nbest-scripts.1

这是一款很好用的工具包

nbest-scripts(1)                                 nbest-scripts(1)



NNAAMMEE
       nbest-scripts,   combine-rover-controls,   compare-sclite,
       compute-sclite, fix-ctm, merge-nbest, nbest-error,  nbest-
       posteriors,  nbest-rover,  nbest-vocab,  nbest2-to-nbest1,
       rescore-acoustic, rescore-decipher, rescore-reweight, sen-
       tid-to-sclite - rescore and evaluate N-best lists

SSYYNNOOPPSSIISS
       rreessccoorree--ddeecciipphheerr  [--bbyytteelloogg] [--nnooddeecciipphheerrllmm] [--mmuullttiiwwoorrddss]
       [--pprreettttyy _m_a_p_f_i_l_e] [--nnggrraamm--ttooooll _p_r_o_g_r_a_m] [--ffiilltteerr  _c_o_m_m_a_n_d]
       [--nnoorreessccoorree]   [--llmm--oonnllyy]   [--ccoouunntt--oooovvss]   [--lliimmiitt--vvooccaabb]
       [--vvooccaabb--aalliiaasseess _m_a_p_f_i_l_e] [--ffaasstt] _n_b_e_s_t_-_f_i_l_e_-_l_i_s_t _s_c_o_r_e_-_d_i_r
       --llmm ...  _l_m_-_o_p_t_i_o_n_s ...
       rreessccoorree--aaccoouussttiicc _o_l_d_-_n_b_e_s_t_-_d_i_r|_o_l_d_-_f_i_l_e_-_l_i_s_t _o_l_d_-_a_c_-_w_e_i_g_h_t
       _n_e_w_-_s_c_o_r_e_-_d_i_r_1  _n_e_w_-_a_c_-_w_e_i_g_h_t_1  ...   _n_e_w_-_n_b_e_s_t_-_d_i_r  [_m_a_x_-
       _n_b_e_s_t]
       rreessccoorree--rreewweeiigghhtt  [--mmuullttiiwwoorrddss]  _s_c_o_r_e_-_d_i_r|_f_i_l_e_-_l_i_s_t  [_l_m_w
       [_w_t_w [_s_c_o_r_e_-_d_i_r_1 _s_c_o_r_e_-_w_e_i_g_h_t_1 ...]  [_m_a_x_-_n_b_e_s_t]]]
       rreessccoorree--mmiinniimmiizzee--wweerr _s_c_o_r_e_-_d_i_r [_l_m_w [_w_t_w [_m_a_x_-_n_b_e_s_t]]]
       nnbbeesstt22--ttoo--nnbbeesstt11 [_n_b_e_s_t_-_f_i_l_e]
       nnbbeesstt--rroovveerr [ _s_e_n_t_i_d_-_l_i_s_t | -- ] _c_o_n_t_r_o_l_-_f_i_l_e [  _p_o_s_t_e_r_i_o_r_-
       _f_i_l_e [ _n_b_e_s_t_-_l_a_t_t_i_c_e_-_o_p_t_i_o_n_s ] ]
       ccoommbbiinnee--rroovveerr--ccoonnttrroollss  [ llaammbbddaa==_w_e_i_g_h_t_s ] _r_o_v_e_r_-_c_o_n_t_r_o_l [
       ... ]
       nnbbeesstt--ppoosstteerriioorrss [ wweeiigghhtt==_W  llmmww==_l_m_w  wwttww==_w_t_w  ppoossttssccaallee==_S
       mmaaxx__nnbbeesstt==_M ] _n_b_e_s_t_-_f_i_l_e
       mmeerrggee--nnbbeesstt [ mmaaxx__nnbbeesstt==_M mmuullttiiwwoorrddss==11 nnooppaauusseess==11 ] _n_b_e_s_t_-
       _f_i_l_e ...
       nnbbeesstt--vvooccaabb [_n_b_e_s_t_-_l_i_s_t...]
       nnbbeesstt--eerrrroorr   _s_c_o_r_e_-_d_i_r|_f_i_l_e_-_l_i_s_t   _r_e_f_s   [_n_b_e_s_t_-_l_a_t_t_i_c_e_-
       _o_p_t_i_o_n...]
       sseennttiidd--ttoo--sscclliittee _h_y_p_s
       sseennttiidd--ttoo--ccttmm _h_y_p_s
       ffiixx--ccttmm _c_t_m_f_i_l_e
       ccoommppuuttee--sscclliittee  --rr  _r_e_f_s --hh _h_y_p_s [--hh _h_y_p_s ...]  [--SS _s_u_b_s_e_t
       ...]  [--mmuullttiiwwoorrddss|--MM] [--nnooppeerriiooddss] [--RR] [--gg _g_l_m_f_i_l_e] [--HH]
       [--vv] [_s_c_l_i_t_e_-_o_p_t_i_o_n_s...]
       ccoommppaarree--sscclliittee  --rr  _r_e_f_s  --hh11  _h_y_p_s_1 --hh22 _h_y_p_s_2 [--SS _s_u_b_s_e_t]
       [--mmuullttiiwwoorrddss|--MM] [_s_c_l_i_t_e_-_o_p_t_i_o_n_s...]

