lattice-tool.html

这是一款很好用的工具包

<DD>
<DT><B>-mix-lambda4</B><I> weight</I><B></B>
<DD>
<DT><B>-mix-lambda5</B><I> weight</I><B></B>
<DD>
<DT><B>-mix-lambda6</B><I> weight</I><B></B>
<DD>
<DT><B>-mix-lambda7</B><I> weight</I><B></B>
<DD>
<DT><B>-mix-lambda8</B><I> weight</I><B></B>
<DD>
<DT><B>-mix-lambda9</B><I> weight</I><B></B>
<DD>
These are the weights for the additional mixture components, corresponding
to
<B> -mix-lm2 </B>
through
<B>-mix-lm9</B>.<B></B>
The weight for the
<B> -mix-lm </B>
model is 1 minus the sum of 
<B> -lambda </B>
and 
<B> -mix-lambda2 </B>
through
<B>-mix-lambda9</B>.<B></B>
<DT><B>-bayes</B><I> length</I><B></B>
<DD>
Set the context length used for Bayesian interpolation.
The default value is 0, giving the standard fixed interpolation weight
specified by
<B>-lambda</B>.<B></B>
<DT><B>-bayes-scale</B><I> scale</I><B></B>
<DD>
Set the exponential scale factor on the context likelihood in conjunction
with the
<B> -bayes </B>
function.
Default value is 1.0.
<DT><B>-compact-expansion</B><I></I><B></B>
<DD>
Use a compact expansion algorithm that uses backoff nodes to reduce the 
size of expanded lattices (see paper reference below).
<DT><B>-old-expansion</B><I></I><B></B>
<DD>
Use older versions of the lattice expansion algorithms (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.
<DT><B>-max-nodes</B><I> M</I><B></B>
<DD>
Abort lattices expansion when the number of nodes (including null and pause
nodes) exceeds 
<I>M</I>.<I></I>
This is another mechanism to avoid spending too much time on very large
lattices.
</DD>
</DL>
<H2> LATTICE EXPANSION ALGORITHMS </H2>
<B> lattice-tool </B>
incorporates several different algorithms to apply LM weights to
lattices.
This section explains what algorithms are applied given what options.
<DL>
<DT><B> Compact LM expansion </B>
<DD>
This expands the nodes and transitions to be able to assign
higher-order probabilities to transitions.
Backoffs in the LM are exploited in the expansion, thereby 
minimizing the number of added nodes (Weng et al., 1998).
This algorithm is triggered by
<B>-compact-expansion</B><B></B>
For the resulting lattices 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
<B> ngram -prune-lowprobs </B>
option.)
Pauses and null nodes are handled during the expansion and do
not have to be removed and restored.
<DT><B> General LM expansion </B>
<DD>
This expands the lattice to apply LMs of arbitrary order, 
without use of backoff transitions.
This algorithm is the default (no
<B>-compact-expansion</B>).<B></B>
<DT><B> Unigram weight replacement </B>
<DD>
This simply replaces the weights on lattice transitions with 
unigram log probabilities.
No modification of the lattice structure is required.
This algorithm is used if 
<B> -old-expansion </B>
and
<B> -order 1 </B>
are specified.
<DT><B> Bigram weight replacement </B>
<DD>
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 
<B> -old-expansion </B>
and
<B> -order 2 </B>
are specified.
</DD>
</DL>
<H2> HTK LATTICES </H2>
<P>
<B> lattice-tool </B>
can optionally read, process, and output lattices in 
HTK Standard Lattice Format.
The following options control HTK lattice processing.
<DL>
<DT><B> -read-htk </B>
<DD>
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.
Transitions 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.
<DT><B>-htk-acscale</B><I> S</I><B></B>
<DD>
<DT><B>-htk-lmscale</B><I> S</I><B></B>
<DD>
<DT><B>-htk-ngscale</B><I> S</I><B></B>
<DD>
<DT><B>-htk-prscale</B><I> S</I><B></B>
<DD>
<DT><B>-htk-duscale</B><I> S</I><B></B>
<DD>
<DT><B>-htk-x1scale</B><I> S</I><B></B>
<DD>
<DT><B>-htk-x2scale</B><I> S</I><B></B>
<DD>
...
<DT><B>-htk-x9scale</B><I> S</I><B></B>
<DD>
<DT><B>-htk-wdpenalty</B><I> S</I><B></B>
<DD>
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 specifications in the lattice files
themselves.
<BR>
Word transition penalties are scaled according to the log base used.
Values specified on the command line are scaled according to 
<B>-htk-logbase</B>,<B></B>
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 
<I>e</I>.<I></I>
<DT><B> -no-htk-nulls </B>
<DD>
Eliminate NULL nodes otherwise created by the conversion of HTK lattices
to PFSGs.
This creates additional links and may or may not reduce the overall
processing time required.
<DT><B>-dictionary</B><I> file</I><B></B>
<DD>
Read a dictionary containing pronunciation probabilities from 
<I>file</I>,<I></I>
and add or replace the pronunciation scores in the lattice accordingly.
This requires that the lattices contain phone alignment information.
<DT><B> -intlogs </B>
<DD>
Assume the dictionary contains log probabilities encoded on the int-log scale,
as used by the SRI Decipher system.
