It looks like your user has embedded HTML character entities in the abstract for some reason. Is this happening for every abstract, or just some of them? Normally, if users just add enter text into the abstract field as they would otherwise enter into a word processor, unadulturated UTF-8 characters would be stored; but in this case, they are either being converted to HTML character entities (ie. ñ becomes ñ), or are being entered as such originally. Try editing an abstract (from the submission's Summary page, click the Edit Metadata link and then edit the abstract) to see how the characters are being displayed there; you may be able to change them to regular UTF-8 in that field.