Update README.md

README.md

@@ -94,8 +94,11 @@ Each pair in the `generation` config will include the same fields except `text1`
 
 **Context**: Most (but not all) of the paraphrase pairs are identified in their document context. By default, these contexts are not included to conserve memory, but can be accessed using the configurations `plain-context` and `classification-context`. These are exactly like `plain` and `classification` with these additional fields:
 
-`context1`: a dictionary with the fields `doctext` (string), `begin` (int), `end` (int). These mean that the paraphrase in `text1` was extracted from `doctext[begin:end]`. In most cases, `doctext[begin:end]` and `text1` are the exact same string, but occassionally that is not the case when e.g. intervening punctuations or other unrelated texts were "cleaned" from `text1` during annotation. In case the context is not available, `doctext` is an empty string and `beg==end==0`
-`context2`: same as `context1` but for `text2`
+`context1`: 
+a dictionary with the fields `doctext` (string), `begin` (int), `end` (int). These mean that the paraphrase in `text1` was extracted from `doctext[begin:end]`. In most cases, `doctext[begin:end]` and `text1` are the exact same string, but occassionally that is not the case when e.g. intervening punctuations or other unrelated texts were "cleaned" from `text1` during annotation. In case the context is not available, `doctext` is an empty string and `beg==end==0`
+
+`context2`: 
+same as `context1` but for `text2`
 
 ### Data Splits