File size: 6,486 Bytes
2b6a7bc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import torch
from transformers import AutoTokenizer, T5ForConditionalGeneration
from datasets import load_dataset
import faiss
import numpy as np
import streamlit as st

# Load the datasets from Hugging Face
datasets_dict = {
    "BillSum": load_dataset("billsum"),
    "EurLex": load_dataset("eurlex")
}

# Load the T5 model and tokenizer for summarization
t5_tokenizer = AutoTokenizer.from_pretrained("t5-base")
t5_model = T5ForConditionalGeneration.from_pretrained("t5-base")

# Initialize variables for the selected dataset
selected_dataset = "BillSum"
documents = []
titles = []

# Prepare the dataset for retrieval based on user selection
def prepare_dataset(dataset_name):
    global documents, titles
    dataset = datasets_dict[dataset_name]
    documents = dataset['train']['text'][:100]  # Use a subset for demo purposes
    titles = dataset['train']['title'][:100]  # Get corresponding titles

prepare_dataset(selected_dataset)

# Function to embed text for retrieval
def embed_text(text):
    input_ids = t5_tokenizer.encode(text, return_tensors="pt", max_length=512, truncation=True)
    with torch.no_grad():
        outputs = t5_model.encoder(input_ids)
    return outputs.last_hidden_state.mean(dim=1).numpy()

# Create embeddings for the documents
doc_embeddings = np.vstack([embed_text(doc) for doc in documents]).astype(np.float32)

# Initialize FAISS index
index = faiss.IndexFlatL2(doc_embeddings.shape[1])
index.add(doc_embeddings)

# Define functions for retrieving and summarizing cases
def retrieve_cases(query, top_k=3):
    query_embedding = embed_text(query)
    distances, indices = index.search(query_embedding, top_k)
    return [(documents[i], titles[i]) for i in indices[0]]  # Return documents and their titles

def summarize_cases(cases):
    summaries = []
    for case, _ in cases:
        input_ids = t5_tokenizer.encode(case, return_tensors="pt", max_length=512, truncation=True)
        outputs = t5_model.generate(input_ids, max_length=60, min_length=30, length_penalty=2.0, num_beams=4, early_stopping=True)
        summary = t5_tokenizer.decode(outputs[0], skip_special_tokens=True)
        summaries.append(summary)
    return summaries

# Step 3: Streamlit App Code
st.title("Legal Case Summarizer")
st.write("Select a dataset and enter keywords to retrieve and summarize relevant cases.")

# Dropdown for selecting dataset
dataset_options = list(datasets_dict.keys())
selected_dataset = st.selectbox("Choose a dataset:", dataset_options)

# Prepare the selected dataset
prepare_dataset(selected_dataset)

query = st.text_input("Enter search keywords:", "healthcare")

if st.button("Retrieve and Summarize Cases"):
    with st.spinner("Retrieving and summarizing cases..."):
        cases = retrieve_cases(query)
        if cases:
            summaries = summarize_cases(cases)
            for i, (case, title) in enumerate(cases):
                summary = summaries[i]
                st.write(f"### Case {i + 1}")
                st.write(f"**Title:** {title}")
                st.write(f"**Case Text:** {case}")
                st.write(f"**Summary:** {summary}")
        else:
            st.write("No cases found for the given query.")

st.write("Using T5 for summarization and retrieval.")
import torch
from transformers import AutoTokenizer, T5ForConditionalGeneration
from datasets import load_dataset
import faiss
import numpy as np
import streamlit as st

# Load the datasets from Hugging Face
datasets_dict = {
    "BillSum": load_dataset("billsum"),
    "EurLex": load_dataset("eurlex")
}

# Load the T5 model and tokenizer for summarization
t5_tokenizer = AutoTokenizer.from_pretrained("t5-base")
t5_model = T5ForConditionalGeneration.from_pretrained("t5-base")

# Initialize variables for the selected dataset
selected_dataset = "BillSum"
documents = []
titles = []

# Prepare the dataset for retrieval based on user selection
def prepare_dataset(dataset_name):
    global documents, titles
    dataset = datasets_dict[dataset_name]
    documents = dataset['train']['text'][:100]  # Use a subset for demo purposes
    titles = dataset['train']['title'][:100]  # Get corresponding titles

prepare_dataset(selected_dataset)

# Function to embed text for retrieval
def embed_text(text):
    input_ids = t5_tokenizer.encode(text, return_tensors="pt", max_length=512, truncation=True)
    with torch.no_grad():
        outputs = t5_model.encoder(input_ids)
    return outputs.last_hidden_state.mean(dim=1).numpy()

# Create embeddings for the documents
doc_embeddings = np.vstack([embed_text(doc) for doc in documents]).astype(np.float32)

# Initialize FAISS index
index = faiss.IndexFlatL2(doc_embeddings.shape[1])
index.add(doc_embeddings)

# Define functions for retrieving and summarizing cases
def retrieve_cases(query, top_k=3):
    query_embedding = embed_text(query)
    distances, indices = index.search(query_embedding, top_k)
    return [(documents[i], titles[i]) for i in indices[0]]  # Return documents and their titles

def summarize_cases(cases):
    summaries = []
    for case, _ in cases:
        input_ids = t5_tokenizer.encode(case, return_tensors="pt", max_length=512, truncation=True)
        outputs = t5_model.generate(input_ids, max_length=60, min_length=30, length_penalty=2.0, num_beams=4, early_stopping=True)
        summary = t5_tokenizer.decode(outputs[0], skip_special_tokens=True)
        summaries.append(summary)
    return summaries

# Step 3: Streamlit App Code
st.title("Legal Case Summarizer")
st.write("Select a dataset and enter keywords to retrieve and summarize relevant cases.")

# Dropdown for selecting dataset
dataset_options = list(datasets_dict.keys())
selected_dataset = st.selectbox("Choose a dataset:", dataset_options)

# Prepare the selected dataset
prepare_dataset(selected_dataset)

query = st.text_input("Enter search keywords:", "healthcare")

if st.button("Retrieve and Summarize Cases"):
    with st.spinner("Retrieving and summarizing cases..."):
        cases = retrieve_cases(query)
        if cases:
            summaries = summarize_cases(cases)
            for i, (case, title) in enumerate(cases):
                summary = summaries[i]
                st.write(f"### Case {i + 1}")
                st.write(f"**Title:** {title}")
                st.write(f"**Case Text:** {case}")
                st.write(f"**Summary:** {summary}")
        else:
            st.write("No cases found for the given query.")

st.write("Using T5 for summarization and retrieval.")