12.12.2012 Views

Festival Speech Synthesis System: - Speech Resource Pages

Festival Speech Synthesis System: - Speech Resource Pages

Festival Speech Synthesis System: - Speech Resource Pages

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

(define (pos_cand_function w)<br />

;; select the appropriate lexicon<br />

(lex.select 'pos_lex)<br />

;; return the list of cands with rprobs<br />

(cadr<br />

(lex.lookup (item.name w) nil)))<br />

The returned candidate list would look somthing like<br />

( (jj -9.872) (vbd -6.284) (vbn -5.565) )<br />

Our part of speech tagger function would look something like this<br />

(define (pos_tagger utt)<br />

(set! get_vit_params<br />

(list<br />

(list 'Relation "Word")<br />

(list 'return_feat 'pos_tag)<br />

(list 'p_word "punc")<br />

(list 'pp_word "nn")<br />

(list 'ngramname "pos-tri-gram")<br />

(list 'cand_function 'pos_cand_function)))<br />

(Gen_Viterbi utt)<br />

utt)<br />

this will assign the optimal part of speech tags to each word in utt.<br />

[ < ] [ > ] [ > ] [Top] [Contents] [Index] [ ? ]<br />

25.5 Linear regression<br />

The linear regression model takes models built from some external package and finds coefficients based on the<br />

features and weights. A model consists of a list of features. The first should be the atom Intercept plus a value.<br />

The following in the list should consist of a feature (see section 14.6 Features) followed by a weight. An optional<br />

third element may be a list of atomic values. If the result of the feature is a member of this list the feature's value is<br />

treated as 1 else it is 0. This third argument allows an efficient way to map categorical values into numeric values.<br />

For example, from the F0 prediction model in `lib/f2bf0lr.scm'. The first few parameters are<br />

(set! f2b_f0_lr_start<br />

'(<br />

( Intercept 160.584956 )<br />

( Word.Token.EMPH 36.0 )<br />

( pp.tobi_accent 10.081770 (H*) )<br />

( pp.tobi_accent 3.358613 (!H*) )<br />

( pp.tobi_accent 4.144342 (*? X*? H*!H* * L+H* L+!H*) )<br />

( pp.tobi_accent -1.111794 (L*) )<br />

...<br />

)<br />

Note the feature pp.tobi_accent returns an atom, and is hence tested with the map groups specified as third<br />

arguments.<br />

Models may be built from feature data (in the same format as `wagon' using the `ols' program distributed with<br />

the speech tools library.<br />

[ < ] [ > ] [ > ] [Top] [Contents] [Index] [ ? ]

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!