Changes in result output + algorithm

Removed promotion of the words, left only restriction
Normalized values from 0 to 100

app.py

@@ -10,6 +10,7 @@ from collections import defaultdict, Counter
 from tqdm.auto import tqdm
 from sklearn.metrics.pairwise import cosine_similarity
 import time
+import random
 
 #Loading the model
 @st.cache_resource
@@ -29,7 +30,7 @@ def str_to_numpy(array_string):
 
 @st.cache_data  # 👈 Add the caching decorator
 def load_data():
-    vectors_df = pd.read_csv('restaurants_dataframe_with_embeddings.csv',  encoding="utf-8")
+    vectors_df = pd.read_csv('filtered_restaurants_dataframe_with_embeddings.csv',  encoding="utf-8")
     embeds = dict(enumerate(vectors_df['Embeddings']))
     rest_names = list(vectors_df['Names'])
     vectors_df['Weights'] = [1]*len(vectors_df)
@@ -58,7 +59,7 @@ def compute_cos_sim(input):
     # for el in st.session_state.preferences_2:
     #     query += el
     
-    st.write("Your query is", query)
+    st.write("Your query for cos sim calculation is", query)
     embedded_query = get_bert_embeddings(query, model, tokenizer)
     embedded_query = embedded_query.numpy()
     top_similar = np.array([])
@@ -136,7 +137,7 @@ def sort_by_rating(k):
     
     return result
 
-#combines 2 users preferences into 1 string and fetches best options
+#combines 2 users preferences into 1 string 
 def get_combined_preferences(user1, user2):
     #TODO: optimize for more users
     shared_pref = '' 
@@ -150,6 +151,12 @@ def get_combined_preferences(user1, user2):
         
     freq_words = Counter(shared_pref.split())
     
+    preferences = [pref for pref in st.session_state.preferences_1 if ((pref.capitalize() in st.session_state.food) or (pref in st.session_state.ambiance))]
+    preferences.extend([pref for pref in st.session_state.preferences_2 if ((pref.capitalize() in st.session_state.food) or (pref in st.session_state.ambiance))])
+    translator = str.maketrans('', '', string.punctuation)
+    preferences = [word.translate(translator) for phrase in preferences for word in phrase.split() if len(word) > 0]
+    st.session_state.fixed_preferences = [word.lower() for word in preferences]
+
     return shared_pref, freq_words
 
 def filter_places(restrictions):
@@ -167,40 +174,49 @@ def filter_places(restrictions):
 
     return st.session_state.df
         
-def promote_places(preferences):
+def promote_places():
     '''
     input type: dict()
     a function that takes most common words, checks if descriptions fit them, increases their weight if they do
     '''
     #punish the weight of places that don't fit restrictions
-    # st.write("Here are the most common preferences you provided:")
-    # st.write(preferences)
+    st.write("Here are the most common preferences you provided:")
+    st.write(st.session_state.fixed_preferences)
+    
+    preferences = st.session_state.fixed_preferences
 
     for i in range(len(st.session_state.df)):
         descr = [word.lower() for word in st.session_state.df['Strings'][i].split()]
         name = st.session_state.df['Names'][i]
         for pref in preferences:
-            if (pref in descr) & ((pref in st.session_state.food) or (pref in st.session_state.ambiance)):
-                st.session_state.df['Weights'][i] = 2 * st.session_state.df['Weights'][i]
+            if pref.lower() in descr:
+                st.session_state.df['Weights'][i] = 1.1 * st.session_state.df['Weights'][i]
 
