lattice-tool.1

这是一款很好用的工具包

       --mmiixx--llaammbbddaa99 _w_e_i_g_h_t
              These  are  the  weights for the additional mixture
              components, corresponding to --mmiixx--llmm22 through --mmiixx--
              llmm99.   The  weight for the --mmiixx--llmm model is 1 minus
              the sum of --llaammbbddaa and --mmiixx--llaammbbddaa22  through  --mmiixx--
              llaammbbddaa99.

       --bbaayyeess _l_e_n_g_t_h
              Set the context length used for Bayesian interpola-
              tion.  The default value is 0, giving the  standard
              fixed interpolation weight specified by --llaammbbddaa.

       --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.

       --ccoommppaacctt--eexxppaannssiioonn
              Use a compact expansion algorithm that uses backoff
              nodes to reduce the size of expanded lattices  (see
              paper reference below).

       --oolldd--eexxppaannssiioonn
              Use  older  versions of the lattice expansion algo-
              rithms (both regular and compact), that handle only
              trigram  models and require elimination of null and
              pause nodes prior to expansion.   Not  recommended,
              but   useful  if  full  backward  compatibility  is
              required.

       --mmaaxx--nnooddeess _M
              Abort lattices expansion when the number  of  nodes
              (including  null  and pause nodes) exceeds _M.  This
              is another mechanism to  avoid  spending  too  much
              time on very large lattices.

LLAATTTTIICCEE EEXXPPAANNSSIIOONN AALLGGOORRIITTHHMMSS
       llaattttiiccee--ttooooll  incorporates several different algorithms to
       apply LM weights to lattices.  This section explains  what
       algorithms are applied given what options.

       CCoommppaacctt LLMM eexxppaannssiioonn
              This  expands  the nodes and transitions to be able
              to assign  higher-order  probabilities  to  transi-
              tions.   Backoffs  in  the  LM are exploited in the
              expansion, thereby minimizing the number  of  added
              nodes (Weng et al., 1998).  This algorithm is trig-
              gered by --ccoommppaacctt--eexxppaannssiioonn For the resulting  lat-
              tices  to  work  correctly, backoff paths in the LM
              must  have  lower  weight  than  the  corresponding
              higher-order  paths.   (For N-gram LMs, this can be
              achieved using the nnggrraamm  --pprruunnee--lloowwpprroobbss  option.)
              Pauses and null nodes are handled during the expan-
              sion and do not have to be removed and restored.

       GGeenneerraall LLMM eexxppaannssiioonn
              This expands the lattice to apply LMs of  arbitrary
              order,  without  use  of backoff transitions.  This
              algorithm is the default (no --ccoommppaacctt--eexxppaannssiioonn).

       UUnniiggrraamm wweeiigghhtt rreeppllaacceemmeenntt
              This simply replaces the weights on lattice transi-
              tions with unigram log probabilities.  No modifica-
              tion of the lattice structure  is  required.   This
              algorithm  is  used  if --oolldd--eexxppaannssiioonn and --oorrddeerr 11
              are specified.

       BBiiggrraamm wweeiigghhtt rreeppllaacceemmeenntt
              This replaces the transition  weights  with  bigram
              log probabilities.  Pause and null nodes have to be
              eliminated prior to the operation, and are restored
              after  weight  replacement.  This algorithm is used
              if --oolldd--eexxppaannssiioonn and --oorrddeerr 22 are specified.

HHTTKK LLAATTTTIICCEESS
       llaattttiiccee--ttooooll can optionally read, process, and output lat-
       tices  in  HTK  Standard  Lattice  Format.   The following
       options control HTK lattice processing.

       --rreeaadd--hhttkk
              Read input lattices in HTK  format.   All  lattices
              are  internally  represented  as  PFSGs; to achieve
              this HTK lattices links are mapped  to  PFSG  nodes
              (with attached word and score information), and HTK
              lattice nodes are mapped to PFSG NULL nodes.  Tran-
              sitions  are  created  so  as to preserve words and
              scores of all paths through the  original  lattice.
              On  output,  this  mapping  is  reversed,  so as to
              create a compact encoding of PFSGs containing  NULL
              nodes as HTK lattices.

