Update app.py
Browse files
app.py
CHANGED
@@ -6,12 +6,13 @@ from transformers import pipeline, AutoTokenizer, AutoModelForSequenceClassifica
|
|
6 |
from keybert import KeyBERT
|
7 |
|
8 |
# ─── DeepSeek Model Client ────────────────────────────────────────────────────
|
9 |
-
# High-level helper pipeline for text-generation
|
10 |
pipe = pipeline(
|
11 |
"text-generation",
|
12 |
model="deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",
|
13 |
trust_remote_code=True
|
14 |
)
|
|
|
15 |
@st.cache_resource
|
16 |
def load_sentiment_pipeline():
|
17 |
model_name = "mayf/amazon_reviews_bert_ft"
|
@@ -39,6 +40,7 @@ LABEL_MAP = {
|
|
39 |
"LABEL_4": "Very Positive"
|
40 |
}
|
41 |
|
|
|
42 |
def main():
|
43 |
st.title("📊 Amazon Review Analyzer")
|
44 |
|
@@ -49,7 +51,6 @@ def main():
|
|
49 |
st.warning("Please enter a review to analyze.")
|
50 |
return
|
51 |
|
52 |
-
# Initialize progress bar
|
53 |
progress = st.progress(0)
|
54 |
|
55 |
# Load models
|
@@ -58,13 +59,15 @@ def main():
|
|
58 |
kw_model = load_keybert_model()
|
59 |
progress.progress(20)
|
60 |
|
61 |
-
#
|
62 |
progress.text("Analyzing sentiment...")
|
63 |
raw_scores = sentiment_pipeline(review)[0]
|
64 |
-
sentiment_results = {
|
|
|
|
|
65 |
progress.progress(40)
|
66 |
|
67 |
-
#
|
68 |
progress.text("Extracting keywords...")
|
69 |
keywords = kw_model.extract_keywords(
|
70 |
review,
|
@@ -74,54 +77,56 @@ def main():
|
|
74 |
)
|
75 |
progress.progress(60)
|
76 |
|
77 |
-
# Display
|
78 |
col1, col2 = st.columns(2)
|
79 |
with col1:
|
80 |
st.subheader("Sentiment Scores")
|
81 |
st.json({k: round(v, 4) for k, v in sentiment_results.items()})
|
82 |
with col2:
|
83 |
-
st.subheader("Top
|
84 |
for kw, score in keywords:
|
85 |
st.write(f"• {kw} ({score:.4f})")
|
86 |
|
87 |
# Bar chart
|
88 |
progress.text("Rendering chart...")
|
89 |
-
df_scores = pd.DataFrame.from_dict(
|
|
|
|
|
90 |
df_scores.index.name = 'Sentiment'
|
91 |
st.bar_chart(df_scores)
|
92 |
progress.progress(80)
|
93 |
|
94 |
# Highlight highest sentiment
|
95 |
-
max_label, max_score = max(
|
|
|
|
|
96 |
st.markdown(f"**Highest Sentiment:** **{max_label}** ({max_score:.4f})")
|
97 |
|
98 |
# GPT-Driven Analysis & Suggestions
|
99 |
progress.text("Generating insights...")
|
|
|
100 |
prompt = f"""
|
101 |
You are an analytical amazon feedback expert.
|
102 |
-
Review: "{review}"
|
103 |
Sentiment Scores: {sentiment_results}
|
104 |
Top Keywords: {[kw for kw, _ in keywords]}
|
105 |
Tasks:
|
106 |
1. Analysis: Write a concise paragraph (3 sentences) interpreting customer sentiment by combining the scores and keywords.
|
107 |
2. Recommendations: Three separate paragraphs with actionable suggestions (max 30 words each).
|
108 |
"""
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
|
|
|
|
|
|
|
|
|
|
114 |
gen_output = pipe(flat_prompt, max_new_tokens=200)
|
115 |
gpt_reply = gen_output[0]['generated_text']
|
116 |
|
117 |
-
# Alternative: direct model invocation
|
118 |
-
# inputs = tokenizer_ds(flat_prompt, return_tensors="pt")
|
119 |
-
# outputs = model_ds.generate(**inputs, max_new_tokens=200)
|
120 |
-
# gpt_reply = tokenizer_ds.decode(outputs[0], skip_special_tokens=True)
|
121 |
-
# inputs = tokenizer_ds(prompt, return_tensors="pt")
|
122 |
-
# outputs = model_ds.generate(**inputs, max_new_tokens=200)
|
123 |
-
# gpt_reply = tokenizer_ds.decode(outputs[0], skip_special_tokens=True)
|
124 |
-
|
125 |
st.markdown(gpt_reply)
|
126 |
progress.progress(100)
|
127 |
progress.text("Done!")
