Home

A tool for the estimation, representation, and computation of statistical language models.

Project Members:

• Christian Hardmeier
• Fabio Brugnara
• Giulio Paci
• Hieu Hoang
• Josh Schroeder
• Marcello Federico (admin)
• Mauro Cettolo
• Nick Ruiz
• Nicola Bertoldi (admin)
• Ondrej Bojar
• Roberto Gretter
• arianna b
• falavigna daniele

[Eric Riebling]

The code refers to http://hlt.fbk.eu/en/irstlm for the User Manual. But this page is not found. Other pages on that site point to this SourceForge project (the project wiki) for the User Manual or documentation, but that does not exist, either. In fact, you are told to go back to hlt.fbk.eu/en/irstlm, forming a circle. Where can a new user find the User Manual?

EDIT: two things seem to have happened: Sourceforge changed their URL formatting so that old links referring to 'mediawiki' no longer work. And the site at hlt.fbk.eu seems to have also dropped the page for /en/irstlm. I can actually find documentation if I compile the .tex formatted documentation in the source. It would just be more convenient to try and find it in HTML format online.

Web searches for strings within the .tex documentation actually link to Sourceforge 'mediawiki' based URLs that no longer work, but used to,
so the changes seem to be mostly the fault of Sourceforge.

[y91]

Hi,

I need to use IRSTLM toolkit to interpolate language models. Following is my requirement. I want to specify interpolation weights for individual LMs and generate a combined LM in ARPA format.

The tool kit provides 'interpolate-lm' script for "estimating new interpolation weights for individual LMs and also calculating the perplexity on a development text". However, it does not seem to output any LM that has resulted after the combination/interpolation process.

Can anyone point me to any program/script for outputing the combined LM in ARPA format?

Thanks
Yash

[Marcello Federico]

Hi
Unfortunately it is impossible to obtain one Arpa LM by interpolating N Arpa LMs. Another option is to use Irstlm's tlm command. We are now facing a problem with the on line documentation as source forge has discontinued the wiki site.

Tell me if you need more help,

Marcello

---

Short from my mobile phone

On 07/set/2014, at 02:48 AM, "Yash" y91@users.sf.net wrote:

Hi,

I need to use IRSTLM toolkit to interpolate language models. Following is my requirement. I want to specify interpolation weights for individual LMs and generate a combined LM in ARPA format.

The tool kit provides 'interpolate-lm' script for "estimating new interpolation weights for individual LMs and also calculating the perplexity on a development text". However, it does not seem to output any LM that has resulted after the combination/interpolation process.

Can anyone point me to any program/script for outputing the combined LM in ARPA format?

Thanks
Yash
Sent from sourceforge.net because you indicated interest in https://sourceforge.net/p/irstlm/wiki/Home/

To unsubscribe from further messages, please visit https://sourceforge.net/auth/subscriptions/

[y91]

Hi Marcello,

Thanks for the above response and info.

I saw the TLM command. It seems to provide functionality for estimating a LM based on training data and some adaptation data.

The task I want to accomplish is adapting a given language model (A.lm) in light of some new data (new-data.txt). Is this task feasible with the IRSTLM toolkit given that I do not have the training text (A-training-data.txt) from which A.lm was estimated??

Thanks!
Yash