How to install new language or technology model
This article explains how to add a new technology model into the current Speech Engine (SPE) instance. In Speech to Text (STT) and Keyword Spotting (KWS), it can be understood also as adding new language to STT or KWS.
Prerequisites
To proceed, you need to have existing installation of SPE. If you do not have one, check other articles, especially
Installation packages are custom created and provided to you by your assigned Phonexia contact. Version of installation package must match the version of your SPE. Your assigned Phonexia contact will ask which version of SPE you are running, but you can speed things up by providing the information already in your request.
How to check SPE version
Go to your SPE installation directory and locate phxadmin
(Linux) or phxadmin.exe
(Windows) binary. Then run following command:
On Linux run: ./phxadmin --version
On Windows run: phxadmin.exe /version
You should see the output similar to this:
Installation
In our example, we will install Spanish (ES_6
) model of Speech to Text and Keyword Spotting (with Phoneme Recognizer) into existing installation.
Installation of new language or technology model can be done in few steps:
- Unzip installation package to a temporary directory.
- Each installation package will contain
phxadmin
(Linux) orphxadmin.exe
(Windows) binary. Using the procedure above verifies that the package version matches your SPE version. If not or in doubt, contact your Phonexia contact. - Stop SPE gracefully by sending
Ctrl+C
to its terminal Window. Wait for it to finish. Based on number of running tasks, it can take some time. - When SPE is stopped, merge content of
bsapi
directory from your temporary folder to your SPE installation.
- In our example, we have installed Spanish model (
ES_6
technology model) of Speech to Text and Keyword Spotting (with Phoneme Recognizer). When you install new languages or models, they are turned off by default (green boxes on screenshot).
To turn the newly added models on, run following command (expectation: SPE is not running).On Windows run:
phxadmin.exe /configure-tech
On Linux run:./phxadmin --configure-tech
You can see that Spanish (ES_6
technology model) now appears as available model. You can enable it by selecting it and assigning number of instances to it. In the above example, 1 instance was configured. - Repeat the above steps for enabling also Phoneme Recognition technology and Speech to Text technology for Spanish
ES_6
model - Remember to save your settings.
- Now you can start SPE back again and use newly installed languages or technology models.
To start SPE, run following command:
On Windows run:phxspe.exe
On Linux run:./phxspe
- Optional: If you want to increase or decrease the number of instances of either technology or either model, you can edit the
SPE/settings/technologies.xml
file with notepad
- Important: in order for the changes to take effect, you need to restart the Speech Engine (
phxspe.exe
on Windows orphxspe
on Linux)
Final note: The binary phxadmin2
| phxadmin2.exe
is production level BETA. We recommend to use only phxadmin
| phxadmin.exe
for production.