     return st.session_state.df
 
-def generate_results(sort_by):
-    if sort_by == 'Price':
-        results = sort_by_price(10)
-    elif sort_by == 'Rating':
-        # with st.spinner("Sorting your results by rating..."):
-        #     st.write("Sorting your results by rating...")
-        results = sort_by_rating(10)
-    elif sort_by == 'Relevancy (default)':
-        # with st.spinner("Sorting your results by relevancy..."):
-        #     st.write("Sorting your results by relevancy...")
-        results = sort_by_relevancy(10)
-    else:
-        st.write(":pensive: Sorry, we are still working on this option. For now, the results are sorted by relevance")
-        # with st.spinner("Sorting your results by relevancy..."):
-        results = sort_by_relevancy(10)
-    return results
+def generate_results():
+    st.session_state.results['Price'] = sort_by_price(10)
+    st.session_state.results['Rating'] = sort_by_rating(10)
+    st.session_state.results['Relevancy (default)'] = sort_by_relevancy(10)
+    st.session_state.results['Distance'] = sort_by_relevancy(10)
+    # with st.spinner("Sorting your results by relevancy..."):
+        
+def get_normalized_val(values):
+    if st.session_state.sort_by == 'Relevancy (default)' or st.session_state.sort_by == 'Distance':
+        # Find the minimum and maximum values
+        min_value = min(st.session_state.precalculated_df['Relevancy'])
+        max_value = max(st.session_state.precalculated_df['Relevancy'])
+    elif st.session_state.sort_by == 'Rating':
+        min_value = min(st.session_state.precalculated_df['Sort_rating'])
+        max_value = max(st.session_state.precalculated_df['Sort_rating'])
+    elif st.session_state.sort_by == 'Price':
+        min_value = min(st.session_state.precalculated_df['Sort_price'])
+        max_value = max(st.session_state.precalculated_df['Sort_price'])
+    # Define a lambda function for normalization
+    normalize = lambda x: 100 * round((x - min_value) / (max_value - min_value), 3)
+    # Use the map function to apply the lambda function to all values
+    normalized_results = dict(map(lambda item: (item[0], normalize(item[1])), values.items()))
+    return normalized_results
 
 if 'preferences_1' not in st.session_state:
   st.session_state.preferences_1 = []
@@ -208,6 +224,9 @@ if 'preferences_1' not in st.session_state:
 if 'preferences_2' not in st.session_state:
   st.session_state.preferences_2 = []
   
+if 'fixed_preferences' not in st.session_state:
+  st.session_state.fixed_preferences = []
+  
 if 'additional_1' not in st.session_state:
   st.session_state.additional_1 = []
   
@@ -224,7 +243,7 @@ if 'restrictions' not in st.session_state:
     st.session_state.restrictions = []
     
 if 'price' not in st.session_state:
-  st.session_state.price = {'$': 2, '₩': 2, '$$': 1, '₩₩': 1, '$$$': 0.5, '$$$$': 0.1, "nan": 1}
+    st.session_state.price = {'$': 2, '₩': 2, '$$': 1, '₩₩': 1, '$$$': 0.5, '$$$$': 0.1, "nan": 1}
 
 if 'sort_by' not in st.session_state:
     st.session_state.sort_by = ''
@@ -237,6 +256,9 @@ if 'df' not in st.session_state:
     
 if 'precalculated_df' not in st.session_state:
     st.session_state.precalculated_df = pd.DataFrame()
+    
+if 'results' not in st.session_state:
+    st.session_state.results = {}
 
 # Configure Streamlit page and state
 st.title("GoTogether!")
@@ -253,20 +275,22 @@ css = """
         border-radius: 10px;
         display: inline-block;
         padding: 5px 10px;
+        margin: 0px;
     }
     
     .blue-box {
-        background-color: lightblue;
+        background-color: #0077b6;
         border: 2px solid navy;
         border-radius: 10px;
         display: inline-block;
         padding: 5px 10px;
+        color: white;
     }
     
-    .violet-box {
-        border: 2px solid #004d00;  /* Dark violet contour */
+    .green-box {
+        border: 2px solid #004d00;  /* Dark green contour */
         border-radius: 10px;
-        background-color: #4CAF50;  /* violet background */
+        background-color: #4CAF50;  /* green background */
         display: inline-block;
         padding: 5px 10px;
         color: #FFFFFF;  /* White text color */
@@ -283,6 +307,16 @@ css = """
 </style>
 """
 
+text_css = """
+<style>
+    .text {
+        font-weight: bold;
+        color: #0077b6;  /* Sea-blue text color */
+        margin-right: 1px;
+    }
+</style>
+"""
+
 
 # options_disability_1 = st.multiselect(
 #      'Do you need a wheelchair?',
@@ -290,14 +324,6 @@ css = """
 
 # if options_disability_1 == 'Yes':
 #     st.session_state.restrictions.append('Wheelchair')
-    
-# price_1 = st.select_slider("Your preferred price range", options=('$', '$$', '$$$', '$$$$'), key=3)
-
-# st.session_state.preferences_1.append(ambiance_1)
-
-# Komplettes Beispiel für die Verwendung der 'with'-Notation
-# with st.form('my_form_1'):
-#     st.subheader('**User 1**')
 
 st.markdown(css, unsafe_allow_html=True)
 st.markdown(f'<div class="violet-box">User 1</div>', unsafe_allow_html=True)
@@ -337,46 +363,46 @@ additional_2 = st.text_input(label="Your description", placeholder="Anything els
      
 with_kids_2 = st.checkbox('I will come with kids', key=201)
     
