c# 中传入grammar，测试无效

[Sooloom]

        var grammar = JsonSerializer.Serialize(new List<string> {
"打开",
"关闭",
"运行",
"启动",
"[unk]"
});
        VoskRecognizer rec = new VoskRecognizer(model, 16000.0f, grammar);

在加载中文语音模型后，语音识别不到这些预定词语，请问如何解决？

[nshmyrev]

你好。 您使用的是大型号还是小型号？ big 不支持语法。 你在日志中看到了什么？

[Sooloom]

你好，中文两种模型我都测试过，另外英文模型没有问题，英文正常，中文两种模型我都测试过，出错信息如下： 如果代码中我用 VoskRecognizer rec = new VoskRecognizer(model, 16000.0f, "[\"打开\",\"关闭\",\"运行\", \"启动\", \"[unk]\"]"); 日志中显示如下 LOG (VoskAPI:ReadDataFiles():model.cc:213) Decoding params beam=12 max-active=5000 lattice-beam=4 LOG (VoskAPI:ReadDataFiles():model.cc:216) Silence phones 1:2:3:4:5:6:7:8:9:10 LOG (VoskAPI:RemoveOrphanNodes():nnet-nnet.cc:948) Removed 0 orphan nodes. LOG (VoskAPI:RemoveOrphanComponents():nnet-nnet.cc:847) Removing 0 orphan components. LOG (VoskAPI:ReadDataFiles():model.cc:248) Loading i-vector extractor from model/ivector/final.ie LOG (VoskAPI:ComputeDerivedVars():ivector-extractor.cc:183) Computing derived variables for iVector extractor LOG (VoskAPI:ComputeDerivedVars():ivector-extractor.cc:204) Done. LOG (VoskAPI:ReadDataFiles():model.cc:282) Loading HCL and G from model/graph/HCLr.fst model/graph/Gr.fst LOG (VoskAPI:ReadDataFiles():model.cc:303) Loading winfo model/graph/phones/word_boundary.int LOG (VoskAPI:Recognizer():recognizer.cc:63) ["打开", "关闭", "运行", "启动", "[unk]"] WARNING (VoskAPI:Recognizer():recognizer.cc:84) Ignoring word missing in vocabulary: '打开' WARNING (VoskAPI:Recognizer():recognizer.cc:84) Ignoring word missing in vocabulary: '关闭' WARNING (VoskAPI:Recognizer():recognizer.cc:84) Ignoring word missing in vocabulary: '运行' WARNING (VoskAPI:Recognizer():recognizer.cc:84) Ignoring word missing in vocabulary: '启动' LOG (VoskAPI:Estimate():language_model.cc:142) Estimating language model with ngram-order=2, discount=0.5 LOG (VoskAPI:OutputToFst():language_model.cc:209) Created language model with 2 states and 2 arcs.

如果代码用json序列化 var grammar = JsonSerializer.Serialize(new List { "打开", "关闭", "运行", "启动", "[unk]" }); VoskRecognizer rec = new VoskRecognizer(model, 16000.0f, grammar); 日志如下 LOG (VoskAPI:ReadDataFiles():model.cc:213) Decoding params beam=12 max-active=5000 lattice-beam=4 LOG (VoskAPI:ReadDataFiles():model.cc:216) Silence phones 1:2:3:4:5:6:7:8:9:10 LOG (VoskAPI:RemoveOrphanNodes():nnet-nnet.cc:948) Removed 0 orphan nodes. LOG (VoskAPI:RemoveOrphanComponents():nnet-nnet.cc:847) Removing 0 orphan components. LOG (VoskAPI:ReadDataFiles():model.cc:248) Loading i-vector extractor from model/ivector/final.ie LOG (VoskAPI:ComputeDerivedVars():ivector-extractor.cc:183) Computing derived variables for iVector extractor LOG (VoskAPI:ComputeDerivedVars():ivector-extractor.cc:204) Done. LOG (VoskAPI:ReadDataFiles():model.cc:282) Loading HCL and G from model/graph/HCLr.fst model/graph/Gr.fst LOG (VoskAPI:ReadDataFiles():model.cc:303) Loading winfo model/graph/phones/word_boundary.int LOG (VoskAPI:Recognizer():recognizer.cc:63) ["\u6253\u5F00", "\u5173\u95ED", "\u8FD0\u884C", "\u542F\u52A8", "[unk]"] WARNING (VoskAPI:Recognizer():recognizer.cc:84) Ignoring word missing in vocabulary: '\u6253\u5F00' WARNING (VoskAPI:Recognizer():recognizer.cc:84) Ignoring word missing in vocabulary: '\u5173\u95ED' WARNING (VoskAPI:Recognizer():recognizer.cc:84) Ignoring word missing in vocabulary: '\u8FD0\u884C' WARNING (VoskAPI:Recognizer():recognizer.cc:84) Ignoring word missing in vocabulary: '\u542F\u52A8' LOG (VoskAPI:Estimate():language_model.cc:142) Estimating language model with ngram-order=2, discount=0.5 LOG (VoskAPI:OutputToFst():language_model.cc:209) Created language model with 2 states and 2 arcs.

