SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1609.06773 · 2016

Joint CTC-Attention based End-to-End Speech Recognition using Multi-task Learning

arXiv · PDF · Open in the Atlas

Code that ran

3 of 10 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 3 have no contract verdict; 0 failures of our instrument; 7 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/1609.06773.json · badge: /api/ran/1609.06773.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
s3prl/End-to-end-ASR-Pytorch community (classifier label)
classifier basis not recorded on the link
2 ran · 2 no recorded run
Alexander-H-Liu/End-to-end-ASR-Pytorch community (classifier label)
classifier basis not recorded on the link
1 ran · 4 no recorded run
neil-zeng/asr community (classifier label)
classifier basis not recorded on the link
0 ran · 1 no recorded run
FunctionStatusWhere it lives
collect_audio_batch Ran
ran on a synthesized input
Alexander-H-Liu/End-to-end-ASR-Pytorch/src/data.py
code served (permissive licence) · get_code("340cb8273953e493")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. It ran in syntology-harvest-sandbox:v5-deps (sha256:6a292aa1aed1) at 2026-09-26T21:52:49Z. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
from pathlib import Path
import tempfile

torch.manual_seed(0)

# Create a mock audio transform that returns a tensor
class MockAudioTransform(nn.Module):
    def __init__(self):
        super().__init__()
    
    def forward(self, audio_path):
        # Return a small tensor with varying lengths
        # Use hash of path to get deterministic but different sizes
        path_hash = hash(audio_path) % 100
        length = 400 + (path_hash % 300)  # Between 400-700 to vary batch sizes
        return torch.randn(length, 4)

audio_transform = MockAudioTransform()

# Create a batch: list of tuples (audio_path, token_list)
# audio_path is a string, token_list is a list of ints
batch = [
    ("audio_file_1.wav", [1, 2, 3, 4, 5]),
    ("audio_file_2.wav", [6, 7, 8]),
]

mode = 'train'

ARGS = {
    'batch': batch,
    'audio_transform': audio_transform,
    'mode': mode,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image syntology-harvest-sandbox:v5-deps (sha256:6a292aa1aed1); not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
collect_text_batch Ran
ran on a synthesized input
s3prl/End-to-end-ASR-Pytorch/src/data.py
code served (permissive licence) · get_code("c4ec9d88dda3f211")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family int, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
from torch.nn.utils.rnn import pad_sequence

torch.manual_seed(0)

# The function expects:
# - batch: list of list of int token ids
# - mode: string, either 'train' or 'eval'

# Create sample batch data: list of sequences where each sequence is a list of token ids
# We'll create a batch with sequences of varying lengths
batch = [
    [10, 20, 30, 40],           # sequence 1: 4 tokens
    [50, 60, 70],               # sequence 2: 3 tokens
]

mode = 'train'

ARGS = {
    'batch': batch,
    'mode': mode
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family int, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
create_dataset Ran
ran on a synthesized input
s3prl/End-to-end-ASR-Pytorch/src/data.py
code served (permissive licence) · get_code("ac71d8ab36c2fae0")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

class SimpleTokenizer:
    """Minimal tokenizer stand-in"""
    def __init__(self):
        self.vocab_size = 100
    
    def encode(self, text):
        return [1, 2, 3]
    
    def decode(self, ids):
        return "decoded"

class MockDataset:
    """Minimal dataset stand-in that mimics LibriDataset/SnipsDataset interface"""
    def __init__(self, path, split, tokenizer, bucket_size, ascending=False, text_mode='character'):
        self.path = path
        self.split = split
        self.tokenizer = tokenizer
        self.bucket_size = bucket_size
        self.ascending = ascending
        self.text_mode = text_mode
        # Create a small dataset with 5 samples
        self.data = [
            (torch.randn(16, 4), torch.tensor([1, 2])),
            (torch.randn(16, 4), torch.tensor([3, 4])),
            (torch.randn(16, 4), torch.tensor([5, 6])),
            (torch.randn(16, 4), torch.tensor([7, 8])),
            (torch.randn(16, 4), torch.tensor([9, 10])),
        ]
    
    def __len__(self):
        return len(self.data)
    
    def __getitem__(self, idx):
        return self.data[idx]

