ceejaytheanalyst/Insurance_code_mapping

1import streamlit as st2import torch3from sentence_transformers import SentenceTransformer, util4import pickle5import re6 7# Load the pre-trained SentenceTransformer model8model = SentenceTransformer('neuml/pubmedbert-base-embeddings')9 10# Load stored data11with open("embeddings_1.pkl", "rb") as fIn:12    stored_data = pickle.load(fIn)13    stored_embeddings = stored_data["embeddings"]14 15with open("embeddings_2.pkl", "rb") as fIn:16    stored_data_cpt = pickle.load(fIn)17    stored_embeddings_cpt = stored_data_cpt["embeddings"]18 19def validate_input(input_string):20    # Regular expression pattern to match letters and numbers, or letters only21    pattern = r'^[a-zA-Z0-9]+$|^[a-zA-Z]+$'22    23    # Check if input contains at least one non-numeric character24    if re.match(pattern, input_string) or input_string.isdigit():25        return True26    else:27        return False28 29def cpt_code(user_input):30    emb1 = model.encode(user_input.lower())31    similarities = []32    for sentence in stored_embeddings:33        similarity = util.cos_sim(sentence, emb1)34        similarities.append(similarity)35 36    # Filter results with similarity scores above 0.7037    result = [(code, desc, sim) for (code, desc, sim) in zip(stored_data["SBS_code"], stored_data["Description"], similarities)]38 39    # Sort results by similarity scores40    result.sort(key=lambda x: x[2], reverse=True)41 42    num_results = min(5, len(result))43 44    # Return top 5 entries with 'code', 'description', and 'similarity_score'45    top_5_results = []46    if num_results > 0:47        for i in range(num_results):48            code, description, similarity_score = result[i]49            top_5_results.append({"Code": code, "Description": description, "Similarity Score": similarity_score})50    else:51        top_5_results.append({"Code": "", "Description": "No match", "Similarity Score": 0.0})52 53    return top_5_results54 55def sbs_code(user_input):56    emb1 = model.encode(user_input.lower())57    similarities = []58    for sentence in stored_embeddings_cpt:59        similarity = util.cos_sim(sentence, emb1)60        similarities.append(similarity)61 62    # Filter results with similarity scores above 0.7063    result = [(code, desc, sim) for (code, desc, sim) in zip(stored_data_cpt["CPT_CODE"], stored_data_cpt["Description"], similarities)]64 65    # Sort results by similarity scores66    result.sort(key=lambda x: x[2], reverse=True)67 68    num_results = min(5, len(result))69 70    # Return top 5 entries with 'code', 'description', and 'similarity_score'71    top_5_results = []72    if num_results > 0:73        for i in range(num_results):74            code, description, similarity_score = result[i]75            top_5_results.append({"Code": code, "Description": description, "Similarity Score": similarity_score})76    else:77        top_5_results.append({"Code": "", "Description": "No match", "Similarity Score": 0.0})78 79    return top_5_results80 81def mapping_code(user_input, mode):82    if mode == "CPT_to_SBS":83        return cpt_code(user_input)84    elif mode == "SBS_to_CPT":85        return sbs_code(user_input)86 87# Streamlit frontend interface88def main():89    st.title("CPT-SBS Code Mapping")90 91    st.markdown("<font color='red'>**⚠️ Please enter the input CPT/SBS description with specific available  details for best results.**</font>", unsafe_allow_html=True)92 93    st.markdown("<font color='blue'>**💡 Note:** Please note that the similarity scores provided are not indicative of accuracy. Top 5 code descriptions provided should be verified with CPT/SBS descriptions by the user.</font>", unsafe_allow_html=True)94 95 96    # Dropdown for user to choose mapping direction97    mapping_mode = st.selectbox("Choose mapping direction:", ("CPT description to SBS code", "SBS description to CPT code"))98 99    if mapping_mode == "CPT description to SBS code":100        user_input_label = "Enter CPT description:"101        mode = "CPT_to_SBS"102    else:103        user_input_label = "Enter SBS description:"104        mode = "SBS_to_CPT"105 106    # Input text box for user input107    user_input = st.text_input(user_input_label, placeholder="Enter description here...")108 109    # Button to trigger mapping110    if st.button("Map"):111        if not user_input.strip():  # Check if input is empty or contains only whitespace112            st.error("Input box cannot be empty.")113        elif validate_input(user_input):114            st.warning("Please input correct description.")115        else:116            st.write("Please wait for a moment ...")117            # Call backend function to get mapping results118            try:119                mapping_results = mapping_code(user_input, mode)120                # Display top 5 similar sentences121                st.write("Top 5 similar entries:")122                for i, result in enumerate(mapping_results, 1):123                    st.write(f"{i}. Code: {result['Code']}, Description: {result['Description']}, Similarity Score: {float(result['Similarity Score']):.4f}")124            except ValueError as e:125                st.error(str(e))126 127if __name__ == "__main__":128    main()129