Line: 1 to 1  

OpenGrm SFst Library: Stochastic FiniteState Transducer Library
 
Changed:  
< <  OpenGrm SFst Version 1.0.0 is now available for download.  
> >  OpenGrm SFst Version 1.1.0 is now available for download.  
SFst is a library for normalizing, sampling, combining, and approximating stochastic (or probabilistic) finitestate transducers. These are weighted finitestate transducers, represented in OpenFst library format, that have two properties: 
Line: 1 to 1  

OpenGrm SFst Library: Stochastic FiniteState Transducer Library  
Changed:  
< <  SFst is a library for normalizing, sampling, combining, and approximating stochastic (or probabilistic) finitestate transducers. These are weighted finitestate transducers, represented in OpenFst library format, that have the following two properties:
For example, an ngram model produced by the OpenGrm NGram Library is a stochastic FST^{2} but many other topologies are possible.  
> >  OpenGrm SFst Version 1.0.0 is now available for download. SFst is a library for normalizing, sampling, combining, and approximating stochastic (or probabilistic) finitestate transducers. These are weighted finitestate transducers, represented in OpenFst library format, that have two properties:
 
Added:  
> >  
Changed:  
< <  ^{1}Computation is done internally assuming the weights are negative log probabilities using Log64Weight. Conversion from the input weight type is done using a WeightConvert functor, predefined for common weight types like TropicalWeight and LogWeight .
^{2}Provided the phi_label is specified to match the backoff label, typically 0, of the ngram model.  
> >  ^{1}Provided the failure label ( phi_label ) is specified to match the backoff label, typically 0, of the ngram model.  

Line: 1 to 1  

OpenGrm SFst Library: Stochastic FiniteState Transducer Library  
Line: 13 to 13  
< * let state
 
Changed:  
< <  For example, an ngram model produced by the OpenGrm NGram Library is a stochastic FST (provided the phi_label is specified to match the backoff label, typically 0, of the ngram model), but many other topologies are possible.  
> >  For example, an ngram model produced by the OpenGrm NGram Library is a stochastic FST^{2} but many other topologies are possible.
^{1}Computation is done internally assuming the weights are negative log probabilities using Log64Weight. Conversion from the input weight type is done using a  
Deleted:  
< <  The following operations are provided for SFSTs:  
Deleted:  
< < 
 
Deleted:  
< <  ^{1}Computation is done internally with Log64Weight. Conversion from the input weight type is done using a WeightConvert functor, predefined for common weight types like TropicalWeight and LogWeight .
^{2}Possible when the sum of weight of all successful paths from the initial state is finite (and the input is trim). 
Line: 1 to 1  

OpenGrm SFst Library: Stochastic FiniteState Transducer Library  
Line: 11 to 11  
 
Changed:  
< < 
 
> > 
 
For example, an ngram model produced by the OpenGrm NGram Library is a stochastic FST (provided the phi_label is specified to match the backoff label, typically 0, of the ngram model), but many other topologies are possible.  
Line: 19 to 19  
 
Changed:  
< < 
 
> > 
 
 
Changed:  
< < 
 
> > 
 

Line: 1 to 1  

OpenGrm SFst Library: Stochastic FiniteState Transducer Library  
Line: 23 to 23  
 
Added:  
> > 
 
 
Deleted:  
< < 
 

Line: 1 to 1  

OpenGrm SFst Library: Stochastic FiniteState Transducer Library  
Line: 35 to 35  
 
Changed:  
< < 
 
> > 
 
^{1}Computation is done internally with Log64Weight. Conversion from the input weight type is done using a 
Line: 1 to 1  

OpenGrm SFst Library: Stochastic FiniteState Transducer Library  
Line: 19 to 19  
 
Changed:  
< < 
 
> > 
 
 
Changed:  
< < 
 
> > 
 
 
Line: 34 to 35  
 
Added:  
> > 
 
^{1}Computation is done internally with Log64Weight. Conversion from the input weight type is done using a WeightConvert functor, predefined for common weight types like TropicalWeight and LogWeight .
^{2}Possible when the sum of weight of all successful paths from the initial state is finite (and the input is trim). 
Line: 1 to 1  

OpenGrm SFst Library: Stochastic FiniteState Transducer LibrarySFst is a library for normalizing, sampling, combining, and approximating stochastic (or probabilistic) finitestate transducers. These are weighted finitestate transducers, represented in OpenFst library format, that have the following two properties:  
Changed:  
< < 
 
> > 
 
 
Deleted:  
< < 
 
For example, an ngram model produced by the OpenGrm NGram Library is a stochastic FST (provided the 
Line: 1 to 1  

