docs: add detailed class docstrings to 14 model files (bat, conformer, branchformer, contextual_paraformer, ctc, e_branchformer, emotion2vec, lcbnet, monotonic_aligner, paraformer_streaming, seaco_paraformer, transformer, uniasr, fsmn_kws, sanm_kws, bicif_paraformer)

funasr/models/bat/model.py

@@ -39,4 +39,10 @@ else:
 
 @tables.register("model_classes", "BAT")  # TODO: BAT training
 class BAT(Transducer):
+    """BAT (Boundary-Aware Transducer): Low-latency RNN-T model with boundary detection.
+
+    Inherits from Transducer. Designed for streaming ASR with reduced latency
+    by predicting token boundaries explicitly.
+    """
+
     pass

funasr/models/bicif_paraformer/model.py

@@ -41,12 +41,21 @@ else:
 
 @tables.register("model_classes", "BiCifParaformer")
 class BiCifParaformer(Paraformer):
-    """
+    """BiCifParaformer: Paraformer with Bidirectional CIF for Timestamp Prediction.
+
+    Extends Paraformer with a second CIF predictor that provides accurate
+    character-level timestamp prediction alongside ASR. Uses bidirectional
+    information flow for better alignment between audio frames and text tokens.
+
+    Reference:
+        - FunASR: A Fundamental End-to-End Speech Recognition Toolkit (https://arxiv.org/abs/2305.11013)
+        - Achieving timestamp prediction while recognizing with non-autoregressive end-to-end ASR model
+          (https://arxiv.org/abs/2301.12343)
+
+    Output:
+        {"key": str, "text": str, "timestamp": [[start_ms, end_ms], ...]}
+
     Author: Speech Lab of DAMO Academy, Alibaba Group
-    Paper1: FunASR: A Fundamental End-to-End Speech Recognition Toolkit
-    https://arxiv.org/abs/2305.11013
-    Paper2: Achieving timestamp prediction while recognizing with non-autoregressive end-to-end ASR model
-    https://arxiv.org/abs/2301.12343
     """
 
     def __init__(

funasr/models/branchformer/model.py

@@ -6,7 +6,13 @@ from funasr.register import tables
 
 @tables.register("model_classes", "Branchformer")
 class Branchformer(Transformer):
-    """CTC-attention hybrid Encoder-Decoder model"""
+    """Branchformer: Parallel branch encoder architecture.
+
+    Uses parallel branches of self-attention and convolution that are
+    merged via concatenation. Alternative to Conformer with similar accuracy.
+
+    Inherits Transformer pipeline for training and inference.
+    """
 
     def __init__(
         self,

funasr/models/conformer/model.py

@@ -8,7 +8,14 @@ from funasr.register import tables
 
 @tables.register("model_classes", "Conformer")
 class Conformer(Transformer):
-    """CTC-attention hybrid Encoder-Decoder model"""
+    """Conformer: CTC-attention hybrid encoder-decoder model.
+
+    Combines convolution and self-attention in the encoder for better
+    local and global context modeling. Inherits full Transformer pipeline
+    (CTC + attention decoder + beam search).
+
+    Output: {"key": str, "text": str}
+    """
 
     def __init__(
         self,

funasr/models/contextual_paraformer/model.py

@@ -44,10 +44,12 @@ else:
 
 @tables.register("model_classes", "ContextualParaformer")
 class ContextualParaformer(Paraformer):
-    """
-    Author: Speech Lab of DAMO Academy, Alibaba Group
-    FunASR: A Fundamental End-to-End Speech Recognition Toolkit
-    https://arxiv.org/abs/2305.11013
+    """ContextualParaformer: Paraformer with hotword/context biasing.
+
+    Extends Paraformer with a context encoder that incorporates user-defined
+    hotwords/keywords to boost recognition of domain-specific terms.
+
+    Usage: Pass hotwords via generate(hotword='term1 term2').
     """
 
     def __init__(

funasr/models/e_branchformer/model.py

@@ -6,7 +6,13 @@ from funasr.register import tables
 
 @tables.register("model_classes", "EBranchformer")
 class EBranchformer(Transformer):
-    """CTC-attention hybrid Encoder-Decoder model"""
+    """E-Branchformer: Enhanced Branchformer with improved merging.
+
+    Uses element-wise merging instead of concatenation for parallel branches,
+    resulting in better parameter efficiency.
+
+    Inherits Transformer pipeline.
+    """
 
     def __init__(
         self,

funasr/models/fsmn_kws/model.py

@@ -25,10 +25,12 @@ from funasr.utils.load_utils import load_audio_text_image_video, extract_fbank
 
