Moses Documentation (Statistical Machine Translation) The mose.ini file format?

Is there any documentation of the moses.ini format for Moses? Running moses on the command line with no arguments returns the available function names, but not their available arguments. Also, the structure of the .ini file is not listed in the manual I see.

+3


source to share


1 answer


The basic idea is that the file contains settings that will be used by the translation model. As such, documentation on values ​​and parameters in moses.ini

should be sought in the Moses function specs.

Here are some excerpts I found on the Internet about moses.ini

.

In Moses Core , we have a few details:

7.6.5 moses.ini


All function functions are indicated in the section [feature]

. It must be in the format:
* Feature-name key1=value1

key2=value2

....
For example,KENLM factor=0 order=3 num-features=1 lazyken=0 path=file.lm.gz


There is also a hint on how to print basic statistics about all the components mentioned in the moses.ini file.



Run the script
analyse_moses_model.pl moses.ini


This can be useful for setting the order of the display steps to avoid hacking translation options, or just to check that the model components are as large and detailed as we expect.

In the Center for Computational Language and Learning (CLEAR) Wiki, there is an example file with some documentation:

Parameters

It is recommended to keep a file .ini

to store all your settings.

input-factors


- Use factor model or not
mapping


- Use LM in memory (T) or directly read file on hard disk (G)
ttable-file


- Specify a number. source-factor, target factor, score and path to the translation table file
lmodel-file


- Specify the type used for LM (0: SRILM, 1: IRSTLM) using factor number, order (n-gram) LM and path to the language model file

If this is not enough, the page has a different description, see the section "Decoder configuration file"

Sections [ttable-file]

and [lmodel-file]

contain pointers to the phrase table file and the language model file, respectively. You can ignore the numbers on these lines. For now, it is enough to know that the last number in the language model specification is the order of the n-gram model .

The configuration file also contains some weights. Note that the section [weight-t]

has 5 weights, one for each function in the phrase table.

The file moses.ini

generated by the tutorial will not work with your decoder without modification , because it relies on a language that does not compile into our decoder. In order for it to work, open the moses.ini file and find the language model in the line immediately after the header [lmodel-file]

. The first number on this line will be 0

, which stands for SRILM. Change it to 8

and leave the rest of the line untouched. Then your config should work.

+1


source







All Articles