Natural    language    processing    (NLP)    is    the    technique  by  which  we  process  the  human  language  with  the  computer.   Parts-of-Speech   (POS)   tagging   is   one   of   the   fundamental  requirements  for  some  NLP  applications.  It  is  considered  as  a  solved  problem  for  some  foreign  languages,  such as English, Chinese, due to higher accuracy (97%), where it   is   still   an   unsolved   problem   for   Bangla   because   of   its   ambiguity. Although making a POS tagger for Bangla is not a new work, but each one of available POS taggers has different kinds  of  limitations.  We  choose  to  develop  an  unsupervised  system  rather  than  a  supervised  system,  because  a  supervised  system  needs  a  huge  data  resource  for  training  purpose  and  available resources in Bangla is really poor.  Here we develop a POS   tagger   mainly   based   on   Bangla   grammar   especially   suffixes.  Because  Bangla  is  a  very  inflectional  language,  where  a single word has many variants based on their suffixes. In this POS  tagger,  we  assign  8  base  POS  tags,  where  some  rules,  based  on  Bangla  grammar  and  suffix,  are  applied  to  identify  POS  tags  with  the  cooperation  of  verb  root  dataset.  To  handle  non-suffix words, a dataset of almost 14500 Bangla words, with having their default POS tags, is added with the system, which helps  to  increase  the  efficiency  of  this  POS  tagger.  A  modified  version  of  previously  used  algorithm  for  suffix  analysis  is  applied, which result in a satisfactory level of about 94.2%