[nshmyrev]

语法字符串必须是 utf-8 编码，而不是 \u 表示法。 在第一个示例中，编码可能是其他东西，而不是 utf-8。

尝试类似的东西：

Console.OutputEncoding = Encoding.UTF8;
JsonSerializerOptions options = new JsonSerializerOptions
{
    Encoder = JavaScriptEncoder.UnsafeRelaxedJsonEscaping,
};
string grammar = JsonSerializer.Serialize<Dictinory>(dictinory, options);

[Sooloom]

我已经更改了代码 Console.OutputEncoding = Encoding.UTF8; JsonSerializerOptions options = new JsonSerializerOptions { Encoder = JavaScriptEncoder.UnsafeRelaxedJsonEscaping, }; List list = new List { "打开", "关闭", "运行", "启动", "[unk]" }; string grammar = JsonSerializer.Serialize(list, options); VoskRecognizer rec = new VoskRecognizer(model, 16000.0f, grammar); 但是运行以后还报错 LOG (VoskAPI:ReadDataFiles():model.cc:213) Decoding params beam=12 max-active=5000 lattice-beam=4 LOG (VoskAPI:ReadDataFiles():model.cc:216) Silence phones 1:2:3:4:5:6:7:8:9:10 LOG (VoskAPI:RemoveOrphanNodes():nnet-nnet.cc:948) Removed 0 orphan nodes. LOG (VoskAPI:RemoveOrphanComponents():nnet-nnet.cc:847) Removing 0 orphan components. LOG (VoskAPI:ReadDataFiles():model.cc:248) Loading i-vector extractor from model/ivector/final.ie LOG (VoskAPI:ComputeDerivedVars():ivector-extractor.cc:183) Computing derived variables for iVector extractor LOG (VoskAPI:ComputeDerivedVars():ivector-extractor.cc:204) Done. LOG (VoskAPI:ReadDataFiles():model.cc:282) Loading HCL and G from model/graph/HCLr.fst model/graph/Gr.fst LOG (VoskAPI:ReadDataFiles():model.cc:308) Loading winfo model/graph/phones/word_boundary.int LOG (VoskAPI:UpdateGrammarFst():recognizer.cc:287) ["", "ر", "", "", "[unk]"] WARNING (VoskAPI:UpdateGrammarFst():recognizer.cc:308) Ignoring word missing in vocabulary: '' WARNING (VoskAPI:UpdateGrammarFst():recognizer.cc:308) Ignoring word missing in vocabulary: 'ر' WARNING (VoskAPI:UpdateGrammarFst():recognizer.cc:308) Ignoring word missing in vocabulary: '' WARNING (VoskAPI:UpdateGrammarFst():recognizer.cc:308) Ignoring word missing in vocabulary: '' LOG (VoskAPI:Estimate():language_model.cc:142) Estimating language model with ngram-order=2, discount=0.5 LOG (VoskAPI:OutputToFst():language_model.cc:209) Created language model with 2 states and 2 arcs.

[nshmyrev]

Same as https://github.com/alphacep/vosk-api/issues/1318

[nshmyrev]

Actually https://github.com/alphacep/vosk-api/issues/1276