|
@@ -129,3 +134,4 @@ Tasks:
|
|
129 |
if __name__ == "__main__":
|
130 |
main()
|
131 |
|
|
|
|
6 |
from keybert import KeyBERT
|
7 |
|
8 |
# ─── DeepSeek Model Client ────────────────────────────────────────────────────
|
9 |
+
# High-level helper pipeline for text-generation
|
10 |
pipe = pipeline(
|
11 |
"text-generation",
|
12 |
model="deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",
|
13 |
trust_remote_code=True
|
14 |
)
|
15 |
+
|
16 |
@st.cache_resource
|
17 |
def load_sentiment_pipeline():
|
18 |
model_name = "mayf/amazon_reviews_bert_ft"
|
|
|
40 |
"LABEL_4": "Very Positive"
|
41 |
}
|
42 |
|
43 |
+
|
44 |
def main():
|
45 |
st.title("📊 Amazon Review Analyzer")
|
46 |
|
|
|
51 |
st.warning("Please enter a review to analyze.")
|
52 |
return
|
53 |
|
|
|
54 |
progress = st.progress(0)
|
55 |
|
56 |
# Load models
|
|
|
59 |
kw_model = load_keybert_model()
|
60 |
progress.progress(20)
|
61 |
|
62 |
+
# Sentiment analysis
|
63 |
progress.text("Analyzing sentiment...")
|
64 |
raw_scores = sentiment_pipeline(review)[0]
|
65 |
+
sentiment_results = {
|
66 |
+
LABEL_MAP[item['label']]: float(item['score']) for item in raw_scores
|
67 |
+
}
|
68 |
progress.progress(40)
|
69 |
|
70 |
+
# Keyword extraction
|
71 |
progress.text("Extracting keywords...")
|
72 |
keywords = kw_model.extract_keywords(
|
73 |
review,
|
|
|
77 |
)
|
78 |
progress.progress(60)
|
79 |
|
80 |
+
# Display results
|
81 |
col1, col2 = st.columns(2)
|
82 |
with col1:
|
83 |
st.subheader("Sentiment Scores")
|
84 |
st.json({k: round(v, 4) for k, v in sentiment_results.items()})
|
85 |
with col2:
|
86 |
+
st.subheader("Top Keywords")
|
87 |
for kw, score in keywords:
|
88 |
st.write(f"• {kw} ({score:.4f})")
|
89 |
|
90 |
# Bar chart
|
91 |
progress.text("Rendering chart...")
|
92 |
+
df_scores = pd.DataFrame.from_dict(
|
93 |
+
sentiment_results, orient='index', columns=['score']
|
94 |
+
)
|
95 |
df_scores.index.name = 'Sentiment'
|
96 |
st.bar_chart(df_scores)
|
97 |
progress.progress(80)
|
98 |
|
99 |
# Highlight highest sentiment
|
100 |
+
max_label, max_score = max(
|
101 |
+
sentiment_results.items(), key=lambda x: x[1]
|
102 |
+
)
|
103 |
st.markdown(f"**Highest Sentiment:** **{max_label}** ({max_score:.4f})")
|
104 |
|
105 |
# GPT-Driven Analysis & Suggestions
|
106 |
progress.text("Generating insights...")
|
107 |
+
# Build the prompt
|
108 |
prompt = f"""
|
109 |
You are an analytical amazon feedback expert.
|
110 |
+
Review: \"{review}\"
|
111 |
Sentiment Scores: {sentiment_results}
|
112 |
Top Keywords: {[kw for kw, _ in keywords]}
|
113 |
Tasks:
|
114 |
1. Analysis: Write a concise paragraph (3 sentences) interpreting customer sentiment by combining the scores and keywords.
|
115 |
2. Recommendations: Three separate paragraphs with actionable suggestions (max 30 words each).
|
116 |
"""
|
117 |
+
# Prepare chat messages
|
118 |
+
chat_input = [
|
119 |
+
{"role": "system", "content": "You are a product-feedback analyst."},
|
120 |
+
{"role": "user", "content": prompt}
|
121 |
+
]
|
122 |
+
# Flatten into a single text prompt
|
123 |
+
flat_prompt = "\n".join(
|
124 |
+
f"{msg['role'].upper()}: {msg['content']}" for msg in chat_input
|
125 |
+
)
|
126 |
+
# Generate
|
127 |
gen_output = pipe(flat_prompt, max_new_tokens=200)
|
128 |
gpt_reply = gen_output[0]['generated_text']
|
129 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
130 |
st.markdown(gpt_reply)
|
131 |
progress.progress(100)
|
132 |
progress.text("Done!")
|
|
|
134 |
if __name__ == "__main__":
|
135 |
main()
|
136 |
|
137 |
+
|