       --hhttkk--aaccssccaallee _S

       --hhttkk--llmmssccaallee _S

       --hhttkk--nnggssccaallee _S

       --hhttkk--pprrssccaallee _S

       --hhttkk--dduussccaallee _S

       --hhttkk--xx11ssccaallee _S

       --hhttkk--xx22ssccaallee _S
              ...

       --hhttkk--xx99ssccaallee _S

       --hhttkk--wwddppeennaallttyy _S
              These options specify the weights for acoustic, LM,
              N-gram, pronunciation, and duration models,  up  to
              nine  extra  scores,  as  well  as  word transition
              penalties to be  used  for  combining  the  various
              scores  contained  in  HTK  lattices.  The combined
              scores are then  used  to  compute  the  transition
              weights   for  the  internal  PFSG  representation.
              Default weights are obtained  from  the  specifica-
              tions in the lattice files themselves.
              Word  transition  penalties are scaled according to
              the log base used.  Values specified on the command
              line  are  scaled according to --hhttkk--llooggbbaassee, or the
              default 10.  Word transition penalties specified in
              the  lattice  file  are scaled according to the log
              base specified in the file, or the default _e.

       --nnoo--hhttkk--nnuullllss
              Eliminate NULL nodes otherwise created by the  con-
              version  of  HTK  lattices  to PFSGs.  This creates
              additional links and may  or  may  not  reduce  the
              overall processing time required.

       --ddiiccttiioonnaarryy _f_i_l_e
              Read  a  dictionary containing pronunciation proba-
              bilities from _f_i_l_e, and add or replace the  pronun-
              ciation  scores  in  the lattice accordingly.  This
              requires that the lattices contain phone  alignment
              information.

       --iinnttllooggss
              Assume  the  dictionary  contains log probabilities
              encoded on the int-log scale, as used  by  the  SRI
              Decipher system.

       --wwrriittee--hhttkk
              Write  output lattices in HTK format.  If the input
              lattices were in  PFSG  format  the  original  PFSG
              weights  will  be  output  as  HTK acoustic scores.
              However, LM rescoring  will  discard  the  original
              PFSG  weights and the results will be encoded as LM
              scores.   Pronunciation  scoring  results  will  be
              encoded as pronunciations scores.  If the --ccoommppuuttee--
              ppoosstteerriioorrss was used in lattice processing the  out-
              put lattices will also contain node posterior prob-
              abilities.  If  the  input  lattices  were  in  HTK
              format,  then acoustic and duration scores are pre-
              served from the input lattices.  The score  scaling
              factors  in  the  lattice  header  will reflect the
              --hhttkk--**ssccaallee options given above.

       --hhttkk--llooggbbaassee _B
              Modify the logarithm base in HTK  lattices  output.
              The default is to use logs base 10, as elsewhere in
              SRILM.  As value of 0 means to output probabilities
              instead  of  log  probabilities.  Note that the log
              base for input lattices is  not  affected  by  this
              option;  it  is encoded in the lattices themselves,
              and defaults to _e according to the HTK SLF  defini-
              tion.

       --hhttkk--wwoorrddss--oonn--nnooddeess
              Output  word labels and other word-related informa-
              tion on HTK lattice nodes, rather than links,  thus
              saving  space.   This  option  is provided only for
              compatibility  with  software  that  requires  word
              information to be attached specifically to links of
              nodes.

       --hhttkk--ssccoorreess--oonn--nnooddeess
              Output scores on HTK  lattice  nodes,  rather  than
              links.

       Note:  The options --nnoo--hhttkk--nnuullllss, --hhttkk--wwoorrddss--oonn--nnooddeess, and
              --hhttkk--ssccoorreess--oonn--nnooddeess defeat the mapping of internal
              PFSG  nodes  back  to  HTK  transitions, and should
              therefore NOT be used when a compact output  repre-
              sentation is desired.

       --hhttkk--qquuootteess
              Enable the HTK string quoting mechanism that allows
              whitespace and other non-printable characters to be
              included in words labels and other fields.  This is
              disabled by default since PFSG lattices  and  other
              SRILM  tools  don't  support  such word labels.  It
              affects both input and output format for  HTK  lat-
              tices.

