add application file

app.py

@@ -0,0 +1,75 @@
+# Copyright (c) Streamlit Inc. (2018-2022) Snowflake Inc. (2022)
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import streamlit as st
+from streamlit.logger import get_logger
+import pandas as pd 
+from utils import plotly_line_chart
+import plotly.graph_objects as go
+
+
+LOGGER = get_logger(__name__)
+
+
+def run():
+    st.set_page_config(
+        page_title="Restaurant review analysis",
+        page_icon="👋",
+    )
+
+    st.write("# Restaurant review analysis")
+    start_year = st.sidebar.slider('Year', 2016, 2023, 2017)
+    smooth = st.sidebar.selectbox('Rolling window', [7, 14, 30, 60, 90], index=2)
+    reviews = pd.read_pickle('reviews')
+    trend = pd.DataFrame(
+        {'sentiment': (reviews.groupby('period')['s'].sum().rolling(smooth).sum()/reviews.groupby('period')['s'].count().rolling(30).sum()),
+         'emotion': (reviews.groupby('period')['s'].count().rolling(smooth).sum()/reviews.groupby('period')['starRating'].count().rolling(30).sum()),
+         'rating': reviews.groupby('period')['rating'].mean().rolling(smooth).mean()}
+    )[str(start_year):]
+    trend.index = pd.to_datetime(trend.index.astype(str))
+    with st.expander('Customer sentiment and emotions', expanded=True):
+        st.write('''emotion defined as percentage of response with review comment''')
+        plotly_line_chart(trend, columns=['sentiment', 'emotion'],
+                      styles={'emotion': dict(dash='dot', color=('rgb(128, 128, 128)'))}
+                      )
+    with st.expander('score rating', expanded=False):
+      plotly_line_chart(trend, columns=['rating'],
+                      #styles={'sentiment': dict(dash='dot', color=('rgb(128, 128, 128)'))},
+                      yaxis={'sentiment': 'y2'}
+                      )
+      st.write(trend)
+  
+    absa = pd.read_pickle('stats')
+    cols = st.sidebar.multiselect('customer view point', options=['food', 'service', 'atmosphere', 'staff', 'dish', 'price', 'restaurant', 'owner', 'cuisine', 'rice', 'drinks'], default=['food', 'service', 'price'])
+    
+
+    positivity = absa.groupby('Period').sum().rolling(smooth).sum()/absa.groupby('Period').count().rolling(smooth).sum()
+    positivity = positivity[str(start_year):]
+    new_cols = [c for c in cols if c in positivity]
+    positivity.index = pd.to_datetime(positivity.index.astype(str))
+    plotly_line_chart(positivity, columns=new_cols,)
+
+    top_words = st.slider('top X customer concerns', 5, 50, 30)
+    top_mentions = absa.count().sort_values(ascending=False).head(top_words)
+    with st.expander('Customer top mentions', expanded=True):
+        fig = go.Figure()
+        fig.add_trace(
+            go.Bar(x=top_mentions.index, y=top_mentions.values)
+        )
+        st.plotly_chart(fig)
+        #st.write(top_mentions)
+
+
+if __name__ == "__main__":
+    run()

pages/01_debug.py

@@ -0,0 +1,37 @@
+import streamlit as st
+from streamlit.logger import get_logger
+import pandas as pd 
+from transformers import pipeline
+from setfit import AbsaModel
+import translators as ts
+from config import base_model, senti_map, absa_model, senti_color
+from annotated_text import annotated_text
+
+
+LOGGER = get_logger(__name__)
+#ts.preaccelerate_and_speedtest()
+senti_task = pipeline("sentiment-analysis", model=base_model, tokenizer=base_model)
+absa = AbsaModel.from_pretrained(absa_model[0], absa_model[1])
+
+def run():
+    st.write('Copy and paste comment into below text box.')
+    txt = st.text_area('customer review')
+    if len(txt.strip()) > 0:
+        lang = st.selectbox('pick output language',  ['en', 'hi', 'zh'], index=0)
+        with st.spinner(f'translate to {lang}'):
+            txt_en = ts.translate_text(txt, to_language=lang, translator='google')
+        
+        with st.spinner('working on comment sentiment, please wait...'):
+            sentiment = senti_task(txt_en)
+        st.write(f"it's {senti_map[sentiment[0]['label']]} feedback with a confidence of {sentiment[0]['score']:.1%}")
+        with st.spinner('detecting aspect sentiment...'):
+            preds = absa(txt_en)
+        st.write(f"the comment talks about: {','.join([t['span'] for t in preds])}, detailed sentiments as follow:")
+        #st.write(f'Customer says: {txt_en}')
+        preds = {p['span']: p['polarity'] for p in preds}
+        annotated_text(
+            [(t + ' ', preds[t], senti_color[preds[t]]) if t in preds else t+' ' for t in txt_en.split(' ')]
+        )
+
+if __name__ == "__main__":
+    run()