index fb9db694e28db886a1a150481c8025a7b94f0750..46390cdeef3fc19ddb9c547f8bca9d6a97f31db7 100644 (file)
This function implements either the MPFE (minimum phone frame error) or SMBR
(state-level minimum bayes risk) forward-backward, depending on whether
"criterion" is "mpfe" or "smbr". It returns the MPFE
- criterion of SMBR criterion for this file, and outputs the posteriors (which
- may be positive or negative) into "arc_post".
- Note: setting one_silence_class to false gives the old traditional behavior,
- true gives a possibly improved behavior which will tend to reduce insertions
- in the trained model.
+ criterion of SMBR criterion for this utterance, and outputs the posteriors (which
+ may be positive or negative) into "post".
+
+ @param [in] trans The transition model. Used to map the
+ transition-ids to phones or pdfs.
+ @param [in] silence_phones A list of integer ids of silence phones. The
+ silence frames i.e. the frames where num_ali
+ corresponds to a silence phones are treated specially.
+ The behavior is determined by 'one_silence_class'
+ being false (traditional behavior) or true.
+ Usually in our setup, several phones including
+ the silence, vocalized noise, non-spoken noise
+ and unk are treated as "silence phones"
+ @param [in] lat The denominator lattice
+ @param [in] num_ali The numerator alignment
+ @param [in] criterion The objective function. Must be "mpfe" or "smbr"
+ for MPFE (minimum phone frame error) or sMBR
+ (state minimum bayes risk) training.
+ @param [in] one_silence_class Determines how the silence frames are treated.
+ Setting this to false gives the old traditional behavior,
+ where the silence frames (according to num_ali) are
+ treated as incorrect. However, this means that the
+ insertions are not penalized by the objective.
+ Setting this to true gives the new behaviour, where we
+ treat silence as any other phone, except that all pdfs
+ of silence phones are collapsed into a single class for
+ the frame-error computation. This can possible reduce
+ the insertions in the trained model. This is closer to
+ the WER metric that we actually care about, since WER is
+ generally computed after filtering out noises, but
+ does penalize insertions.
+ @param [out] post The "MBR posteriors" i.e. derivatives w.r.t to the
+ pseudo log-likelihoods of states at each frame.
*/
BaseFloat LatticeForwardBackwardMpeVariants(
const TransitionModel &trans,
used in our normal MMI training recipes, where it's instead done using various command
line programs that each do a part of the job. This function was written for use in
neural-net MMI training.
- If drop_frames is true, it will not compute any posteriors on frames where the num and
- den have disjoint pdf-ids.
- If "convert_to_pdf_ids" is true, it will convert the output to be at the level of pdf-ids,
- not transition-ids.
- If "cancel" is true, it will cancel out any positive and negative parts from
- the same transition-id (or pdf-id, if convert_to_pdf_ids == true).
+
+ @param [in] trans The transition model. Used to map the
+ transition-ids to phones or pdfs.
+ @param [in] lat The denominator lattice
+ @param [in] num_ali The numerator alignment
+ @param [in] drop_frames If "drop_frames" is true, it will not compute any
+ posteriors on frames where the num and den have disjoint
+ pdf-ids.
+ @param [in] convert_to_pdf_ids If "convert_to_pdfs_ids" is true, it will
+ convert the output to be at the level of pdf-ids, not
+ transition-ids.
+ @param [in] cancel If "cancel" is true, it will cancel out any positive and
+ negative parts from the same transition-id (or pdf-id,
+ if convert_to_pdf_ids == true).
+ @param [out] arc_post The output MMI posteriors of transition-ids (or
+ pdf-ids if convert_to_pdf_ids == true) at each frame
+ i.e. the difference between the numerator
+ and denominator posteriors.
+
It returns the forward-backward likelihood of the lattice. */
BaseFloat LatticeForwardBackwardMmi(
const TransitionModel &trans,