Hi team,

In a document containing multiple source segments where only numerical values differ, automatic number replacement works incorrectly in some cases.

Here is how to reproduce the issue:

1. First, transalate the following English source to Chinses as below:

EN: LPCM (4ch monaural 48 kHz/24 bit) * 4
CN: LPCM（4声道单声道 48 kHz/24 位）* 4

2. Then, the following source is automatically translated as below as a 100% match:

EN: LPCM (4ch monaural 48 kHz/24 bit) * 3
中: LPCM（3声道单声道 48 kHz/24 位）* 4

As you see, the number in the part correponding to "4ch" is incorrectly replaced as "3声道" while the number in the part corresponding "* 3" is incorrectly replaced as "* 4".

However, if you first translate "LPCM (4ch monaural 48 kHz/24 bit) * 3" to "LPCM（4声道单声道 48 kHz/24 位）* 3", "LPCM (4ch monaural 48 kHz/24 bit) * 4" is correctly translated as "LPCM（4声道单声道 48 kHz/24 位）* 4" with the numbers properly replaced.

This inconsistency in the number replacement behaviors (misreplacement) could lead to serious mistranslation, so it would be great that the issue is resolved in future releases.