nbest-lattice.1

这是一款很好用的工具包

              If  all  three of _a_m_w, _l_m_w, and _w_t_w are set to zero
              the  posteriors  are  computed  directly  from  the
              aggregate scores stored in the N-best input.

       --vvooccaabb _f_i_l_e
              Read  the  N-best  list vocabulary from _f_i_l_e.  This
              option is mostly redundant since words found in the
              N-best  input  are  implicitly added to the vocabu-
              lary.

       --vvooccaabb--aalliiaasseess _f_i_l_e
              Reads vocabulary alias definitions from _f_i_l_e,  con-
              sisting of lines of the form
                   _a_l_i_a_s _w_o_r_d
              This  causes all tokens _a_l_i_a_s to be mapped to _w_o_r_d.

       --ttoolloowweerr
              Map vocabulary to lowercase, eliminating case  dis-
              tinctions.

       --mmuullttiiwwoorrddss
              Split  multiwords  (words joined by '_') into their
              components when reading N-best lists.

       --nnooiissee _n_o_i_s_e_-_t_a_g
              Designate _n_o_i_s_e_-_t_a_g as a vocabulary item that is to
              be  ignored  in aligning hypotheses with each other
              (the same as the -pau- word).   This  is  typically
              used to identify a noise marker.

       --nnooiissee--vvooccaabb _f_i_l_e
              Read  several  noise tags from _f_i_l_e, instead of, or
              in addition to, the single noise tag  specified  by
              --nnooiissee.

       --kkeeeepp--nnooiissee
              Do  not  remove pause or noise tokens from hypothe-
              ses. The default is  to  preserve  noise  tags  but
              still eliminate pauses.

       --nnbbeesstt--eerrrroorr
              Compute  the  N-best  error (minimum word error) of
              the N-best list read with --nnbbeesstt.  Pause and  noise
              tokens  (as  specified  with  --nnooiissee) in the N-best
              list are ignored.

       --dduummpp--ppoosstteerriioorrss
              Output  posterior  probabilities  of   all   N-best
              hypotheses instead of choosing the best hypothesis.
              In N-best mode, only the posterior probability  for
              each  hypothesis  is  output.  In lattice mode, the
              hyp posterior is followed by word posterior  proba-
              bilities  for  each (non-pause, non-noise) token in
              the hypothesis.  The --mmaaxx--rreessccoorree option limits the
              number of hypotheses per N-best list processed.

       --dduummpp--eerrrroorrss
              Output  word  correctness indicators for all N-best
              hypotheses instead of choosing the best hypothesis.
              For  each  hypothesis,  a line is output containing
              first the total number of errors and  the  list  of
              indicators  of  whether  the  corresponding word is
              correct, substituted or inserted  relative  to  the
              reference string.  The location of deleted words is
              also indicated  by  a  corresponding  marker.   The
              --mmaaxx--rreessccoorree option limits the number of hypotheses
              per N-best list processed.

       --rreeffeerreennccee _w_1 _w_2 _._._.
              Specifies a reference word string for --dduummpp--eerrrroorrss,
              --nnbbeesstt--eerrrroorr,  and  --llaattttiiccee--eerrrroorr  options.  Addi-
              tionally, in --uussee--mmeesshh mode,  the  reference  words
              are  recorded  in  the  word mesh and can be output
              with --wwrriittee, indicating which word in  each  align-
              ment position is the correct one.

       --rreeffss _r_e_f_e_r_e_n_c_e_s
              Read  a  table  of  reference transcripts from file
              _r_e_f_e_r_e_n_c_e, for when multiple N-best lists are  pro-
              cessed (see --nnbbeesstt--ffiilleess).  Each line in _r_e_f_e_r_e_n_c_e_s
              must contain the sentence ID (the last component in
              the  N-best filename path, minus any suffixes) fol-
              lowed by zero or more reference words.

       The following options only affect lattice mode.

       --rreeaadd _f_i_l_e
              Reads an initial lattice from _f_i_l_e,  to  be  merged
              with  additional  paths constructed from the N-best
              hypotheses.

       --llaattttiiccee--ffiilleess _f_i_l_e
              Reads the names of one or more lattices  from  _f_i_l_e
              and  aligns  those  lattices  with the main lattice
              being built.  Each line of _f_i_l_e must contain a lat-
              tice filename, optionally followed by a weight.

