Update transparency_note.md

transparency_note.md

@@ -9,14 +9,14 @@ We have released this tool as-is. Thanks!
 
 We note the following about our approach
 
- - Deep Image Networks might be biased towards particular shapes, forms, races, and cultures as the networks we use are trained on the ImageNet dataset, which might over- or underrepresent certain groups or demonstrate stereotypes
+ - Deep Image Networks might be biased towards particular shapes, forms, races, and cultures as the networks we use are trained on the ImageNet dataset, which might over- or underrepresent certain groups or demonstrate stereotypes.
 
 - The Artwork used in our approach is from the open access collections of the MET and Rijksmusem and represent a particular curatorial view that is not necessarily representative of the worlds art.
 
 - Artwork itself may be insensitive to some as morals and societal concepts of fairness vary across culture and time.
 
-- We can only use cultures and media from the art collections and this might be not very accurate or this information can be biased towards a particular culture
+- We can only use cultures and media from the art collections and this might be not very accurate or this information can be biased towards a particular culture.
 
-- We provide many matches for each image and allow users to match with any image in the collection and any culture. The flexibility of this tool can allow for many items to be matched together including ones that might be insensitive. For example you could search for an animal and then ask it to match against photos of people, the method will return the most visually similiar pair,but this does not mean that this should be interpreted as our algorithm saying person X is an animal/ looks like an animalLikewise the method might just be flaky and find matches that are more random than anything else, thus we want to ensure people do not take the matches as definitive or as the only way two cultures are connected
+- We provide many matches for each image and allow users to match with any image in the collection and any culture. The flexibility of this tool can allow for many items to be matched together including ones that might be insensitive. For example you could search for an animal and then ask it to match against photos of people, the method will return the most visually similiar pair,but this does not mean that this should be interpreted as our algorithm saying person X is an animal/ looks like an animal. Likewise the method might find matches that are more random than anything else, thus we want to ensure people do not take the matches as definitive or as the only way two cultures are connected.
 
 - We want people to use this as a tool of cultural celebration and discovery!