# Monkey-patch the corpus modules to return our mock dataset
import sys
from types import ModuleType

librispeech_module = ModuleType('librispeech')
librispeech_module.LibriDataset = MockDataset
sys.modules['corpus.librispeech'] = librispeech_module
sys.modules['corpus'] = ModuleType('corpus')
sys.modules['corpus'].librispeech = librispeech_module

snips_module = ModuleType('snips')
snips_module.SnipsDataset = MockDataset
sys.modules['corpus.snips'] = snips_module
sys.modules['corpus'].snips = snips_module

ARGS = {
    'tokenizer': SimpleTokenizer(),
    'ascending': False,
    'name': 'librispeech',
    'path': '/fake/path/librispeech',
    'bucketing': True,
    'batch_size': 32,
    'eval_batch_size': 64,
    'train_split': ['train-clean-100'],
    'dev_split': ['dev-clean'],
    'test_split': ['test-clean'],
    'text_mode': 'character'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
corpus, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
collect_audio_batch Not yet run
unverified (harvested, no recorded run)
s3prl/End-to-end-ASR-Pytorch/src/data.py
code served (permissive licence) · get_code("bded34b0a51d533c")
How it ran

no invocation recorded.

collect_text_batch Not yet run
unverified (harvested, no recorded run)
Alexander-H-Liu/End-to-end-ASR-Pytorch/src/data.py
code served (permissive licence) · get_code("ca15b9ec596a30d0")
How it ran

no invocation recorded.

create_dataset Not yet run
unverified (harvested, no recorded run)
Alexander-H-Liu/End-to-end-ASR-Pytorch/src/data.py
code served (permissive licence) · get_code("e2ec1a4e3221772d")
How it ran

no invocation recorded.

create_dataset Not yet run
unverified (harvested, no recorded run)
neil-zeng/asr/src/data.py
code served (permissive licence) · get_code("3929b38d6fca37e2")
How it ran

no invocation recorded.

generate_embedding Not yet run
unverified (harvested, no recorded run)
Alexander-H-Liu/End-to-end-ASR-Pytorch/src/bert_embedding.py
code served (permissive licence) · get_code("a7310a3d1f046472")
How it ran

no invocation recorded.

get_testcases Not yet run
unverified (harvested, no recorded run)
s3prl/End-to-end-ASR-Pytorch/slot_evaluate.py
code served (permissive licence) · get_code("9a8e285c706f7fe8")
How it ran

no invocation recorded.

speech_aug_scheduler Not yet run
unverified (harvested, no recorded run)
Alexander-H-Liu/End-to-end-ASR-Pytorch/src/optim.py
code served (permissive licence) · get_code("90a0a3f3574b92a6")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 10 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Recently, there has been an increasing interest in end-to-end speech recognition that directly transcribes speech to text without any predefined alignments. One approach is the attention-based encoder-decoder framework that learns a mapping between variable-length input and output sequences in one step using a purely data-driven method. The attention model has often been shown to improve the performance over another end-to-end approach, the Connectionist Temporal Classification (CTC), mainly because it explicitly uses the history of the target character without any conditional independence assumptions. However, we observed that the performance of the attention has shown poor results in noisy condition and is hard to learn in the initial training stage with long input sequences. This is because the attention model is too flexible to predict proper alignments in such cases due to the lack of left-to-right constraints as used in CTC. This paper presents a novel method for end-to-end speech recognition to improve robustness and achieve fast convergence by using a joint CTC-attention model within the multi-task learning framework, thereby mitigating the alignment issue. An experiment on the WSJ and CHiME-4 tasks demonstrates its advantages over both the CTC and attention-based encoder-decoder baselines, showing 5.4-14.6% relative improvements in Character Error Rate (CER).

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("1609.06773")
get_code_for_paper("1609.06773")
have("1609.06773")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/1609.06773.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/1609.06773.svg)](https://syntology.ai/paper/1609.06773)

Connect an agent — have() is free.