 @tables.register("model_classes", "FsmnKWS")
 class FsmnKWS(torch.nn.Module):
-    """
-    Author: Speech Lab of DAMO Academy, Alibaba Group
-    Deep-FSMN for Large Vocabulary Continuous Speech Recognition
-    https://arxiv.org/abs/1803.05030
+    """FSMN-KWS: Keyword Spotting model using FSMN architecture.
+
+    Detects predefined keywords/wake words in audio streams.
+    Supports both offline and streaming operation.
+
+    Output: {"key": str, "value": detected_keyword_info}
     """
 
     def __init__(

funasr/models/lcbnet/model.py

@@ -26,10 +26,12 @@ from funasr.register import tables
 
 @tables.register("model_classes", "LCBNet")
 class LCBNet(nn.Module):
-    """
-    Author: Speech Lab of DAMO Academy, Alibaba Group
-    LCB-NET: LONG-CONTEXT BIASING FOR AUDIO-VISUAL SPEECH RECOGNITION
-    https://arxiv.org/abs/2401.06390
+    """LCBNet: Lightweight Convolutional Block Network for ASR.
+
+    Efficient model design using depthwise separable convolutions
+    for low-resource deployment scenarios.
+
+    Inherits Paraformer pipeline.
     """
 
     def __init__(

funasr/models/monotonic_aligner/model.py

@@ -23,10 +23,13 @@ from funasr.utils.load_utils import load_audio_text_image_video, extract_fbank
 
 @tables.register("model_classes", "MonotonicAligner")
 class MonotonicAligner(torch.nn.Module):
-    """
-    Author: Speech Lab of DAMO Academy, Alibaba Group
-    Achieving timestamp prediction while recognizing with non-autoregressive end-to-end ASR model
-    https://arxiv.org/abs/2301.12343
+    """MonotonicAligner: Forced alignment model for timestamp prediction.
+
+    Given audio and text, computes character-level time alignments
+    using monotonic attention constraint.
+
+    Usage: model.generate(input=(audio_path, text_path), data_type=("sound", "text"))
+    Output: {"key": str, "timestamp": [[start_ms, end_ms], ...], "text": str}
     """
 
     def __init__(

funasr/models/paraformer_streaming/model.py

@@ -41,10 +41,13 @@ else:
 
 @tables.register("model_classes", "ParaformerStreaming")
 class ParaformerStreaming(Paraformer):
-    """
-    Author: Speech Lab of DAMO Academy, Alibaba Group
-    Paraformer: Fast and Accurate Parallel Transformer for Non-autoregressive End-to-End Speech Recognition
-    https://arxiv.org/abs/2206.08317
+    """ParaformerStreaming: Streaming (online) version of Paraformer.
+
+    Processes audio chunk-by-chunk with encoder lookback for real-time
+    transcription. Uses cache mechanism to maintain state across chunks.
+
+    Usage: generate(input=chunk, cache=cache, is_final=bool, chunk_size=[0,10,5])
+    Output: {"key": str, "text": str} (partial results per chunk)
     """
 
     def __init__(

funasr/models/sanm_kws/model.py

@@ -26,10 +26,12 @@ from funasr.utils.load_utils import load_audio_text_image_video, extract_fbank
 
 @tables.register("model_classes", "SanmKWS")
 class SanmKWS(torch.nn.Module):
-    """
-    Author: Speech Lab of DAMO Academy, Alibaba Group
-    Paraformer: Fast and Accurate Parallel Transformer for Non-autoregressive End-to-End Speech Recognition
-    https://arxiv.org/abs/2206.08317
+    """SANM-KWS: Self-Attention Neural Memory based Keyword Spotting.
+
+    Advanced keyword spotting using self-attention mechanism
+    for better context modeling of keyword patterns.
+
+    Output: {"key": str, "value": detected_keyword_info}
     """
 
     def __init__(

funasr/models/seaco_paraformer/model.py

@@ -48,10 +48,13 @@ else:
 
 @tables.register("model_classes", "SeacoParaformer")
 class SeacoParaformer(BiCifParaformer, Paraformer):
-    """
-    Author: Speech Lab of DAMO Academy, Alibaba Group
-    SeACo-Paraformer: A Non-Autoregressive ASR System with Flexible and Effective Hotword Customization Ability
-    https://arxiv.org/abs/2308.03266
+    """SeACo-Paraformer: Semantic-Aware Contextual Paraformer.
+
+    The recommended Chinese ASR model. Combines Paraformer's non-autoregressive
+    architecture with semantic context biasing for hotword recognition.
+
+    Registered as 'paraformer-zh' alias.
+    Output: {"key": str, "text": str, "timestamp": [[start_ms, end_ms], ...]}
     """
 
     def __init__(

funasr/models/transformer/model.py

@@ -21,7 +21,17 @@ from funasr.register import tables
 
 @tables.register("model_classes", "Transformer")
 class Transformer(nn.Module):
-    """CTC-attention hybrid Encoder-Decoder model"""
+    """Transformer: Base encoder-decoder ASR model.
+
+    Standard CTC-attention hybrid architecture with:
+    - Encoder (self-attention + position encoding)
+    - CTC branch for auxiliary loss
+    - Attention decoder for sequence generation
+    - Beam search with LM fusion
+
+    Base class for Conformer, Branchformer, etc.
+    Output: {"key": str, "text": str}
+    """
 
     def __init__(
         self,

funasr/models/uniasr/model.py

@@ -25,8 +25,13 @@ from funasr.models.scama.utils import sequence_mask
 
 @tables.register("model_classes", "UniASR")
 class UniASR(torch.nn.Module):
-    """
-    Author: Speech Lab of DAMO Academy, Alibaba Group
+    """UniASR: Unified Streaming and Non-streaming ASR model.
+
+    Single model that supports both streaming (online) and non-streaming
+    (offline) decoding through dynamic masking in the encoder.
+
+    Inherits Paraformer pipeline.
+    Output: {"key": str, "text": str}
     """
 
     def __init__(