使用sherpa-onnx给文字添加标点符号

上篇文章https://blog.csdn.net/AWNUXCVBN/article/details/138372795 识别出的文字都是没有标点符号的，虽然可以添加整句断行，但总觉得怪怪的…… 于是看了一下，ncnn没有添加标点符号的功能，翻了一下onnx有示例，但没有C#的，于是照着搞了一番

using System.Runtime.InteropServices;

namespace SherpaOnnx
{
    public class OfflinePunctuation : IDisposable
    {
        private HandleRef _handle;

        public OfflinePunctuation(SherpaOnnxOfflinePunctuationConfig config)
        {
            IntPtr intPtr = OfflinePunctuation.SherpaOnnxCreateOfflinePunctuation(config);
            this._handle = new HandleRef(this, intPtr);
        }

        public string AddPunctuation(string text)
        {
           return SherpaOfflinePunctuationAddPunct(this._handle.Handle,text);
        }

        public void Dispose()
        {
            this.Cleanup();
            GC.SuppressFinalize(this);
        }

        ~OfflinePunctuation()
        {
            this.Cleanup();
        }

        private void Cleanup()
        {
            OfflinePunctuation.DestroyOfflinePunctuation(this._handle.Handle);
            this._handle = new HandleRef(this, IntPtr.Zero);
        }

        [DllImport("sherpa-onnx-c-api", EntryPoint = "SherpaOnnxDestroyOfflinePunctuation")]
        private static extern IntPtr DestroyOfflinePunctuation(IntPtr handle);

        [DllImport("sherpa-onnx-c-api", EntryPoint = "SherpaOnnxCreateOfflinePunctuation")]
        private static extern IntPtr SherpaOnnxCreateOfflinePunctuation(SherpaOnnxOfflinePunctuationConfig config);

        [DllImport("sherpa-onnx-c-api", EntryPoint = "SherpaOfflinePunctuationAddPunct")]
        private static extern string SherpaOfflinePunctuationAddPunct(IntPtr ptr, string text);
    }
}

namespace SherpaOnnx
{
    public struct SherpaOnnxOfflinePunctuationConfig
    {
        public SherpaOnnxOfflinePunctuationModelConfig model;
    }
}

namespace SherpaOnnx
{
    public struct SherpaOnnxOfflinePunctuationModelConfig 
    {
        public string ctTransformer = "";
        public int numThreads = 1;
        public bool debug = true;
        public string provider = "cpu";

        public SherpaOnnxOfflinePunctuationModelConfig (string ctTransformer, int numThreads, bool debug, string provider)
        {
            this.ctTransformer = ctTransformer;
            this.numThreads = numThreads;
            this.debug = debug;
            this.provider = provider;
        }
    }
}

SherpaOnnx.SherpaOnnxOfflinePunctuationConfig soopc = new SherpaOnnx.SherpaOnnxOfflinePunctuationConfig();
SherpaOnnx.SherpaOnnxOfflinePunctuationModelConfig soopmc =
    new SherpaOnnx.SherpaOnnxOfflinePunctuationModelConfig
    (Environment.CurrentDirectory + "/sherpa-onnx-punct-ct-transformer-zh-en-vocab272727-2024-04-12/model.onnx", 1, false, "cpu");
soopc.model = soopmc;
offlinePunctuation = new SherpaOnnx.OfflinePunctuation(soopc);
Console.writeLine(offlinePunctuation.AddPunctuation("When you only have one life that's what makes it special"));

看起来一切顺利，运行一下，不出意外的出意外了

微信群里@fangjun大佬指点了一番，说是要设置系统使用utf8编码

设置重启之后，正常了！

另外，大佬说，这个加标点模型，支持中英文。默认所有的标点，都是中文标点，即使你输入的都是英文。

最后，案发现场在https://github.com/xue-fei/uSherpaServer.git