Update README.md

README.md

@@ -81,7 +81,8 @@ mask_filler = pipeline("fill-mask",
 mask_filler(f"1820 [DATE] We received a letter from [MASK] Majesty.")
 ```
 
-Returns as most likely prediction
+Returns as most likely prediction:
+
 ```python
 {'score': 0.8527863025665283,
   'token': 2010,
@@ -95,11 +96,22 @@ However, if we change the date at the start of the sentence to 1850:
 mask_filler(f"1820 [DATE] We received a letter from [MASK] Majesty.")
 ```
 
-Will put more probability mass on the token 'her`.
+Will put most of probability mass on the token "her" and only a little bit on "him".
+
+```python
+{'score': 0.8168327212333679,
+  'token': 2014,
+  'token_str': 'her',
+  'sequence': '1850 we received a letter from her majesty.'}
+```
+
+You can repeat this experiment for yourself using the example sentences in the **Hosted inference API** at the top right.
+
+Okay, but why is this interesting? 
 
-You can try this for yourself in Example sentences at the top right.
+Firstly, eyeballing some toy-examples (but also using more rigorous metrics such as perplexity) shows that MLMs can perform more accurate predictions when it has access to temporal metadata. In other words, ERWT's prediction reflects historical language use more accurately. Model that are sensitive to historical context could 
 
-But why is this interesting? Firstly 
+Secondly, we anticipate the MDMA may reduce bias, or at least gives us more of a handle on this problem. Admittedly, we have to prove this more formally, but some experiments at least hint in this direction.
 
 ### Date Prediction