OpenGrm SFst Library: Stochastic FiniteState Transducer Library  
Line: 9 to 9  
 
Changed:  
< < 
 
> > 
 
< * let state

Line: 1 to 1  

OpenGrm SFst Library: Stochastic FiniteState Transducer Library  
Line: 19 to 19  
The following operations are provided for SFSTs:
 
Added:  
> > 
 
 
Added:  
> > 
 
 
Changed:  
< < 
 
> > 
 
 
Changed:  
< < 
 
> > 
 
^{1}Computation is done internally with Log64Weight. Conversion from the input weight type is done using a  
Changed:  
< <  ^{2}Possible when the sum of all successful paths from the initial state is finite (and the input is trim).  
> >  ^{2}Possible when the sum of weight of all successful paths from the initial state is finite (and the input is trim). 
Line: 1 to 1  

OpenGrm SFst Library: Stochastic FiniteState Transducer Library  
Line: 14 to 14  
 
Changed:  
< <  For example, an ngram model produced by the OpenGrm NGram Library is a stochastic FST (provided the phi_label is specified to match the backoff label, typically 0, of the ngram model), but many other topologies are possible.  
> >  For example, an ngram model produced by the OpenGrm NGram Library is a stochastic FST (provided the phi_label is specified to match the backoff label, typically 0, of the ngram model), but many other topologies are possible.  
The following operations are provided for SFSTs:
 
Added:  
> > 
 

Line: 1 to 1  

 
Changed:  
< <  OpenGrm SFst Library: Stochastic FiniteState Transducer Library  
> >  OpenGrm SFst Library: Stochastic FiniteState Transducer Library  
SFst is a library for normalizing, sampling, combining, and approximating stochastic (or probabilistic) finitestate transducers. These are weighted finitestate transducers, represented in OpenFst library format, that have the following two properties: 
Line: 1 to 1  

OpenGrm SFst Library: Stochastic FiniteState Transducer Library  
Line: 27 to 27  
 
Changed:  
< < 
 
> > 
 
^{1}Computation is done internally with Log64Weight. Conversion from the input weight type is done using a 
Line: 1 to 1  

OpenGrm SFst Library: Stochastic FiniteState Transducer Library  
Line: 22 to 22  
 
Changed:  
< < 
 
> > 
 
 
Changed:  
< < 
 
> > 
 
^{1}Computation is done internally with Log64Weight. Conversion from the input weight type is done using a 
Line: 1 to 1  

OpenGrm SFst Library: Stochastic FiniteState Transducer Library  
Line: 12 to 12  
 
Added:  
> > 
 
For example, an ngram model produced by the OpenGrm NGram Library is a stochastic FST (provided the phi_label is specified to match the backoff label, typically 0, of the ngram model), but many other topologies are possible. 
Line: 1 to 1  

OpenGrm SFst Library: Stochastic FiniteState Transducer Library  
Line: 10 to 10  
 
Changed:  
< < 
 
> >  < * let state  
For example, an ngram model produced by the OpenGrm NGram Library is a stochastic FST (provided the 
Line: 1 to 1  

OpenGrm SFst Library: Stochastic FiniteState Transducer Library  
Changed:  
< <  SFst is a library for normalizing, sampling, combining, and approximating stochastic (or probabilistic) finitestate transducers. These are weighted finitestate transducers, represented in OpenFst library format, that have the following properties:  
> >  SFst is a library for normalizing, sampling, combining, and approximating stochastic (or probabilistic) finitestate transducers. These are weighted finitestate transducers, represented in OpenFst library format, that have the following two properties:  
Changed:  
< < 
 
> > 
 
For example, an ngram model produced by the OpenGrm NGram Library is a stochastic FST (provided the phi_label is specified to match the backoff label, typically 0, of the ngram model), but many other topologies are possible. 
Line: 1 to 1  

OpenGrm SFst Library: Stochastic FiniteState Transducer Library  
Line: 14 to 14  
 
Changed:  
< < 
 
> > 
 
 
Line: 22 to 22  
^{1}Computation is done internally with Log64Weight. Conversion from the input weight type is done using a  
Added:  
> >  ^{2}Possible when the sum of all successful paths from the initial state is finite (and the input is trim). 
Line: 1 to 1  

Added:  
> > 
OpenGrm SFst Library: Stochastic FiniteState Transducer LibrarySFst is a library for normalizing, sampling, combining, and approximating stochastic (or probabilistic) finitestate transducers. These are weighted finitestate transducers, represented in OpenFst library format, that have the following properties:
For example, an ngram model produced by the OpenGrm NGram Library is a stochastic FST (provided the The following operations are provided for SFSTs:
^{1}Computation is done internally with Log64Weight. Conversion from the input weight type is done using a 