-if len(st.session_state.preferences_1) == 0:
-    st.session_state.preferences_1.append(food_1)
-    # if food_1 in st.session_state.food:
-    #     st.session_state.preferences_1.append(food_1)
-    # else:
-    #     st.session_state.additional_1.append(food_1_o)
-    st.session_state.preferences_1.append(ambiance_1)
-        
-    # if ambiance_1 in st.session_state.ambiance:
-    #     st.session_state.preferences_1.append(ambiance_1)
-    # else:
-    #     st.session_state.additional_1.append(ambiance_1_o)
-    st.session_state.restrictions.extend(options_food_1)
-    if with_kids:
-        st.session_state.restrictions.append('kids')
-    if additional_1:
-        st.session_state.preferences_1.append(additional_1)
-    
-if len(st.session_state.preferences_2) == 0:
-    st.session_state.preferences_2.append(food_2)
-    # if food_2 in st.session_state.food:
-    #     st.session_state.preferences_2.append(food_2)
-    # else:
-    #     st.session_state.additional_2.append(food_2_o)
-    st.session_state.preferences_2.append(ambiance_2)
-    # if ambiance_2 in st.session_state.ambiance:
-    #     st.session_state.preferences_2.append(ambiance_2)
-    # else:
-    #     st.session_state.additional_2.append(ambiance_2_o)
-    st.session_state.restrictions.extend(options_food_2)
-    if additional_2:
-        st.session_state.preferences_2.append(additional_2)
-    if with_kids_2:
-        st.session_state.restrictions.append('kids')
-
 submitted = st.button('Submit!')
     
 if submitted:
     with st.spinner('Processing your request...'):
         time.sleep(1)
+        if len(st.session_state.preferences_1) == 0:
+            st.session_state.preferences_1.append(food_1)
+            # if food_1 in st.session_state.food:
+            #     st.session_state.preferences_1.append(food_1)
+            # else:
+            #     st.session_state.additional_1.append(food_1_o)
+            st.session_state.preferences_1.append(ambiance_1)
+                
+            # if ambiance_1 in st.session_state.ambiance:
+            #     st.session_state.preferences_1.append(ambiance_1)
+            # else:
+            #     st.session_state.additional_1.append(ambiance_1_o)
+            st.session_state.restrictions.extend(options_food_1)
+            if with_kids:
+                st.session_state.restrictions.append('kids')
+            if additional_1:
+                st.session_state.preferences_1.append(additional_1)
+    
+        if len(st.session_state.preferences_2) == 0:
+            st.session_state.preferences_2.append(food_2)
+            # if food_2 in st.session_state.food:
+            #     st.session_state.preferences_2.append(food_2)
+            # else:
+            #     st.session_state.additional_2.append(food_2_o)
+            st.session_state.preferences_2.append(ambiance_2)
+            # if ambiance_2 in st.session_state.ambiance:
+            #     st.session_state.preferences_2.append(ambiance_2)
+            # else:
+            #     st.session_state.additional_2.append(ambiance_2_o)
+            st.session_state.restrictions.extend(options_food_2)
+            if additional_2:
+                st.session_state.preferences_2.append(additional_2)
+            if with_kids_2:
+                st.session_state.restrictions.append('kids')
+
     st.success("Thanks, we received your preferences!")
     