NN--BBEESSTT DDEECCOODDIINNGG
       The  option  --nnbbeesstt--ddeeccooddee  triggers  generation of N-best
       lists, according to the aggregate score of  paths  encoded
       in  the  lattice.   The output format for N-best lists and
       associated additional score files is compatible with other
       SRILM  tools  that  process  N-best  lists,  such as those
       described in nnbbeesstt--llaattttiiccee(1) and  nnbbeesstt--ssccrriippttss(1).   The
       following options control the location of output files:

       --oouutt--nnbbeesstt--ddiirr _d_i_r
              The  directory to which N-best list files are writ-
              ten.  These contain acoustic model scores, language
              model  scores, word counts, and the word hypotheses
              themselves, in SRILM format as described in  nnbbeesstt--
              ffoorrmmaatt(5).

       --oouutt--nnbbeesstt--ddiirr--nnggrraamm _d_i_r
              Output  directory  for separate N-gram LM scores as
              may be encoded in HTK lattices.

       --oouutt--nnbbeesstt--ddiirr--pprroonn _d_i_r
              Output directory for pronunciation  scores  encoded
              in HTK lattices.

       --oouutt--nnbbeesstt--ddiirr--dduurr _d_i_r
              Output  directory for duration model scores encoded
              in HTK lattices.

       --oouutt--nnbbeesstt--ddiirr--xxssccoorree11 _d_i_r

       --oouutt--nnbbeesstt--ddiirr--xxssccoorree22 _d_i_r
              ...

       --oouutt--nnbbeesstt--ddiirr--xxssccoorree99 _d_i_r
              Output score directories for up to nine  additional
              knowledge sources encoded in HTK lattices.

       --oouutt--nnbbeesstt--ddiirr--rrttttmm _d_i_r
              N-best  hypotheses in NIST RTTM format.  This func-
              tion is experimental and  makes  assumptions  about
              the  input  file naming conventions to infer timing
              information.

SSEEEE AALLSSOO
       ngram(1),  pfsg-scripts(1),  nbest-lattice(1),   pfsg-for-
       mat(5),   ngram-format(5),   classes-format(5),  wlat-for-
       mat(5), nbest-format(5).
       F. Weng, A. Stolcke, and A.  Sankar,  ``Efficient  Lattice
       Representation  and  Generation.''   _P_r_o_c_.  _I_n_t_l_. _C_o_n_f_. _o_n
       _S_p_o_k_e_n _L_a_n_g_u_a_g_e _P_r_o_c_e_s_s_i_n_g, vol. 6, pp. 2531-2534, Sydney,
       1998.
       S.   Young   et  al.,  _T_h_e  _H_T_K  _B_o_o_k,  HTK  version  3.1.
       http://htk.eng.cam.ac.uk/prot-docs/htk_book.shtml

BBUUGGSS
       Not all LM types supported by nnggrraamm(1) are handled by llaatt--
       ttiiccee--ttooooll..

       Care must be taken when processing multiword lattices with
       --uunnkk and --mmuullttiiwwoorrddss or --sspplliitt--mmuullttiiwwoorrddss.  Multiwords not
       listed  in  the  LM (or the explicit vocabulary specified)
       will be considered ``unknown'', even though  their  compo-
       nents might be in-vocabulary.

       The  --nnbbeesstt--dduupplliiccaatteess  option does not work together with
       --nnbbeesstt--vviitteerrbbii.

       Input lattice in HTK format may contain node or link  pos-
       terior  information.   However, this information is effec-
       tively discarded; posteriors are  always  recomputed  from
       scores when needed for pruning or output.

       The  --nnoo--nnuullllss,  --nnoo--ppaauussee and --ccoommppaacctt--ppaauussee options dis-
       card the acoustic information  associated  with  NULL  and
       pause nodes in HTK lattice input, and should therefore not
       be used if equivalent HTK lattice output is intended.

AAUUTTHHOORRSS
       Fuliang Weng <[e-mail]>
       Andreas Stolcke <[e-mail]>
       Dustin Hillard <[e-mail]>
       Jing Zheng <[e-mail]>
       Copyright 1997-2006 SRI International



SRILM Tools        $Date: 2006/09/20 21:05:57 $   lattice-tool(1)