<DT><B> -write-htk </B>
<DD>
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 
<B> -compute-posteriors </B>
was used in lattice processing the output lattices will also contain
node posterior probabilities.
If the input lattices were in HTK format, then
acoustic and duration scores are preserved from the input lattices.
The score scaling factors in the lattice header will reflect the 
<B> -htk-*scale </B>
options given above.
<DT><B>-htk-logbase</B><I> B</I><B></B>
<DD>
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
<I> e </I>
according to the HTK SLF definition.
<DT><B> -htk-words-on-nodes </B>
<DD>
Output word labels and other word-related information 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.
<DT><B> -htk-scores-on-nodes </B>
<DD>
Output scores on HTK lattice nodes, rather than links.
<DT>Note:
<DD>
The options
<B>-no-htk-nulls</B>,<B></B>
<B>-htk-words-on-nodes</B>,<B></B>
and
<B>-htk-scores-on-nodes</B><B></B>
defeat the mapping of internal PFSG nodes back to HTK transitions, and should
therefore NOT be used when a compact output representation is desired.
<DT><B> -htk-quotes </B>
<DD>
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 lattices.
</DD>
</DL>
<H2> N-BEST DECODING </H2>
The option
<B> -nbest-decode </B>
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 
<A HREF="nbest-lattice.html">nbest-lattice(1)</A>
and 
<A HREF="nbest-scripts.html">nbest-scripts(1)</A>.
The following options control the location of output files:
<DL>
<DT><B>-out-nbest-dir</B><I> dir</I><B></B>
<DD>
The directory to which N-best list files are written.
These contain acoustic model scores, language model scores,
word counts, and the word hypotheses themselves,
in SRILM format as described in
<A HREF="nbest-format.html">nbest-format(5)</A>.
<DT><B>-out-nbest-dir-ngram</B><I> dir</I><B></B>
<DD>
Output directory for separate N-gram LM scores as may be encoded in 
HTK lattices.
<DT><B>-out-nbest-dir-pron</B><I> dir</I><B></B>
<DD>
Output directory for pronunciation scores encoded in HTK lattices.
<DT><B>-out-nbest-dir-dur</B><I> dir</I><B></B>
<DD>
Output directory for duration model scores encoded in HTK lattices.
<DT><B>-out-nbest-dir-xscore1</B><I> dir</I><B></B>
<DD>
<DT><B>-out-nbest-dir-xscore2</B><I> dir</I><B></B>
<DD>
...
<DT><B>-out-nbest-dir-xscore9</B><I> dir</I><B></B>
<DD>
Output score directories for up to nine additional knowledge sources
encoded in HTK lattices.
<DT><B>-out-nbest-dir-rttm</B><I> dir</I><B></B>
<DD>
N-best hypotheses in NIST RTTM format.
This function is experimental and makes assumptions about the input 
file naming conventions to infer timing information.
</DD>
</DL>
<H2> SEE ALSO </H2>
<A HREF="ngram.html">ngram(1)</A>, <A HREF="pfsg-scripts.html">pfsg-scripts(1)</A>, <A HREF="nbest-lattice.html">nbest-lattice(1)</A>,
<A HREF="pfsg-format.html">pfsg-format(5)</A>, <A HREF="ngram-format.html">ngram-format(5)</A>, <A HREF="classes-format.html">classes-format(5)</A>, <A HREF="wlat-format.html">wlat-format(5)</A>,
<A HREF="nbest-format.html">nbest-format(5)</A>.
<BR>
F. Weng, A. Stolcke, and A. Sankar,
``Efficient Lattice Representation and Generation.''
<I>Proc. Intl. Conf. on Spoken Language Processing</I>, vol. 6, pp. 2531-2534,
Sydney, 1998.
<BR>
S. Young et al., <I>The HTK Book</I>, HTK version 3.1.
http://htk.eng.cam.ac.uk/prot-docs/htk_book.shtml
<H2> BUGS </H2>
Not all LM types supported by 
<A HREF="ngram.html">ngram(1)</A>
are handled by 
<B> lattice-tool. </B>
<P>
Care must be taken when processing multiword lattices with 
<B> -unk </B>
and 
<B> -multiwords </B>
or 
<B>-split-multiwords</B>.<B></B>
Multiwords not listed in the LM (or the explicit vocabulary specified) will
be considered ``unknown'', even though their components might be 
in-vocabulary.
<P>
The 
<B> -nbest-duplicates </B>
option does not work together with
<B>-nbest-viterbi</B>.<B></B>
<P>
Input lattice in HTK format may contain node or link posterior information.
However, this information is effectively discarded; posteriors are always
recomputed from scores when needed for pruning or output.
<P>
The 
<B>-no-nulls</B>,<B></B>
<B> -no-pause </B>
and
<B> -compact-pause </B>
options discard 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.
<H2> AUTHORS </H2>
Fuliang Weng &lt;[e-mail]&gt;
<BR>
Andreas Stolcke &lt;[e-mail]&gt;
<BR>
Dustin Hillard &lt;[e-mail]&gt;
<BR>
Jing Zheng &lt;[e-mail]&gt;
<BR>
Copyright 1997-2006 SRI International
</BODY>
</HTML>