segment-nbest.1

这是一款很好用的工具包

segment-nbest(1)                                 segment-nbest(1)



NNAAMMEE
       segment-nbest  -  rescore  and  segment N-best lists using
       hidden segment N-gram model

SSYYNNOOPPSSIISS
       sseeggmmeenntt--nnbbeesstt [--hheellpp] option ...  nbest-file-list ...

DDEESSCCRRIIPPTTIIOONN
       sseeggmmeenntt--nnbbeesstt processes a  series  of  consecutive  N-best
       lists  from  a speech recognizer and applies a hidden seg-
       ment N-gram language model to them.  The language model is
       a  standard  backoff  N-gram model in ARPA nnggrraamm--ffoorrmmaatt(5)
       modeling sentence segmentation using the boundary tags <s>
       and  </s>.  The program reads in all N-best lists and out-
       puts the hypotheses that have the highest aggregate  (com-
       bined  acoustic  and  language model) score.  Hypothesized
       sentence boundaries are marked by <s> tags.

OOPPTTIIOONNSS
       Each filename argument can be an ASCII  file,  or  a  com-
       pressed file (name ending in .Z or .gz), or ``-'' to indi-
       cate stdin/stdout.

       --hheellpp  Print option summary.

       --vveerrssiioonn
              Print version information.

       --oorrddeerr _n
              Set the maximal N-gram order to be used, by default
              3.   NOTE:  The order of the model is not set auto-
              matically when a model file is read,  so  the  same
              file can be used at various orders.

       --ddeebbuugg _l_e_v_e_l
              Set  the  debugging output level (0 means no debug-
              ging  output).   Debugging  messages  are  sent  to
              stderr.

       --llmm _f_i_l_e
              Read the N-gram model from _f_i_l_e.

       --ttoolloowweerr
              Map  all  vocabulary  to lowercase.  Useful if case
              conventions for N-best  lists  and  language  model
              differ.

       --mmiixx--llmm _f_i_l_e
              Read a second, standard N-gram model for interpola-
              tion purposes.

       --llaammbbddaa _w_e_i_g_h_t
              Set the weight of the main model when interpolating
              with --mmiixx--llmm.  Default value is 0.5.

       --bbaayyeess _l_e_n_g_t_h
              Interpolate  the  second  and  the main model using
              posterior probabilities for  local  N-gram-contexts
              of  length  _l_e_n_g_t_h.  The --llaammbbddaa value is used as a
              prior mixture weight in this case.

       --bbaayyeess--ssccaallee _s_c_a_l_e
              Set the exponential scale  factor  on  the  context
              likelihood in conjunction with the --bbaayyeess function.
              Default value is 1.0.

       --nnbbeesstt--ffiilleess _l_i_s_t
              Specifies a list of N-best files.   The  file  _l_i_s_t
              should  contain  a list of filenames, one per line,
              each corresponding to an N-best file in one of  the
              formats  described  in nnbbeesstt--ffoorrmmaatt(5).  The N-best
              files should correspond to consecutive speech wave-
              forms in the order listed.

       --ffbb--rreessccoorree
              Perform Forward-backward rescoring.  This generates
              new N-best lists as output whose LM scores  reflect
              the  posterior probability of each hypothesis.  The
              default is to perform Viterbi rescoring and  output
              only the best combined hypothesis.

       --wwrriittee--nnbbeesstt--ddiirr _d_i_r
              Write   rescored  N-best  lists  to  directory  _d_i_r
              instead of to stdout.  The filenames from the input
              are preserved.

       --mmaaxx--nnbbeesstt _n
              Limits  the  number of hypotheses read from each N-
              best list to the first _n.

       --mmaaxx--rreessccoorree _m
              Only choose among the top _m hypotheses of each list
              (after  reordering hypotheses, see below).  This is
              an effective way to limit the quadratic computation
              of the Viterbi or forward/backward dynamic program-
              ming.

       --nnoo--rreeoorrddeerr
              Do not reorder the hypotheses before  limiting  the
              computation  to the top _m.  By default the hypothe-
              ses will first be sorted according to the  acoustic
              and  language  model  scores recorded in the N-best
              lists.

       --rreessccoorree--llmmww _w_e_i_g_h_t
              Specifies the language model weight to  be  use  in
              combining  acoustic  and  language  model scores to
              select the best hypotheses.

       --rreessccoorree--wwttww _w_e_i_g_h_t
              Specifies the word transition weight to be used  in
              selecting the best hypotheses.

       --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 by the LM.  (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.

       --ddeecciipphheerr--llmm _m_o_d_e_l_-_f_i_l_e
              Designates  the  N-gram  backoff model (typically a
              bigram) that was used by  the  Decipher(TM)  recog-
              nizer  in computing composite scores.  Used to com-
              pute acoustic scores from the composite  scores  if
              the N-best lists are in "NBestList1.0" format.

       --ddeecciipphheerr--llmmww _w_e_i_g_h_t
              Specifies  the  language  model  weight used by the
              recognizer.  Used to compute acoustic  scores  from
              the composite scores.

       --ddeecciipphheerr--wwttww _w_e_i_g_h_t
              Specifies  the  word  transition weight used by the
              recognizer.  Used to compute acoustic  scores  from
              the composite scores.

       --ssttaagg _s_t_r_i_n_g
              Use  _s_t_r_i_n_g  to mark segment boundaries in the out-
              put.   Default  is  the  start-of-sentence   symbol
              defined in the language model (<s>).

       --bbiiaass _b
              Make  a  segment boundary a priori more likely by a
              factor of _b.  If _b is 0, the dynamic program  algo-
              rithm  is  restricted to never consider hidden sen-
              tence boundaries; this is useful when sseeggmmeenntt--nnbbeesstt
              is  used  merely  for  its  ability to apply the LM
              across N-best boundaries.

       --ssttaarrtt--ttaagg _s_t_r_i_n_g
              Insert a tag _s_t_r_i_n_g at the front  of  every  N-best
              hypothesis read in.

       --eenndd--ttaagg _s_t_r_i_n_g
              Insert  a  tag  _s_t_r_i_n_g  at  the end of every N-best
              hypothesis read in.  This and the  previous  option
              are useful if the LM marks acoustic waveform bound-
              aries with a special tag.

       sseeggmmeenntt--nnbbeesstt will also process any command line arguments
       following  the  options  as lists of N-best lists, as with
       the --nnbbeesstt--ffiilleess option.   Each  _n_b_e_s_t_-_f_i_l_e_-_l_i_s_t  will  be
       processed  in  turn, with individual output delimited by a
       line of the form
            <nbestfile _n_b_e_s_t_-_f_i_l_e_-_l_i_s_t>

SSEEEE AALLSSOO
       ngram-count(1),  segment(1),  ngram-format(5),  nbest-for-
       mat(5).
       A.  Stolcke, ``Modeling Linguistic Segment and Turn Bound-
       aries for N-best Rescoring of Spontaneous Speech,''  _P_r_o_c_.
       _E_u_r_o_s_p_e_e_c_h, 2779-2782, 1997.

BBUUGGSS
       N-gram models of arbitrary order can be used, but the con-
       text at the beginning of a hypothesis never extends beyond
       the words from the preceding N-best list.

AAUUTTHHOORR
       Andreas Stolcke <[e-mail]>.
       Copyright 1997-2004 SRI International



SRILM Tools        $Date: 2004/12/03 17:59:01 $  segment-nbest(1)