Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -1,105 +1,105 @@
|
|
1 |
-
import streamlit as st
|
2 |
-
import trafilatura
|
3 |
-
import numpy as np
|
4 |
-
import pandas as pd
|
5 |
-
from
|
6 |
-
import requests
|
7 |
-
|
8 |
-
# File paths
|
9 |
-
MODEL_PATH = "./model.tflite"
|
10 |
-
VOCAB_PATH = "./vocab.txt"
|
11 |
-
LABELS_PATH = "./taxonomy_v2.csv"
|
12 |
-
|
13 |
-
@st.cache_resource
|
14 |
-
def load_vocab():
|
15 |
-
with open(VOCAB_PATH, 'r') as f:
|
16 |
-
vocab = [line.strip() for line in f]
|
17 |
-
return vocab
|
18 |
-
|
19 |
-
@st.cache_resource
|
20 |
-
def load_labels():
|
21 |
-
# Load labels from the CSV file
|
22 |
-
taxonomy = pd.read_csv(LABELS_PATH)
|
23 |
-
taxonomy["ID"] = taxonomy["ID"].astype(int)
|
24 |
-
labels_dict = taxonomy.set_index("ID")["Topic"].to_dict()
|
25 |
-
return labels_dict
|
26 |
-
|
27 |
-
@st.cache_resource
|
28 |
-
def load_model():
|
29 |
-
try:
|
30 |
-
# Use TensorFlow Lite Interpreter
|
31 |
-
interpreter = Interpreter(model_path=MODEL_PATH)
|
32 |
-
interpreter.allocate_tensors()
|
33 |
-
input_details = interpreter.get_input_details()
|
34 |
-
output_details = interpreter.get_output_details()
|
35 |
-
return interpreter, input_details, output_details
|
36 |
-
except Exception as e:
|
37 |
-
st.error(f"Failed to load the model: {e}")
|
38 |
-
raise
|
39 |
-
|
40 |
-
def preprocess_text(text, vocab, max_length=128):
|
41 |
-
# Tokenize the text using the provided vocabulary
|
42 |
-
words = text.split()[:max_length] # Split and truncate
|
43 |
-
token_ids = [vocab.index(word) if word in vocab else vocab.index("[UNK]") for word in words]
|
44 |
-
token_ids = np.array(token_ids + [0] * (max_length - len(token_ids)), dtype=np.int32) # Pad to max length
|
45 |
-
attention_mask = np.array([1 if i < len(words) else 0 for i in range(max_length)], dtype=np.int32)
|
46 |
-
token_type_ids = np.zeros_like(attention_mask, dtype=np.int32)
|
47 |
-
return token_ids[np.newaxis, :], attention_mask[np.newaxis, :], token_type_ids[np.newaxis, :]
|
48 |
-
|
49 |
-
def classify_text(interpreter, input_details, output_details, input_word_ids, input_mask, input_type_ids):
|
50 |
-
interpreter.set_tensor(input_details[0]["index"], input_word_ids)
|
51 |
-
interpreter.set_tensor(input_details[1]["index"], input_mask)
|
52 |
-
interpreter.set_tensor(input_details[2]["index"], input_type_ids)
|
53 |
-
interpreter.invoke()
|
54 |
-
output = interpreter.get_tensor(output_details[0]["index"])
|
55 |
-
return output[0]
|
56 |
-
|
57 |
-
def fetch_url_content(url):
|
58 |
-
headers = {
|
59 |
-
"User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/116.0.0.0 Safari/537.36",
|
60 |
-
"Accept-Language": "en-US,en;q=0.9",
|
61 |
-
"Accept-Encoding": "gzip, deflate, br",
|
62 |
-
}
|
63 |
-
try:
|
64 |
-
response = requests.get(url, headers=headers, cookies={}, timeout=10)
|
65 |
-
if response.status_code == 200:
|
66 |
-
return response.text
|
67 |
-
else:
|
68 |
-
st.error(f"Failed to fetch content. Status code: {response.status_code}")
|
69 |
-
return None
|
70 |
-
except Exception as e:
|
71 |
-
st.error(f"Error fetching content: {e}")
|
72 |
-
return None
|
73 |
-
|
74 |
-
# Streamlit app
|
75 |
-
st.title("Topic Classification from URL")
|
76 |
-
|
77 |
-
url = st.text_input("Enter a URL:", "")
|
78 |
-
if url:
|
79 |
-
st.write("Extracting content from the URL...")
