Update app.py

app.py

@@ -7,20 +7,19 @@ from sentence_transformers import SentenceTransformer
 import gradio as gr
 #%%
 etalon = pd.read_csv("etalon_prod.csv")
-df = pd.read_csv("preprocessed_train_classify_rec_spec_filtered_by_etalon.csv")
-df = df[df['is_match'] == 1]
+df = pd.read_csv("preprocessed_complaints.csv")
 
-model = SentenceTransformer('sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2')
+model = SentenceTransformer('sentence-transformers/multi-qa-distilbert-cos-v1')
 
 
 unique_complaints = df['Жалобы'].values.tolist()
 
 with open("embeddings.npy", 'rb') as f:
-    unique_complaints_embeddings_st = np.load(f)
+    embeddings = np.load(f)
 
 def get_recommend(user_input, 
                   top_k_spec = 3,
-                  top_k_services = 10,
+                  top_k_services = 5,
                   treshold = 0.8):
     
     cols_for_top_k = ["Специальность врача",
@@ -28,7 +27,7 @@ def get_recommend(user_input,
     
     usr_embeddings = model.encode(user_input)
 
-    cos_similarity = cos_sim(usr_embeddings, unique_complaints_embeddings_st).detach().numpy()
+    cos_similarity = cos_sim(usr_embeddings, embeddings).detach().numpy()
     sorted_idx = cos_similarity[0].argsort()[::-1]
     cos_similarity.sort()
     
@@ -42,6 +41,7 @@ def get_recommend(user_input,
     for col in cols_for_top_k:
         result[col] = sorted_df[col].value_counts()[:top_k_spec].index.tolist()
     result['Жалобы'] = sorted_df['Жалобы'].value_counts()[:top_k_services].index.tolist()
+    result['Диагноз МКБ'] = sorted_df['Диагноз МКБ'].value_counts()[:top_k_services].index.tolist()
     
     lst = []    
     categories = ['Инструментальная диагностика', 'Лабораторная диагностика']
@@ -56,7 +56,7 @@ def get_recommend(user_input,
 gradio_app = gr.Interface(
     get_recommend,
     inputs='text',
-    outputs=gr.JSON(label='s'),
+    outputs=gr.JSON(label='Рекомендации: '),
     # title="Предсказание топ-10 наиболее схожих услуг",
     description="Введите услугу:"
 )