edits

README.md

@@ -1,8 +1,8 @@
 ---
-title: Ferret Demo
-emoji: 🌖
-colorFrom: red
-colorTo: red
+title: ferret
+emoji: 🦡
+colorFrom: blue
+colorTo: purple
 sdk: streamlit
 sdk_version: 1.10.0
 app_file: app.py

requirements.txt

@@ -1,3 +1,3 @@
 transformers==4.20.1
-ferret-xai==0.3.0
+ferret-xai==0.3.1
 sentencepiece

single.py

@@ -30,14 +30,14 @@ def body():
         You are working now on the *single instance* mode -- i.e., you will work and
         inspect one textual query at a time.
  
-        Post-hoc explanation techniques discose the rationale behind a given prediction a model
-        makes while detecting a sentiment out of a text. In a sense the let you *poke* inside the model.
+        Post-hoc explanation techniques disclose 🔎 the rationale behind a given prediction a model
+        makes while detecting a sentiment out of a text. In a sense, they let you *poke* inside the model.
 
-        But **who watches the watchers**?
+        But **who watches the watchers**? Are these explanations *accurate*? Can you *trust* them?
 
         Let's find out!
 
-        Let's choose your favourite sentiment classification mode and let ferret do the rest.
+        Let's choose your favourite mode and let *ferret* do the rest.
         We will:
 
         1. download your model - if you're impatient, here it is a [cute video](https://www.youtube.com/watch?v=0Xks8t-SWHU) 🦜 for you;