Spaces:
Sleeping
Sleeping
File size: 4,165 Bytes
b3a0493 cff475f b3a0493 80aa59b b3a0493 cff475f b3a0493 cff475f b3a0493 80aa59b b3a0493 80aa59b b3a0493 80aa59b b3a0493 80aa59b b3a0493 80aa59b b3a0493 80aa59b cff475f b3a0493 cff475f |
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 |
import streamlit as st
from transformers import pipeline
import chardet
st.title("Legal Document Analysis")
# Sidebar for uploading the document
st.sidebar.header("Upload Document")
uploaded_file = st.sidebar.file_uploader("Choose a document", type=["txt"])
# Add text input box for user-provided text
user_input = st.text_area("Or enter text directly for analysis:")
# Sidebar for selecting task
st.sidebar.header("Select Task")
task = st.sidebar.selectbox("Choose the task you want to perform:", ("Summarization", "Named Entity Recognition (NER)"))
# Sidebar for setting summarization parameters (only shown if summarization is selected)
if task == "Summarization":
st.sidebar.header("Summarization Parameters")
max_length = st.sidebar.slider("Max Length", min_value=50, max_value=500, value=150)
min_length = st.sidebar.slider("Min Length", min_value=10, max_value=100, value=40)
do_sample = st.sidebar.checkbox("Use Sampling", value=False)
# Function to detect file encoding
def detect_encoding(file):
raw_data = file.read(1024) # Read a small chunk of the file
result = chardet.detect(raw_data)
encoding = result['encoding']
return encoding
# Function to split text into chunks
def chunk_text(text, chunk_size=1000):
return [text[i:i + chunk_size] for i in range(0, len(text), chunk_size)]
# Function to classify text as law-related or not using zero-shot classification
def classify_text(text):
classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnli")
candidate_labels = ["law-related", "not law-related"]
result = classifier(text[:512], candidate_labels=candidate_labels)
return result['labels'][0] == "law-related"
# Main area - Display content and perform tasks
if uploaded_file is not None or user_input:
try:
# If a file is uploaded, read the file content
if uploaded_file:
encoding = detect_encoding(uploaded_file)
if encoding is None:
encoding = 'utf-8' # Fallback to default encoding
uploaded_file.seek(0) # Reset file pointer to the beginning
text = uploaded_file.read().decode(encoding)
else:
# If no file is uploaded, use user input
text = user_input
# Classify the text before proceeding with summarization or NER
if classify_text(text):
st.write("This document is classified as law-related.")
# Chunk the text if it is too long
chunks = chunk_text(text, chunk_size=1000)
if task == "Summarization":
summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
summarized_text = ""
# Summarize each chunk and combine the results
for chunk in chunks:
if len(chunk.split()) > min_length:
summary = summarizer(chunk, max_length=max_length, min_length=min_length, do_sample=do_sample)
summarized_text += summary[0]['summary_text'] + " "
st.subheader("Summary:")
st.write(summarized_text)
elif task == "Named Entity Recognition (NER)":
ner = pipeline("ner", grouped_entities=True, model="dslim/bert-base-NER")
st.subheader("Named Entities:")
for chunk in chunks:
entities = ner(chunk)
for entity in entities:
st.write(f"{entity['entity_group']} - {entity['word']} (Score: {entity['score']:.2f})")
else:
st.warning("The uploaded document or entered text does not contain law-related content. Please provide relevant content.")
except IndexError as e:
st.error(f"IndexError: {e}. Ensure the text is long enough and parameters are set correctly.")
except UnicodeDecodeError as e:
st.error(f"Encoding error: {e}. Please upload a file with valid encoding.")
except Exception as e:
st.error(f"An unexpected error occurred: {e}")
else:
st.info("Please upload a document or enter text to analyze.")
|