I agree with davido - your question is too vague. What I see in the OP are hexadecimal entity references for Chinese characters. Maybe you have entity references to other non-ASCII characters as well? If you just want to delete all such entity references:

    s/\&#x\w+;//g;
[download]

(Note that the number of hex digits per character may vary.) In the case of the OP data sample, that regex leaves 7 lines that are either empty or contain just hyphens and/or spaces.