DDEESSCCRRIIPPTTIIOONN
       These scripts perform common tasks on N-best hypotheses in
       nnbbeesstt--ffoorrmmaatt(5), especially those needed for rescoring and
       extracting and evaluating 1-best hypotheses.

       rreessccoorree--ddeecciipphheerr applies a language model  implemented  by
       nnggrraamm(1)  to  the  N-best lists listed in _n_b_e_s_t_-_f_i_l_e_-_l_i_s_t.
       The  N-best  files  may  be  in  compressed  format.   The
       rescored  N-best  lists are stored in directory _s_c_o_r_e_-_d_i_r.
       All following arguments are passed  to  nnggrraamm(1)  and  are
       used to control the language model.  The following options
       are handled by rreessccoorree--ddeecciipphheerr itself:

       --bbyytteelloogg            causes scores  to  be  output  on  the
                           bytelog scale (see nnbbeesstt--ffoorrmmaatt(1)).

       --nnooddeecciipphheerrllmm       indicates that the recognizer language
                           model  is  not  being  provided  (with
                           --ddeecciipphheerr--llmm).  (This is only possible
                           if  the  N-best  lists  are   not   in
                           ``NBestList1.0'' format.)

       --mmuullttiiwwoorrddss         specifies  that  N-best  lists contain
                           words joined by underscores, which are
                           to be split into their component prior
                           to rescoring.

       --pprreettttyy _m_a_p_f_i_l_e     specifies a  word  mapping  file  that
                           allows individual words to be globally
                           replaced by strings of  zero  or  more
                           other  words,  e.g., to remove vocabu-
                           lary mismatches between the  input  N-
                           best  lists and the rescoring LM.  The
                           _m_a_p_f_i_l_e contains one mapping per line,
                           the first field specifying the word to
                           be  replaced  and  subsequent   fields
                           forming the replacement string.

       --nnggrraamm--ttooooll _p_r_o_g_r_a_m specifies  a  non-standard  _p_r_o_g_r_a_m to
                           perform the actual LM  evaluation  (by
                           default,  nnggrraamm(1)  is  used).  Such a
                           program must understand  nnggrraamm's  com-
                           mand-line  options  related  to N-best
                           rescoring.

       --ffiilltteerr _c_o_m_m_a_n_d     specifies a _c_o_m_m_a_n_d that  is  used  to
                           filter  the N-best hypotheses prior to
                           evaluating the language  model.   This
                           may  be  used for more general textual
                           rewriting so that non-standard LMs can
                           be  applied.   The output N-best lists
                           will contain the filtered  hypotheses.

       --nnoorreessccoorree          causes   N-best  lists  to  be  simply
                           reformatted from one of  the  Decipher
                           formats  into the SRILM N-best format,
                           separating  acoustic  and  LM  scores,
                           without   replacing  the  existing  LM
                           scores.   In  this   case   only   the
                           nnggrraamm(1)   options  --ddeecciipphheerr--llmmww  and
                           --ddeecciipphheerr--wwttww are relevant, and others
                           are  ignored.   --nnoorreessccoorree and --ffiilltteerr
                           may be used together to  perform  tex-
                           tual rewriting of N-best lists.