 else:
@@ -389,29 +415,33 @@ if submit or (not st.session_state.precalculated_df.empty):
     with st.spinner("Please wait while we are finding the best solution..."):
         if st.session_state.precalculated_df.empty:
             query = get_combined_preferences(st.session_state.preferences_1, st.session_state.preferences_2)
-            st.write("Your query is:", query[0])
             #sort places based on restrictions
             st.session_state.precalculated_df = filter_places(st.session_state.restrictions)
             #sort places by elevating preferrences
-            st.session_state.precalculated_df = promote_places(query[1])
+            
+            # st.session_state.precalculated_df = promote_places()
             st.session_state.precalculated_df = compute_cos_sim(query[0])
         sort_by = st.selectbox(('Sort by:'), st.session_state.options, key=400, 
                                index=st.session_state.options.index('Relevancy (default)'))
         if sort_by:
             st.session_state.sort_by = sort_by
             with st.spinner(f"Sorting your results by {sort_by.lower()}..."):
-                results = generate_results(st.session_state.sort_by)
+                if len(st.session_state.results) == 0:
+                    generate_results()
+                results = st.session_state.results[sort_by]
+                if sort_by == 'Distance':
+                    st.write(":pensive: Sorry, we are still working on this option. For now, the results are sorted by relevance")      
                 k = 10
                 st.write(f"Here are the best {k} matches to your preferences:")
                 i = 1
                 nums = list(range(1, 11))
                 words = ['one', 'two', 'three', 'four', 'five', 'six', 'seven', 'eight', 'nine', 'one: :zero']
                 nums_emojis = dict(zip(nums, words))
-                for name, score in results.items():
-                    
+                results = get_normalized_val(results)
+                for name, score in results.items(): 
                     condition = st.session_state.precalculated_df['Names'] == name
                     rating = st.session_state.precalculated_df.loc[condition, 'Rating'].values[0]
-                    with st.expander(f":{nums_emojis[i]}: **{name}** **({str(rating)}**:star:): match score: {score}"):
+                    with st.expander(f":{nums_emojis[i]}: **{name}** **({str(rating)}**:star:): match score: {score}%"):
                         
                         #f":{nums_emojis[i]}: **{name}** **({str(rating)}**:star:) :", 'match score:', score
                         try:
@@ -419,29 +449,36 @@ if submit or (not st.session_state.precalculated_df.empty):
                                 st.write("Price category:", st.session_state.precalculated_df.loc[condition, 'Price'].values[0])
                         except:
                             pass
-                            
-                        # Use the condition to extract the value(s)
-                        # description = st.session_state.precalculated_df.loc[condition, 'Strings']
-                        # st.write(description)
                         
+                        descr = st.session_state.precalculated_df.loc[condition, 'Strings'].values[0]
+                        for word in set([word.lower() for word in descr.split()]):
+                            if word in st.session_state.fixed_preferences:
+                                st.markdown(f'✅{word.capitalize()}')
+                               
+                        
+                        #Restaurant category
                         type = [item for item in eval(st.session_state.precalculated_df.loc[condition, 'Category'].values[0])]
+                        st.markdown(text_css, unsafe_allow_html=True)
+                        st.markdown('<div class="text">Category</div>', unsafe_allow_html=True)
+                        
                         # Display HTML with the custom styles
                         for word in type:
                             st.markdown(css, unsafe_allow_html=True)
                             st.markdown(f'<div class="blue-box">{word}</div>', unsafe_allow_html=True)
-                        # st.write("Restaurant type:", str(type))
-                        
+                           
                         keywords = [item[0] for item in eval(st.session_state.precalculated_df.loc[condition, 'Keywords'].values[0]) if item[1] > 2]
+                        if len(keywords) > 0:
+                            st.markdown(text_css, unsafe_allow_html=True)
+                            st.markdown('<div class="text">Other users say:</div>', unsafe_allow_html=True)
+                        
                         for pair in keywords[:3]:
                             st.markdown(css, unsafe_allow_html=True)
-                            st.markdown(f'<div class="orange-box">{pair[0]} {pair[1]}</div>', unsafe_allow_html=True)
-                        # st.write("Restaurant type:", str(type))
-                        
+                            st.markdown(f'<div class="orange-box">{pair[0]} {pair[1]}</div>', unsafe_allow_html=True)                        
                         
                         url = st.session_state.precalculated_df.loc[condition, 'URL'].values[0]
                         st.write(f"_Check on the_ [_map_]({url})")
-                        
-                        st.write(st.session_state.precalculated_df.loc[condition, 'Strings'].values[0])
+                         
+                        st.write(descr)
                                         
                     i+=1
                             
@@ -449,7 +486,8 @@ if submit or (not st.session_state.precalculated_df.empty):
                 # st.markdown("<span style='font-size: 24px;'>This is larger text</span>", unsafe_allow_html=True)
 
         
-
+            st.session_state.preferences_1, st.session_state.preferences_2 = [], []
+            st.session_state.restrictions = []
         
 stop = st.button("New search!", type='primary', key=500)  
 if stop:               
@@ -460,8 +498,6 @@ if stop:
     st.session_state.sort_by = ""
     st.session_state.df = init_df
     st.session_state.precalculated_df = pd.DataFrame()
-            
-# #TODO: implement price range as a sliding bar
-# When the user presses "New search", erase everything
-# Propose URLs
-# Show keywords instead of whole strings
+    st.session_state.results = {}
+    st.session_state.fixed_preferences = []
+