Update app.py

app.py

@@ -1,5 +1,6 @@
 import json
 import requests
+import re
 from relbert import RelBERT
 import gradio as gr
 
@@ -21,6 +22,12 @@ def cosine_similarity(a, b, zero_vector_mask: float = -100):
     return sum(map(lambda x: x[0] * x[1], zip(a, b)))/(norm_a * norm_b)
 
 
+def clean(text):
+    text = re.sub(r"\A\s+", "", text)
+    text = re.sub(r"\s+\Z", "", text)
+    return text
+
+
 def greet(
         query,
         candidate_1,
@@ -29,7 +36,7 @@ def greet(
         candidate_4,
         candidate_5,
         candidate_6):
-    query = query.split(',')
+    query = [clean(i) for i in query.split(',')]
     # validate query
     if len(query) == 0:
         raise ValueError(f'ERROR: query is empty {query}')
@@ -50,7 +57,7 @@ def greet(
     ]):
         if i == '':
             continue
-        candidate = i.split(',')
+        candidate = [clean(x) for x in i.split(',')]
         if len(candidate) == 1:
             raise ValueError(f'ERROR: candidate {n + 1} contains single word {candidate}')
         if len(candidate) > 2:
@@ -64,9 +71,6 @@ def greet(
     sims = []
     for v in vectors:
         sims.append(cosine_similarity(v, vector_q))
-    # output = list(zip(pairs_id, sims, pairs))
-    # output = sorted(list(zip(pairs_id, sims, pairs)), key=lambda _x: _x[1], reverse=True)
-    # output = {f'candidate {i}: [{p[0]}, {p[1]}]': s for n, (i, s, p) in enumerate(output)}
     output = {f'candidate {i}: [{p[0]}, {p[1]}]': s for i, s, p in zip(pairs_id, sims, pairs)}
     return output