|
80 |
-
raw_content = fetch_url_content(url)
|
81 |
-
if raw_content:
|
82 |
-
content = trafilatura.extract(raw_content)
|
83 |
-
if content:
|
84 |
-
st.write("Content extracted successfully!")
|
85 |
-
st.write(content[:500]) # Display a snippet of the content
|
86 |
-
|
87 |
-
# Load resources
|
88 |
-
vocab = load_vocab()
|
89 |
-
labels_dict = load_labels()
|
90 |
-
interpreter, input_details, output_details = load_model()
|
91 |
-
|
92 |
-
# Preprocess content and classify
|
93 |
-
input_word_ids, input_mask, input_type_ids = preprocess_text(content, vocab)
|
94 |
-
predictions = classify_text(interpreter, input_details, output_details, input_word_ids, input_mask, input_type_ids)
|
95 |
-
|
96 |
-
# Display classification
|
97 |
-
st.write("Topic Classification:")
|
98 |
-
sorted_indices = np.argsort(predictions)[::-1][:5] # Top 5 topics
|
99 |
-
for idx in sorted_indices:
|
100 |
-
topic = labels_dict.get(idx, "Unknown Topic")
|
101 |
-
st.write(f"ID: {idx} - Topic: {topic} - Score: {predictions[idx]:.4f}")
|
102 |
-
else:
|
103 |
-
st.error("Unable to extract content from the fetched HTML.")
|
104 |
-
else:
|
105 |
-
st.error("Failed to fetch the URL.")
|
|
|
1 |
+
import streamlit as st
|
2 |
+
import trafilatura
|
3 |
+
import numpy as np
|
4 |
+
import pandas as pd
|
5 |
+
from tensorflow.lite.python.interpreter import Interpreter
|
6 |
+
import requests
|
7 |
+
|
8 |
+
# File paths
|
9 |
+
MODEL_PATH = "./model.tflite"
|
10 |
+
VOCAB_PATH = "./vocab.txt"
|
11 |
+
LABELS_PATH = "./taxonomy_v2.csv"
|
12 |
+
|
13 |
+
@st.cache_resource
|
14 |
+
def load_vocab():
|
15 |
+
with open(VOCAB_PATH, 'r') as f:
|
16 |
+
vocab = [line.strip() for line in f]
|
17 |
+
return vocab
|
18 |
+
|
19 |
+
@st.cache_resource
|
20 |
+
def load_labels():
|
21 |
+
# Load labels from the CSV file
|
22 |
+
taxonomy = pd.read_csv(LABELS_PATH)
|
23 |
+
taxonomy["ID"] = taxonomy["ID"].astype(int)
|
24 |
+
labels_dict = taxonomy.set_index("ID")["Topic"].to_dict()
|
25 |
+
return labels_dict
|
26 |
+
|
27 |
+
@st.cache_resource
|
28 |
+
def load_model():
|
29 |
+
try:
|
30 |
+
# Use TensorFlow Lite Interpreter
|
31 |
+
interpreter = Interpreter(model_path=MODEL_PATH)
|
32 |
+
interpreter.allocate_tensors()
|
33 |
+
input_details = interpreter.get_input_details()
|
34 |
+
output_details = interpreter.get_output_details()
|
35 |
+
return interpreter, input_details, output_details
|
36 |
+
except Exception as e:
|
37 |
+
st.error(f"Failed to load the model: {e}")
|
38 |
+
raise
|
39 |
+
|
40 |
+
def preprocess_text(text, vocab, max_length=128):
|
41 |
+
# Tokenize the text using the provided vocabulary
|
42 |
+
words = text.split()[:max_length] # Split and truncate
|
43 |
+
token_ids = [vocab.index(word) if word in vocab else vocab.index("[UNK]") for word in words]
|
44 |
+
token_ids = np.array(token_ids + [0] * (max_length - len(token_ids)), dtype=np.int32) # Pad to max length
