Calculate Perplexity

Perplexity is a metric used in the language modeling. It indicates the average branching factor for a typical language.

To measure perplexity, you need:

To measure perplexity of a language model

  1. Normalize the sample text (see Run Text Normalization).
  2. Send an AddTask action to HPE IDOL Speech Server, and set the following parameters:

    Type The task name. Set to LmPerplexity.
    Tlm The base language model to analyze.
    Text The sample text file.
    Out The log file to generate.

For example:

http://localhost:13000/action=AddTask&Type=LmPerplexity&Text=C:\data\transcript.txt&Tlm=C:\LP\ENUK\ver-ENUK-5.0.tlm&Out=PerpScore.ctm

This action uses port 13000 to instruct HPE IDOL Speech Server, which is located on the local machine, to calculate the perplexity of the ver-ENUK-5.0.tlm language model using the transcript.txt sample text, and to write the results to the PerpScore.ctm file.

This action returns a token. You can use the token to:


_HP_HTML5_bannerTitle.htm