Update README.md

README.md

@@ -23,4 +23,17 @@ model-index:
     metrics:
       - type: 'bleu'         
         value: 17.2    
----
+---
+
+This is a BART-large model finetuned on roughly 58000 aligned sentence pairs in English and Middle English, collected from the works of Geoffrey Chaucer, John Wycliffe, and the Gawain Poet.
+<br>
+It includes special characters such as þ. 
+<br>
+This model reflects the spelling inconsistencies characteristic of Middle English.
+<br>
+Because the model is trained largely on poetry and some prose, it is best at translating those sorts of tasks.
+<br>
+Performance can be improved by sentence tokenizing input data and translating sentence-by-sentence.
+<br>
+Removing contractions (hadn't -> had not) also boosts performance.
+