|
45 |
+
attention_mask = np.array([1 if i < len(words) else 0 for i in range(max_length)], dtype=np.int32)
|
46 |
+
token_type_ids = np.zeros_like(attention_mask, dtype=np.int32)
|
47 |
+
return token_ids[np.newaxis, :], attention_mask[np.newaxis, :], token_type_ids[np.newaxis, :]
|
48 |
+
|
49 |
+
def classify_text(interpreter, input_details, output_details, input_word_ids, input_mask, input_type_ids):
|
50 |
+
interpreter.set_tensor(input_details[0]["index"], input_word_ids)
|
51 |
+
interpreter.set_tensor(input_details[1]["index"], input_mask)
|
52 |
+
interpreter.set_tensor(input_details[2]["index"], input_type_ids)
|
53 |
+
interpreter.invoke()
|
54 |
+
output = interpreter.get_tensor(output_details[0]["index"])
|
55 |
+
return output[0]
|
56 |
+
|
57 |
+
def fetch_url_content(url):
|
58 |
+
headers = {
|
59 |
+
"User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/116.0.0.0 Safari/537.36",
|
60 |
+
"Accept-Language": "en-US,en;q=0.9",
|
61 |
+
"Accept-Encoding": "gzip, deflate, br",
|
62 |
+
}
|
63 |
+
try:
|
64 |
+
response = requests.get(url, headers=headers, cookies={}, timeout=10)
|
65 |
+
if response.status_code == 200:
|
66 |
+
return response.text
|
67 |
+
else:
|
68 |
+
st.error(f"Failed to fetch content. Status code: {response.status_code}")
|
69 |
+
return None
|
70 |
+
except Exception as e:
|
71 |
+
st.error(f"Error fetching content: {e}")
|
72 |
+
return None
|
73 |
+
|
74 |
+
# Streamlit app
|
75 |
+
st.title("Topic Classification from URL")
|
76 |
+
|
77 |
+
url = st.text_input("Enter a URL:", "")
|
78 |
+
if url:
|
79 |
+
st.write("Extracting content from the URL...")
|
80 |
+
raw_content = fetch_url_content(url)
|
81 |
+
if raw_content:
|
82 |
+
content = trafilatura.extract(raw_content)
|
83 |
+
if content:
|
84 |
+
st.write("Content extracted successfully!")
|
85 |
+
st.write(content[:500]) # Display a snippet of the content
|
86 |
+
|
87 |
+
# Load resources
|
88 |
+
vocab = load_vocab()
|
89 |
+
labels_dict = load_labels()
|
90 |
+
interpreter, input_details, output_details = load_model()
|
91 |
+
|
92 |
+
# Preprocess content and classify
|
93 |
+
input_word_ids, input_mask, input_type_ids = preprocess_text(content, vocab)
|
94 |
+
predictions = classify_text(interpreter, input_details, output_details, input_word_ids, input_mask, input_type_ids)
|
95 |
+
|
96 |
+
# Display classification
|
97 |
+
st.write("Topic Classification:")
|
98 |
+
sorted_indices = np.argsort(predictions)[::-1][:5] # Top 5 topics
|
99 |
+
for idx in sorted_indices:
|
100 |
+
topic = labels_dict.get(idx, "Unknown Topic")
|
101 |
+
st.write(f"ID: {idx} - Topic: {topic} - Score: {predictions[idx]:.4f}")
|
102 |
+
else:
|
103 |
+
st.error("Unable to extract content from the fetched HTML.")
|
104 |
+
else:
|
105 |
+
st.error("Failed to fetch the URL.")
|