mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-12-11 06:55:27 +00:00
Code Issues Packages Projects Releases Wiki Activity

Support combining two LMs

Browse Source
This commit is contained in:
Fangjun Kuang 2023-04-10 12:13:36 +08:00
parent 482efb8020
commit c3c6d4c3c4
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
icefall/shared/combine_lm.py
View File

@ -7,11 +7,11 @@
Given two LMs "A" and "B", this script modifies
probabilities in A such that
P_{A_{new}}(w_n|w0,w1,...,w_{n-1}) = P_{A_{original}}(w_n|w0,w1,...,w_{n-1}) / P_{B}(w_n|w0,w1,...,w_{n-1})
P_{A_{new}}(w_n|w_0,w_1,...,w_{n-1}) = P_{A_{original}}(w_n|w_0,w_1,...,w_{n-1}) / P_{B}(w_n|w_0,w_1,...,w_{n-1})
When it is formulated in log-space, it becomes
\log P_{A_{new}}(w_n|w0,w1,...,w_{n-1}) = \log P_{A_{original}}(w_n|w0,w1,...,w_{n-1}) - \log P_{B}(w_n|w0,w1,...,w_{n-1})
\log P_{A_{new}}(w_n|w_0,w_1,...,w_{n-1}) = \log P_{A_{original}}(w_n|w_0,w_1,...,w_{n-1}) - \log P_{B}(w_n|w_0,w_1,...,w_{n-1})
Optionally, you can pass a scale for the LM "B", such that