       --llmm--oonnllyy            dumps  out  LM scores only, instead of
                           complete N-best lists.

       --ccoouunntt--oooovvss         writes the count of  out-of-vocabulary
                           and zero-probability words to the out-
                           put score files (instead  of  rescored
                           N-best lists).

       --lliimmiitt--vvooccaabb        saves memory by arranging for nnggrraamm(1)
                           to load only those  N-gram  parameters
                           that are relevant to the vocabulary of
                           the  N-best  lists  to  be   rescored.
                           After  determining  the N-best vocabu-
                           lary the --lliimmiitt--vvooccaabb option is passed
                           to nnggrraamm(1).

       --vvooccaabb--aalliiaasseess _m_a_p  declares  that certain words are to be
                           treated as  alternative  spellings  of
                           the  same  word for LM evaluation; see
                           the same option for nnggrraamm(1).  The _m_a_p
                           is  filtered of unused words when used
                           in conjunction with --lliimmiitt--vvooccaabb,  and
                           then passed on to nnggrraamm(1).

       --ffaasstt               performs  rescoring  using  only func-
                           tions  built  into   nnggrraamm(1).    This
                           avoids   some  computational  and  I/O
                           overhead and  therefore  runs  faster,
                           but  the options --ffiilltteerr, --pprreettttyy, and
                           --llmm--oonnllyy are not supported, and --nnooddee--
                           cciipphheerrllmm is obligatory.

       rreessccoorree--aaccoouussttiicc  replaces the acoustic scores in a set of
       N-best lists by a weighted combination of new scores.  The
       old  N-best  lists  are  given  by either a directory _o_l_d_-
       _s_c_o_r_e_-_d_i_r or a filelist  _o_l_d_-_f_i_l_e_-_l_i_s_t;  _o_l_d_-_a_c_-_w_e_i_g_h_t  is
       the  weight given to the old scores.  Directories contain-
       ing the new scores are listed alternating with the  corre-
       sponding  weights;  each  score directory must contain one
       file per waveform segment, each having the same file base-
       names as the original N-best lists.  The new scores should
       appear in a single column per file, one per line.  The  N-
       best lists containing the new combined acoustic scores are
       written to _n_e_w_-_n_b_e_s_t_-_d_i_r.  The optional _m_a_x_-_n_b_e_s_t argument
       can  be  used to limit the length of the N-best lists out-
       put.  Also, When a new score file is encountered  contain-
       ing fewer than _m_a_x_-_n_b_e_s_t lines, the missing scores are set
       to the lowest score encountered so far.

       rreessccoorree--rreewweeiigghhtt combines the scores in N-best lists  with
       a  set  of weights and outputs the 1-best hypotheses.  The
       N-best files are found in directory _s_c_o_r_e_-_d_i_r or listed in
       _f_i_l_e_-_l_i_s_t.   Optional  arguments  set  the  language model
       weight _l_m_w (default 8), the  word  transition  weight  _w_t_w
       (default  0), and the maximum number _m_a_x_-_n_b_e_s_t of hypothe-
       ses to consider (default all).  Optionally, any number  of
       additional score directories and associated weights _s_c_o_r_e_-
       _d_i_r_1 _s_c_o_r_e_-_w_e_i_g_h_t_1 _s_c_o_r_e_-_d_i_r_2  _s_c_o_r_e_-_w_e_i_g_h_t_2  ...  can  be
       specified,  following the _w_t_w parameter.  These additional
       scores are combined with those  contained  in  the  N-best
       lists themselves as in rreessccoorree--aaccoouussttiicc (using unit weight
       for the original acoustic scores).  --mmuullttiiwwoorrddss  indicates
       that  multi-words  are  to be split into their components.
       The output format for 1-best hypotheses is
            _s_e_n_t_i_d _w_1 _w_2 ...
       where _s_e_n_t_i_d is the sentence ID derived  from  the  N-best