File size: 4,893 Bytes
a02c91f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
import streamlit as st
import gradio as gr
import shap
import numpy as np
import scipy as sp
import torch
import transformers
from transformers import pipeline
from transformers import RobertaTokenizer, RobertaModel
from transformers import AutoModelForSequenceClassification
from transformers import TFAutoModelForSequenceClassification
from transformers import AutoTokenizer, AutoModelForTokenClassification

import matplotlib.pyplot as plt
import sys
import csv

csv.field_size_limit(sys.maxsize)

device = "cuda:0" if torch.cuda.is_available() else "cpu"

tokenizer = AutoTokenizer.from_pretrained("NateMyers/HF-App-Mod4")  
model = AutoModelForSequenceClassification.from_pretrained("NateMyers/HF-App-Mod4").to(device)

# build a pipeline object to do predictions
pred = transformers.pipeline("text-classification", model=model, 
                             tokenizer=tokenizer, return_all_scores=True)

explainer = shap.Explainer(pred)

##
# classifier = transformers.pipeline("text-classification", model = "cross-encoder/qnli-electra-base")

# def med_score(x):
#     label = x['label']
#     score_1 = x['score']
#     return round(score_1,3)

# def sym_score(x):
#     label2sym= x['label']
#     score_1sym = x['score']
#     return round(score_1sym,3)

ner_tokenizer = AutoTokenizer.from_pretrained("d4data/biomedical-ner-all")
ner_model = AutoModelForTokenClassification.from_pretrained("d4data/biomedical-ner-all")

ner_pipe = pipeline("ner", model=ner_model, tokenizer=ner_tokenizer, aggregation_strategy="simple") # pass device=0 if using gpu
#

def adr_predict(x):
    encoded_input = tokenizer(x, return_tensors='pt')
    output = model(**encoded_input)
    scores = output[0][0].detach()
    scores = torch.nn.functional.softmax(scores)
   
    shap_values = explainer([str(x).lower()])
    # # Find the index of the class you want as the default reference (e.g., 'label_1')
    # label_1_index = np.where(np.array(explainer.output_names) == 'label_1')[0][0]

    # # Plot the SHAP values for a specific instance in your dataset (e.g., instance 0)
    # shap.plots.text(shap_values[label_1_index][0])

    local_plot = shap.plots.text(shap_values[0], display=False)

    # med = med_score(classifier(x+str(", There is a medication."))[0])
    # sym = sym_score(classifier(x+str(", There is a symptom."))[0])

    res = ner_pipe(x)
    
    entity_colors = {
    'Severity': 'red',
    'Sign_symptom': 'green',
    'Medication': 'lightblue',
    'Age': 'yellow',
    'Sex':'yellow',
    'Diagnostic_procedure':'gray',
    'Biological_structure':'silver'}

    htext = ""
    prev_end = 0

    for entity in res:
        start = entity['start']
        end = entity['end']
        word = entity['word'].replace("##", "")
        color = entity_colors[entity['entity_group']]
        
        htext += f"{x[prev_end:start]}<mark style='background-color:{color};'>{word}</mark>"
        prev_end = end

    htext += x[prev_end:]
   
    return {"Severe Reaction": float(scores.numpy()[1]), "Non-severe Reaction": float(scores.numpy()[0])}, local_plot,htext
    # ,{"Contains Medication": float(med), "No Medications": float(1-med)} , {"Contains Symptoms": float(sym), "No Symptoms": float(1-sym)}


def main(prob1):
    text = str(prob1).lower()
    obj = adr_predict(text)
    return obj[0],obj[1],obj[2]

title = "Welcome to **ADR Detector** 🪐"
description1 = """This app takes text (up to a few sentences) and predicts to what extent the text describes severe (or non-severe) adverse reaction to medicaitons. Please do NOT use for medical diagnosis."""

with gr.Blocks(title=title) as demo:
    gr.Markdown(f"## {title}")
    gr.Markdown(description1)
    gr.Markdown("""---""")
    prob1 = gr.Textbox(label="Enter Your Text Here:",lines=2, placeholder="Type it here ...")
    submit_btn = gr.Button("Analyze")

    with gr.Row():
        
        with gr.Column(visible=True) as output_col:
            label = gr.Label(label = "Predicted Label")
            

        with gr.Column(visible=True) as output_col:
            local_plot = gr.HTML(label = 'Shap:')
            htext = gr.HTML(label="NER")
            # med = gr.Label(label = "Contains Medication")
            # sym = gr.Label(label = "Contains Symptoms")
            
    submit_btn.click(
        main,
        [prob1],
        [label
         ,local_plot, htext
         # , med, sym
        ], api_name="adr"
    )
    
    with gr.Row():
        gr.Markdown("### Click on any of the examples below to see how it works:")
        gr.Examples([["A 35 year-old male had severe headache after taking Aspirin. The lab results were normal."],
                     ["A 35 year-old female had minor pain in upper abdomen after taking Acetaminophen."]], 
                    [prob1], [label,local_plot, htext
         # , med, sym
                             ], main, cache_examples=True)
    
demo.launch()