The limitation of target length in _break_query

[loelee307]

Hi,

I'm trying to use text search to generate segments from my audio-text resources, based on the recipe of libriheavy. But occasionally I found some cases with incomplete text to cover corresponding audio contents.

I have identified that the issue is likely in the matching._break_query function. From the codes, it appears that when the query is not split, the target length seems to be restricted to less than or equal to query_end - query_start. Is it designed for lower the complexity of levenshtein alignment? What would you advise me to solve these cases?

Thanks so much for your time.