Python API for speaker diarization. (#1400)

sherpa-onnx/csrc/offline-speaker-diarization-pyannote-impl.h

@@ -103,7 +103,7 @@ class OfflineSpeakerDiarizationPyannoteImpl
     auto chunk_speaker_samples_list_pair = GetChunkSpeakerSampleIndexes(labels);
     Matrix2D embeddings =
         ComputeEmbeddings(audio, n, chunk_speaker_samples_list_pair.second,
-                          callback, callback_arg);
+                          std::move(callback), callback_arg);
 
     std::vector<int32_t> cluster_labels = clustering_.Cluster(
         &embeddings(0, 0), embeddings.rows(), embeddings.cols());

sherpa-onnx/csrc/offline-speaker-diarization-result.h

@@ -28,6 +28,8 @@ class OfflineSpeakerDiarizationSegment {
   const std::string &Text() const { return text_; }
   float Duration() const { return end_ - start_; }
 
+  void SetText(const std::string &text) { text_ = text; }
+
   std::string ToString() const;
 
  private:

sherpa-onnx/csrc/offline-speaker-diarization.h

@@ -34,10 +34,13 @@ struct OfflineSpeakerDiarizationConfig {
   OfflineSpeakerDiarizationConfig(
       const OfflineSpeakerSegmentationModelConfig &segmentation,
       const SpeakerEmbeddingExtractorConfig &embedding,
-      const FastClusteringConfig &clustering)
+      const FastClusteringConfig &clustering, float min_duration_on,
+      float min_duration_off)
       : segmentation(segmentation),
         embedding(embedding),
-        clustering(clustering) {}
+        clustering(clustering),
+        min_duration_on(min_duration_on),
+        min_duration_off(min_duration_off) {}
 
   void Register(ParseOptions *po);
   bool Validate() const;