No hits from pubmed fix

app.py

@@ -9,8 +9,8 @@ from langdetect import detect
 from typing import Dict, List
 
 
-if "valid_inputs_received" not in st.session_state:
-    st.session_state["valid_inputs_received"] = False
+if 'valid_inputs_received' not in st.session_state:
+    st.session_state['valid_inputs_received'] = False
 
 
 def get_articles(query, fetcher) -> Dict[List[str], List[str]]:
@@ -42,11 +42,11 @@ def get_articles(query, fetcher) -> Dict[List[str], List[str]]:
                 titles.append(title)
                 links.append(article_link)
         except Exception as e:
-            logging.warning(f"Error reading article: {article_id}: ", exc_info=e)
+            logging.warning(f'Error reading article: {article_id}: ', exc_info=e)
 
     return {
-        "Conclusions": conclusions,
-        "Links": links
+        'Conclusions': conclusions,
+        'Links': links
     }
 
 
@@ -59,30 +59,30 @@ def load_cross_encoder():
 
 @st.cache_resource
 def load_pubmed_fetcher():
-    pubmed = PubMed(tool="PubmedFactChecker", email="[email protected]")
+    pubmed = PubMed(tool='PubmedFactChecker', email='[email protected]')
     return pubmed
 
 
 def run_ui():
     # This function controls the whole app flow.
-    st.set_page_config(page_title="PUBMED FACT-CHECKER", page_icon="📖")
+    st.set_page_config(page_title='PUBMED FACT-CHECKER', page_icon='📖')
 
     sidebar = st.sidebar
     sidebar.title('ABOUT')
-    sidebar.write("""
+    sidebar.write('''
                 The PubMed fact-checker app enables users to verify biomedical claims by comparing them against 
                 research papers available on PubMed. \n
                 As the number of self-proclaimed experts continues to rise, 
                 so does the risk of harmful misinformation. This app showcases the potential of Large Language Models 
                 to provide accurate and valuable information to people.
-            """)
+            ''')
     sidebar.title('EXAMPLES')
     sidebar.write('Try one of the below examples to see PubMed fact-checker in action.')
 
     st.title('PubMed FACT CHECKER')
-    with st.form(key="fact_form"):
-        fact = st.text_input('Fact:', placeholder='Enter your fact')
-        submitted = st.form_submit_button("Fact-Check")
+    with st.form(key='fact_form'):
+        fact = st.text_input('Fact:', placeholder='Enter your fact', key='form_input')
+        submitted = st.form_submit_button('Fact-Check')
 
     if sidebar.button('Mediterranean diet helps with weight loss.', use_container_width=250):
         submitted = True
@@ -97,7 +97,7 @@ def run_ui():
         fact = 'Vaccines are a cause of autism.'
     
     sidebar.title('HOW IT WORKS')
-    sidebar.write('Source code and in-depth app description available at:')
+    sidebar.write('Source code and an in-depth app description available at:')
     sidebar.info('**GitHub: [@jacinthes](https://github.com/jacinthes/slovene-nli-benchmark)**', icon="💻")
     sidebar.title('DISCLAIMER')
     sidebar.write('This project is meant for educational and research purposes. \n'
@@ -112,7 +112,8 @@ def run_ui():
         st.stop()
 
     elif submitted and not detect(fact) == 'en':
-        st.warning('Please enter valid text in English. For short inputs, language detection is sometimes inaccurate.')
+        st.warning('Please enter valid text in English. For short inputs, language detection is sometimes inaccurate.'
+                   ' Try making the fact more verbose.')
         st.session_state.valid_inputs_received = False
         st.stop()
 
@@ -130,7 +131,17 @@ def run_ui():
 
         article_conclusions = articles['Conclusions']
         article_links = articles['Links']
+        if len(article_conclusions) == 0:
+            # If nothing is returned by pymed, inform user.
+            st.info(
+                "Unfortunately, I couldn't find anything for your search.\n"
+                "Don't let that discourage you, I have over 35 million citations in my database.\n"
+                "I am sure your next search will be more successful."
+            )
+            st.stop()
+
         cross_inp = [[fact, conclusions] for conclusions in article_conclusions]
+
         with st.spinner('Assessing article relevancy...'):
             cross_encoder = load_cross_encoder()
             cross_scores = cross_encoder.predict(cross_inp)  # Calculate relevancy using the defined cross-encoder.
@@ -142,7 +153,7 @@ def run_ui():
         })
         df.sort_values(by=['Score'], ascending=False, inplace=True)
         df = df[df['Score'] > 0]  # Only keep articles with relevancy score above 0.
-        if df.shape[0] == 0:  # If no relevant article si found, inform the user.
+        if df.shape[0] == 0:  # If no relevant article is found, inform the user.
             st.info(
                 "Unfortunately, I couldn't find anything for your search.\n"
                 "Don't let that discourage you, I have over 35 million citations in my database.\n"
@@ -151,7 +162,7 @@ def run_ui():
             st.stop()
 
         df = df.head(10)  # Keep only 10 most relevant articles. This is done to control OpenAI costs and load time.
-        progress_text = "Assessing the validity of the fact based on relevant research papers."
+        progress_text = 'Assessing the validity of the fact based on relevant research papers.'
         fact_checking_bar = st.progress(0, text=progress_text)
         step = 100/df.shape[0]
         percent_complete = 0