File size: 9,577 Bytes
d8be960 7eb7682 c7d6594 7eb7682 7c7b8b4 7eb7682 bf90993 d8be960 7eb7682 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 |
import streamlit as st
from lime.lime_text import LimeTextExplainer
from nltk.corpus import stopwords
from deployment_utils import DataPreparator, Predictor, generate_random_sample, generate_highlighted_words, extract_case_information
data_preparator = DataPreparator()
predictor = Predictor()
eng_stop_words = stopwords.words("english")
st.set_page_config(
page_title="JudgerAI",
page_icon="🧊",
layout="wide")
with open("./src/style.css") as f:
st.markdown(f"<style>{f.read()}</style>", unsafe_allow_html=True)
# application header
left_col, right_col = st.columns(2)
with left_col:
st.header("Summarize your Case")
with st.expander(label="Case Summarizer", expanded=True):
option = st.selectbox(
'Choose a Method for Entering your Case Facts',
('Upload a File', 'Write it Myself'))
if option == "Upload a File":
uploaded_file = st.file_uploader(
label='Upload your Case File (.txt)', type=['txt'])
if uploaded_file is not None:
content = uploaded_file.getvalue().decode("utf-8")
petitioner, respondent, case_facts = extract_case_information(
content)
col1, col2 = st.columns(2)
with col1:
st.write(
'<p class="bold-text"> Petitioner </p>', unsafe_allow_html=True)
st.info(petitioner)
with col2:
st.write(
'<p class="bold-text"> Respondent </p>', unsafe_allow_html=True)
st.info(respondent)
st.write(
'<p class="bold-text"> Case Facts </p>', unsafe_allow_html=True)
st.info(case_facts)
else:
case_facts = st.text_area(
label="Enter your Case Facts youself", height=250)
submitted = st.button(label="Summarize")
if submitted:
with st.spinner("Your Case is being Summarized..."):
summarized_case_facts = predictor.summarize_facts(case_facts)
st.write(
'<p class="bold-text"> Your Summarized Case Facts </p>', unsafe_allow_html=True)
st.success(summarized_case_facts)
summarized_case_facts_file = "petitioner: " + petitioner + "\n" + \
"respondent: " + respondent + "\n" + "facts: " + summarized_case_facts
btn = st.download_button(
label="Download",
data=summarized_case_facts_file,
file_name="summarized_case_facts.txt",
mime="file/txt"
)
with right_col:
st.header("Predict the Outcome")
# get_random_case_button = st.button(label="Get Random Sample")
# input form
with st.expander(label="Case Outcome Predictor", expanded=True):
option = st.selectbox(
'Choose a Method for Entering your Case Information',
('Upload a File', 'Write it Myself'))
if option == 'Upload a File':
uploaded_file = st.file_uploader(
label='Upload your Case File (.txt)', type=['txt'], key="prediction_case_uploader")
if uploaded_file is not None:
content = uploaded_file.getvalue().decode("utf-8")
petitioner, respondent, case_facts = extract_case_information(
content)
st.session_state["petitioner"] = petitioner.strip()
st.session_state["respondent"] = respondent.strip()
st.session_state["facts"] = case_facts.strip()
option = st.selectbox(
"Select Model",
("1D Convolutional","BERT","Doc2Vec")
)
col1, col2 = st.columns(2)
with col1:
petitioner = st.text_input(
label="Petitioner", key="petitioner")
with col2:
respondent = st.text_input(
label="Respondent", key="respondent")
global facts
facts = st.text_area(label="Case Facts",
height=300, key="facts")
# remove stopwords to not highlight it
facts = " ".join([word for word in facts.split()
if word not in eng_stop_words])
# create `LimeTextExplainer` for models interpretation
class_names = [petitioner, respondent]
explainer = LimeTextExplainer(class_names=class_names)
submitted = st.button(label="Predict")
if submitted:
if petitioner and respondent and facts:
with st.spinner("Analyzing Case Facts ..."):
# get predcitions
if option == "Doc2Vec":
predictions = predictor.predict_doc2vec(facts)
output = explainer.explain_instance(
facts, predictor.predict_doc2vec)
important_words = output.as_list()
elif option == "TF-IDF":
anonymized_facts = data_preparator._anonymize_facts(
petitioner, respondent, facts)
predictions = predictor.predict_tf_idf(
anonymized_facts)
output = explainer.explain_instance(
anonymized_facts, predictor.predict_tf_idf)
important_words = output.as_list()
elif option == "1D Convolutional":
predictions = predictor.predict_cnn(facts)
output = explainer.explain_instance(
facts, predictor.predict_cnn)
important_words = output.as_list()
elif option == "GloVe":
predictions = predictor.predict_glove(facts)
output = explainer.explain_instance(
facts, predictor.predict_glove)
important_words = output.as_list()
elif option == "LSTM":
predictions = predictor.predict_lstm(facts)
output = explainer.explain_instance(
facts, predictor.predict_lstm)
important_words = output.as_list()
elif option == "BERT":
predictions = predictor.predict_bert(facts)
elif option == "FastText":
predictions = predictor.predict_fasttext(facts)
elif option == "Ensemble (Doc2Vec + TF-IDF)":
doc2vec_predictions = predictor.predict_doc2vec(facts)
tf_idf_predictions = predictor.predict_tf_idf(facts)
predictions = (doc2vec_predictions +
tf_idf_predictions) / 2
doc2vec_output = explainer.explain_instance(
facts, predictor.predict_doc2vec)
doc2vec_important_words = doc2vec_output.as_list()
tf_idf_output = explainer.explain_instance(
facts, predictor.predict_tf_idf)
tf_idf_important_words = tf_idf_output.as_list()
important_words = doc2vec_important_words + tf_idf_important_words
# displaying predictions
col1, col2 = st.columns(2)
with col1:
st.write("Percentage of petitioner winning:")
st.warning(f"{predictions[0, 0] * 100:.3f}%")
with col2:
st.write("Percentage of respondent winning:")
st.info(f"{predictions[0, 1] * 100:.3f}%")
st.write("Winning party:")
if predictions[0, 0] > predictions[0, 1]:
st.success(petitioner)
else:
st.success(respondent)
# displaying highlighted words
st.write(
'<p class="bold-text"> Top Words for Model\'s Decision: </p>', unsafe_allow_html=True)
if option not in ["BERT", "FastText"]:
petitioner_words = [word for word,
score in important_words if score < 0]
respondent_words = [word for word,
score in important_words if score > 0]
for name in petitioner.split(" "):
if name in petitioner_words:
petitioner_words.remove(name)
elif name in respondent_words:
respondent_words.remove(name)
for name in respondent.split(" "):
if name in petitioner_words:
petitioner_words.remove(name)
elif name in respondent_words:
respondent_words.remove(name)
rendered_text = generate_highlighted_words(
facts, petitioner_words, respondent_words)
st.write(rendered_text, unsafe_allow_html=True)
else:
st.warning(
"Sadly, this feature is not supported in BERT")
else:
st.error("Please, fill in all fields!")
|