       --wwrriittee _f_i_l_e
              Writes the resulting word posterior lattice or mesh
              to _f_i_l_e, in wwllaatt--ffoorrmmaatt(5).

       --wwrriittee--ddiirr _d_i_r_e_c_t_o_r_y
              Write the resulting N-best lattices  to  _d_i_r_e_c_t_o_r_y,
              in  files  named  after the input N-best lists, for
              when  multiple  N-best  lists  are  processed  (see
              --nnbbeesstt--ffiilleess).

       --pprriimmee--llaattttiiccee
              Start building the lattice with the best hypothesis
              obtained from N-best error minimization.  This pro-
              duces  slightly  better  alignments  and  sometimes
              lower error rates.  The default is  to  start  with
              the top-scoring hypothesis.

       --pprriimmee--wwiitthh--11bbeesstt
              Similar  to --pprriimmee--llaattttiiccee, but uses the top-ranked
              sentence hypothesis for priming.  (Experience shows
              that    --nnoo--rreeoorrddeerr   --pprriimmee--llaattttiiccee   gives   best
              results.)

       --pprriimmee--wwiitthh--rreeffss
              Similar to --pprriimmee--llaattttiiccee, but uses  the  reference
              words for priming.

       --nnoo--mmeerrggee
              Build  a lattice from the N-best hypotheses without
              merging  edges  (string/lattice  alignment).   This
              creates  a  lattice  with  one  disjoint  path  per
              hypothesis, and is useful mainly for debugging pur-
              poses.

       --llaattttiiccee--eerrrroorr
              Compute  the  lattice error (minimum word error) of
              the lattice read with --rreeaadd or built with --nnbbeesstt.

       --ddiiccttiioonnaarryy file
              Use word pronunciations listed in _f_i_l_e to construct
              word  alignments  when  building word meshes.  This
              will use an alignment cost function  that  reflects
              the  number of inserted/deleted/substituted phones,
              rather than words.  The dictionary _f_i_l_e should con-
              tain one pronunciation per line, each naming a word
              in the first field, followed by a string  of  phone
              symbols.

       --hhiiddddeenn--vvooccaabb file
              Read  a  subvocabulary from _f_i_l_e and constrain word
              meshes to only align those words  that  are  either
              all  in  or outside the subvocabulary.  This may be
              used to keep ``hidden event''  tags  from  aligning
              with regular words.

       --rreeccoorrdd--hhyyppss
              Record  the  ranks of the hyps contributing to each
              word hypothesis in the resulting word lattice;  the
              information is included in --wwrriittee output.

SSEEEE AALLSSOO
       ngram(1),  nbest-optimize(1), nbest-scripts(1), nbest-for-
       mat(5), wlat-format(5).
       A. Stolcke, Y. Konig, and M.  Weintraub,  ``Explicit  Word
       Error  Minimization  in  N-best  List  Rescoring,''  _P_r_o_c_.
       _E_u_r_o_s_p_e_e_c_h, 163-166, 1997.
       The ``word meshes'' used here are equivalent to the ``con-
       fusion networks'' described in: L. Mangu, E. Brill, and A.
       Stolcke, ``Finding Consensus  Among  Words:  Lattice-based
       Word  Error  Minimization.''  _P_r_o_c_.  _E_u_r_o_s_p_e_e_c_h,  vol.  1,
       495-498, 1999.

BBUUGGSS
       Several functions are not uniformly  implemented  for  all
       rescoring   modes   (e.g.,   --llaattttiiccee--ffiilleess,  --ddiiccttiioonnaarryy,
       --rreeccoorrdd--hhyyppss, and --nnbbeesstt--bbaacckkttrraaccee are currently effective
       only in mesh-lattice mode).
       It  is  a common mistake (not a bug) to use the default LM
       weight with N-best lists directly from Decipher.  Decipher
       N-best lists have the recognizer's LM weight already built
       in, so they should be processed with
            nbest-lattice -rescore-lmw 1 -posterior-scale _L_M_W
       where _L_M_W is the LM weight during  recognition.   This  is
       not  an  issue if the N-best lists have been rescored with
       rreessccoorree--ddeecciipphheerr.

AAUUTTHHOORR
       Andreas Stolcke <[e-mail]>.
       Copyright 1996-2004 SRI International



SRILM Tools        $Date: 2006/07/05 08:24:08 $  nbest-lattice(1)