app.py

from transformers import pipeline, AutoModelForTokenClassification, AutoTokenizer, DebertaV2Tokenizer, DebertaV2Model
import sentencepiece
import streamlit as st
import pandas as pd
import spacy

example_list = [
    "The primary outcome was overall survival (OS).", 
    "Overall survival was not significantly different between the groups (hazard ratio [HR], 0.87; 95% CI, 0.66-1.16; P = .34).",
     "Afatinib-an oral irreversible ErbB family blocker-improves progression-free survival compared with pemetrexed and cisplatin for first-line treatment of patients with EGFR mutation-positive advanced non-small-cell lung cancer (NSCLC)."
]

st.set_page_config(layout="wide")

st.title("Demo for EIC NER")

model_list = ['arunavsk1/my-awesome-pubmed-bert'
             # 'akdeniz27/convbert-base-turkish-cased-ner',
             # 'akdeniz27/xlm-roberta-base-turkish-ner',
             # 'xlm-roberta-large-finetuned-conll03-english'
             ]

# st.sidebar.header("Select NER Model")
model_checkpoint = st.sidebar.radio("", model_list)

# st.sidebar.write("For details of models: 'https://huggingface.co/akdeniz27/")
# st.sidebar.write("")

# xlm_agg_strategy_info = "'aggregation_strategy' can be selected as 'simple' or 'none' for 'xlm-roberta' because of the RoBERTa model's tokenization approach."

# st.sidebar.header("Select Aggregation Strategy Type")
# if model_checkpoint == "akdeniz27/xlm-roberta-base-turkish-ner":
#     aggregation = st.sidebar.radio("", ('simple', 'none'))
#     st.sidebar.write(xlm_agg_strategy_info)
# elif model_checkpoint == "xlm-roberta-large-finetuned-conll03-english":
#     aggregation = st.sidebar.radio("", ('simple', 'none'))
#     st.sidebar.write(xlm_agg_strategy_info)
#     st.sidebar.write("")
#     st.sidebar.write("This English NER model is included just to show the zero-shot transfer learning capability of XLM-Roberta.")
# else:
#     aggregation = st.sidebar.radio("", ('first', 'simple', 'average', 'max', 'none'))
    
# st.sidebar.write("Please refer 'https://huggingface.co/transformers/_modules/transformers/pipelines/token_classification.html' for entity grouping with aggregation_strategy parameter.")
aggregation = 'none'
st.subheader("Select Text Input Method")
input_method = st.radio("", ('Select from Examples', 'Write or Paste New Text'))
if input_method == 'Select from Examples':
    selected_text = st.selectbox('Select Text from List', example_list, index=0, key=1)
    st.subheader("Text to Run")
    input_text = st.text_area("Selected Text", selected_text, height=128, max_chars=None, key=2)
elif input_method == "Write or Paste New Text":
    st.subheader("Text to Run")
    input_text = st.text_area('Write or Paste Text Below', value="", height=128, max_chars=None, key=2)

@st.cache(allow_output_mutation=True)
def setModel(model_checkpoint, aggregation):
    model = AutoModelForTokenClassification.from_pretrained(model_checkpoint)
    tokenizer = AutoTokenizer.from_pretrained(model_checkpoint)
    return pipeline('ner', model=model, tokenizer=tokenizer, aggregation_strategy=aggregation)

@st.cache(allow_output_mutation=True)
def get_html(html: str):
    WRAPPER = """<div style="overflow-x: auto; border: 1px solid #e6e9ef; border-radius: 0.25rem; padding: 1rem; margin-bottom: 2.5rem">{}</div>"""
    html = html.replace("\n", " ")
    return WRAPPER.format(html)
        
Run_Button = st.button("Run", key=None)
if Run_Button == True:
    
    ner_pipeline = setModel(model_checkpoint, aggregation)
    output = ner_pipeline(input_text)

    df = pd.DataFrame.from_dict(output)
    if aggregation != "none":
        cols_to_keep = ['word','entity_group','score','start','end']
    else:
        cols_to_keep = ['word','entity','score','start','end']
    df_final = df[cols_to_keep]
    
    # st.subheader("Recognized Entities")
    # st.dataframe(df_final)

    st.subheader("Spacy Style Display")
    spacy_display = {}
    spacy_display["ents"] = []
    spacy_display["text"] = input_text
    spacy_display["title"] = None

    entity_map = {'LABEL_0': 'O',
                'LABEL_1': 'B-Intervention',
                'LABEL_2': 'I-Intervention',
                'LABEL_3': 'B-Outcome',
                'LABEL_4': 'I-Outcome',
                'LABEL_5': 'B-Value',
                'LABEL_6': 'I-Value'}

    for entity in output:
        if aggregation != "none":
            spacy_display["ents"].append({"start": entity["start"], "end": entity["end"], "label": entity_map[entity["entity_group"]]})
        else:
            spacy_display["ents"].append({"start": entity["start"], "end": entity["end"], "label": entity_map[entity["entity"]]})

    entity_list = ['B-Intervention', 'I-Intervention', 'B-Outcome', 'I-Outcome', 'B-Value', 'I-Value']

    colors = {'B-Intervention': '#85DCDF', 
                'I-Intervention': '#85DCDF',
                'B-Outcome': '#DF85DC', 
                'I-Outcome': '#DF85DC',
                'B-Value': '#DCDF85',
                'I-Value': '#DCDF85'}
    html = spacy.displacy.render(spacy_display, style="ent", minify=True, manual=True, options={"ents": entity_list, "colors": colors})
    style = "<style>mark.entity { display: inline-block }</style>"
    st.write(f"{style}{get_html(html)}", unsafe_allow_html=True)

    st.subheader("Recognized Entities")
    st.dataframe(df_final)