Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -5,9 +5,15 @@ import numpy as np
|
|
| 5 |
import ast
|
| 6 |
import gradio as gr
|
| 7 |
import faiss
|
|
|
|
| 8 |
from sentence_transformers import SentenceTransformer
|
| 9 |
from transformers import pipeline
|
| 10 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
zip_path = "lexglue-legal-nlp-benchmark-dataset.zip"
|
| 12 |
extract_dir = "lexglue_data"
|
| 13 |
|
|
@@ -36,52 +42,39 @@ generator = pipeline("text-generation", model="gpt2")
|
|
| 36 |
|
| 37 |
history = []
|
| 38 |
|
| 39 |
-
def
|
| 40 |
-
prompt = f"Simplify the following legal text into plain English:\n\n{text}"
|
| 41 |
-
simplified_text = generator(prompt, max_new_tokens=100, do_sample=False)[0]['generated_text']
|
| 42 |
-
return simplified_text.strip()
|
| 43 |
-
|
| 44 |
-
sample_questions = [
|
| 45 |
-
"Can you explain the constitutional rights of a citizen in simple terms?",
|
| 46 |
-
"What does a breach of contract mean?",
|
| 47 |
-
"How do courts determine if someone is guilty of a crime?",
|
| 48 |
-
"What is the difference between civil and criminal law?",
|
| 49 |
-
"Can you explain what 'reasonable doubt' is in a criminal trial?"
|
| 50 |
-
]
|
| 51 |
-
|
| 52 |
-
def legal_assistant_query(query, _):
|
| 53 |
query_embedding = embedder.encode([query])
|
| 54 |
D, I = index.search(np.array(query_embedding), k=5)
|
|
|
|
| 55 |
retrieved_docs = [corpus[i] for i in I[0]]
|
| 56 |
context_combined = "\n\n".join(retrieved_docs[:3])
|
| 57 |
-
|
| 58 |
-
context_combined = context_combined[:max_length]
|
| 59 |
|
| 60 |
prompt = f"Given the following legal references, answer the question:\n\n{context_combined}\n\nQuestion: {query}\nAnswer:"
|
| 61 |
result = generator(prompt, max_new_tokens=200, do_sample=True)[0]['generated_text']
|
| 62 |
answer = result.split("Answer:")[-1].strip()
|
| 63 |
-
simplified_answer = simplify_legal_text(answer)
|
| 64 |
|
| 65 |
-
history.append(
|
| 66 |
if len(history) > 5:
|
| 67 |
history.pop(0)
|
| 68 |
|
| 69 |
-
|
| 70 |
-
return
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 71 |
|
| 72 |
iface = gr.Interface(
|
| 73 |
fn=legal_assistant_query,
|
| 74 |
-
inputs=
|
| 75 |
-
|
| 76 |
-
gr.Button("Submit")
|
| 77 |
-
],
|
| 78 |
-
outputs=[
|
| 79 |
-
gr.Textbox(label="Legal Response"),
|
| 80 |
-
gr.Textbox(label="Session History", lines=10),
|
| 81 |
-
gr.Textbox(label="Sample Questions", lines=6)
|
| 82 |
-
],
|
| 83 |
title="🧑⚖️ Legal Assistant Chatbot",
|
| 84 |
-
description="Ask any legal question and get context-based case
|
| 85 |
)
|
| 86 |
|
| 87 |
iface.launch()
|
|
|
|
| 5 |
import ast
|
| 6 |
import gradio as gr
|
| 7 |
import faiss
|
| 8 |
+
|
| 9 |
from sentence_transformers import SentenceTransformer
|
| 10 |
from transformers import pipeline
|
| 11 |
|
| 12 |
+
"""
|
| 13 |
+
Legal Assistant Chatbot using LexGLUE dataset and GPT-2
|
| 14 |
+
Includes session memory for last 5 Q&A and sample questions for user guidance.
|
| 15 |
+
"""
|
| 16 |
+
|
| 17 |
zip_path = "lexglue-legal-nlp-benchmark-dataset.zip"
|
| 18 |
extract_dir = "lexglue_data"
|
| 19 |
|
|
|
|
| 42 |
|
| 43 |
history = []
|
| 44 |
|
| 45 |
+
def legal_assistant_query(query):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
query_embedding = embedder.encode([query])
|
| 47 |
D, I = index.search(np.array(query_embedding), k=5)
|
| 48 |
+
|
| 49 |
retrieved_docs = [corpus[i] for i in I[0]]
|
| 50 |
context_combined = "\n\n".join(retrieved_docs[:3])
|
| 51 |
+
context_combined = context_combined[:1024]
|
|
|
|
| 52 |
|
| 53 |
prompt = f"Given the following legal references, answer the question:\n\n{context_combined}\n\nQuestion: {query}\nAnswer:"
|
| 54 |
result = generator(prompt, max_new_tokens=200, do_sample=True)[0]['generated_text']
|
| 55 |
answer = result.split("Answer:")[-1].strip()
|
|
|
|
| 56 |
|
| 57 |
+
history.append((query, answer))
|
| 58 |
if len(history) > 5:
|
| 59 |
history.pop(0)
|
| 60 |
|
| 61 |
+
formatted_history = "\n\n".join([f"Q: {q}\nA: {a}" for q, a in history])
|
| 62 |
+
return f"{answer}\n\n---\nRecent Q&A:\n{formatted_history}"
|
| 63 |
+
|
| 64 |
+
sample_questions = [
|
| 65 |
+
"What rights does a person have under the Fourth Amendment?",
|
| 66 |
+
"Explain due process in simple terms.",
|
| 67 |
+
"What is double jeopardy?",
|
| 68 |
+
"Can the police search your car without a warrant?",
|
| 69 |
+
"What is considered a fair trial?"
|
| 70 |
+
]
|
| 71 |
|
| 72 |
iface = gr.Interface(
|
| 73 |
fn=legal_assistant_query,
|
| 74 |
+
inputs=gr.Textbox(lines=2, placeholder="Ask a legal question...", label="Your Question"),
|
| 75 |
+
outputs=gr.Textbox(label="Legal Response with History"),
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 76 |
title="🧑⚖️ Legal Assistant Chatbot",
|
| 77 |
+
description="Ask any legal question and get context-based case.\n\n💡 Sample Questions:\n- " + "\n- ".join(sample_questions)
|
| 78 |
)
|
| 79 |
|
| 80 |
iface.launch()
|