speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

识别结果中有英语时，缺少空格或者第一个单词的问题 (#2284)

Browse Source 
* Update ct-transformer-online.cpp

修复最后两个单词之间没有空格的问题

* Update ct-transformer-online.cpp

解决语音中连续两句英语，offline结果丢失第二句第一个单词的情况。
This commit is contained in:
Haitao 2024-12-13 13:47:15 +08:00 committed by GitHub
parent 2139ef696b
commit 7263fb08e9
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 6 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
runtime/onnxruntime/src/ct-transformer-online.cpp
View File

@ -42,6 +42,11 @@ string CTTransformerOnline::AddPunc(const char* sz_input, vector<string> &arr_ca
vector<int> InputData;
string strText; //full_text
strText = accumulate(arr_cache.begin(), arr_cache.end(), strText);
// 如果上一句的结尾是英语字母，并且这一句的开始也是英语字母，应该添加空格
if ((strText.size() > 0 and !(strText[strText.size()-1] & 0x80)) && (strlen(sz_input) > 0 && !(sz_input[0] & 0x80)))
strText += " ";
strText += sz_input; // full_text = precache + text
m_tokenizer.Tokenize(strText.c_str(), strOut, InputData);
@ -107,7 +112,7 @@ string CTTransformerOnline::AddPunc(const char* sz_input, vector<string> &arr_ca
{
if (!(sentence_words_list[i][0] & 0x80) && (i + 1) < sentence_words_list.size() && !(sentence_words_list[i + 1][0] & 0x80))
{
sentence_words_list[i] = " " + sentence_words_list[i];
sentence_words_list[i] = sentence_words_list[i] + " ";
}
if (nSkipNum < arr_cache.size()) // if skip_num < len(cache):
nSkipNum++;