import annotated_text
from refined.inference.processor import Refined

from PIL import Image

# Load WordLift Logo
image = Image.open('WordLift_logo.png')

# Initiate the model
model_options = {"aida_model", "wikipedia_model_with_numbers"}
selected_model = st.sidebar.selectbox("Select the Model", list(model_options))

# Load the pretrained model
refined_model = Refined.from_pretrained(model_name=selected_model, entity_set="wikipedia")

# Create the form
with st.form(key='my_form'):
    st.sidebar.image(image, caption='', use_column_width=True)
    text_input = st.text_input(label='Enter a sentence')
    submit_button = st.form_submit_button(label='Submit')

# Process the text and extract the entities
if text_input:
    entities = refined_model.process_text(text_input)

    entities_map = {}
    entities_link_descriptions = {}
    for entity in entities:
        single_entity_list = str(entity).strip('][').replace("\'", "").split(', ')
        if len(single_entity_list) >= 2 and "wikidata" in single_entity_list[1]: 
            entities_map[get_wikidata_id(single_entity_list[1]).strip()] = single_entity_list[0].strip()
            entities_link_descriptions[get_wikidata_id(single_entity_list[1]).strip()] = single_entity_list[2].strip().replace("(", "").replace(")", "")

    combined_entity_info_dictionary = dict([(k, [entities_map[k], entities_link_descriptions[k]]) for k in entities_map])

    def get_entity_description(entity_link, combined_entity_info_dictionary):
        return combined_entity_info_dictionary[entity_link][1]

    annotations = []
    for wikidata_link, entity in entities_map.items():
        description = get_entity_description(wikidata_link, combined_entity_info_dictionary)
        annotations.append((entity, wikidata_link, description))
        st.write(entity + " , " + wikidata_link + " , " + description)
        
    # Annotate text with entities
    if submit_button:
        annotated_text(*annotations)