The following is a summary of the principal changes made to the POS 
taglist since the "Arabic Treebank: Part2 v 2.0" corpus [LDC2004T02]):

(1) the accusative/genitive marker ACCGEN has been decoupled 
    into ACC and GEN:

    CASE_DEF_ACCGEN             -->  CASE_DEF_ACC, CASE_DEF_GEN
    CASE_INDEF_ACCGEN           -->  CASE_INDEF_ACC, CASE_INDEF_GEN
    NSUFF_FEM_DU_ACCGEN         -->  NSUFF_FEM_DU_ACC, NSUFF_FEM_DU_GEN
    NSUFF_FEM_DU_ACCGEN_POSS    -->  NSUFF_FEM_DU_ACC_POSS, NSUFF_FEM_DU_GEN_POSS
    NSUFF_MASC_DU_ACCGEN        -->  NSUFF_MASC_DU_ACC, NSUFF_MASC_DU_GEN
    NSUFF_MASC_DU_ACCGEN_POSS   -->  NSUFF_MASC_DU_ACC_POSS, NSUFF_MASC_DU_GEN_POSS
    NSUFF_MASC_PL_ACCGEN        -->  NSUFF_MASC_PL_ACC, NSUFF_MASC_PL_GEN
    NSUFF_MASC_PL_ACCGEN_POSS   -->  NSUFF_MASC_PL_ACC_POSS, NSUFF_MASC_PL_GEN_POSS

(2) "long" POS tags have been re-spelled in shorter form:

    VERB_IMPERATIVE           -->  CV
    VERB_IMPERFECT            -->  IV
    VERB_IMPERFECT_PASSIVE    -->  IV_PASS
    VERB_PERFECT              -->  PV
    VERB_PERFECT_PASSIVE      -->  PV_PASS
    EMPHATIC_PART             -->  EMPH_PART
    RESULT_CLAUSE_PARTICLE    -->  RC_PART
    SUBJUNC                   -->  SUB

(3) new POS tags have been introduced for more precise tagging:

    FOCUS_PART  focus particle
    JUS         jussive prefix (e.g. li/JUS+ya/IV3MS+kun/IV+o/IVSUFF_MOOD:J 
                                     let + him/it + be + [jus.]