SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2009.13902 · 2020

Utterance-level Dialogue Understanding: An Empirical Study

arXiv · PDF · Open in the Atlas

Code that ran

6 of 11 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 6 have no contract verdict; 0 failures of our instrument; 5 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/2009.13902.json · badge: /api/ran/2009.13902.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
declare-lab/dialogue-understanding canonical (classifier label)
classifier basis: census
6 ran · 5 no recorded run
FunctionStatusWhere it lives
ShuffledDialogLoader Ran
ran on a synthesized input
declare-lab/dialogue-understanding/glove-end-to-end/dataloader.py
code served (permissive licence) · get_code("d41f3696ddc7adb8")
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 DataLoader. 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 tempfile
import os
import numpy as np

# Create temporary files with the expected format
temp_dir = tempfile.mkdtemp()

filename1 = os.path.join(temp_dir, "file1.txt")
filename2 = os.path.join(temp_dir, "file2.txt")
filename3 = os.path.join(temp_dir, "file3.txt")
filename4 = os.path.join(temp_dir, "file4.txt")

# Create test data files with tab-separated values
# Format: ID\tvalue1\tvalue2\tvalue3...
with open(filename1, 'w') as f:
    f.write("id\tutterance_0\tutterance_1\tutterance_2\n")
    f.write("id\tutterance_3\tutterance_4\tutterance_5\n")

with open(filename2, 'w') as f:
    f.write("id\t0\t1\t0\n")
    f.write("id\t1\t1\t0\n")

with open(filename3, 'w') as f:
    f.write("id\t1\t1\t1\n")
    f.write("id\t1\t1\t1\n")

with open(filename4, 'w') as f:
    f.write("id\t0\t1\t0\n")
    f.write("id\t1\t2\t0\n")

ARGS = {
    "filename1": filename1,
    "filename2": filename2,
    "filename3": filename3,
    "filename4": filename4,
    "batch_size": 2,
    "shuffle": True,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataLoader — a DataLoader output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataLoader output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
get_scores Ran
ran on a synthesized input
declare-lab/dialogue-understanding/glove-end-to-end/train_shuffled_context.py
code served (permissive licence) · get_code("53d8a07aaa1a0e6b")
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 list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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 numpy as np

# Set random seed for reproducibility
np.random.seed(0)

# Create sample data for the function
# The function accesses valid_losses with np.argmin and np.argmax
# It accesses valid_fscores and test_fscores as lists of arrays indexed by emotion/act type

# For 'dailydialog' + 'emotion' case: needs 6 emotion types
# For 'dailydialog' + 'act' case: needs 3 act types
# For other cases: needs 1 type

# Let's create data for the 'dailydialog' + 'emotion' case (most complex)
num_epochs = 10

# valid_losses: shape (num_epochs,) - scores for each epoch
valid_losses = np.array([0.5, 0.45, 0.4, 0.38, 0.37, 0.36, 0.365, 0.37, 0.375, 0.38])

# valid_fscores: list of 6 arrays (one per emotion type), each with num_epochs scores
valid_fscores = [
    np.array([0.6, 0.65, 0.68, 0.70, 0.71, 0.72, 0.715, 0.71, 0.705, 0.70]),  # emotion 0
    np.array([0.55, 0.60, 0.63, 0.65, 0.66, 0.67, 0.665, 0.66, 0.655, 0.65]),  # emotion 1
    np.array([0.58, 0.62, 0.65, 0.67, 0.68, 0.69, 0.685, 0.68, 0.675, 0.67]),  # emotion 2
    np.array([0.52, 0.57, 0.60, 0.62, 0.63, 0.64, 0.635, 0.63, 0.625, 0.62]),  # emotion 3
    np.array([0.54, 0.59, 0.62, 0.64, 0.65, 0.66, 0.655, 0.65, 0.645, 0.64]),  # emotion 4
    np.array([0.56, 0.61, 0.64, 0.66, 0.67, 0.68, 0.675, 0.67, 0.665, 0.66]),  # emotion 5
]

# test_fscores: list of 6 arrays (one per emotion type), each with num_epochs scores
test_fscores = [
    np.array([0.61, 0.66, 0.69, 0.71, 0.72, 0.73, 0.725, 0.72, 0.715, 0.71]),  # emotion 0
    np.array([0.56, 0.61, 0.64, 0.66, 0.67, 0.68, 0.675, 0.67, 0.665, 0.66]),  # emotion 1
    np.array([0.59, 0.63, 0.66, 0.68, 0.69, 0.70, 0.695, 0.69, 0.685, 0.68]),  # emotion 2
    np.array([0.53, 0.58, 0.61, 0.63, 0.64, 0.65, 0.645, 0.64, 0.635, 0.63]),  # emotion 3
    np.array([0.55, 0.60, 0.63, 0.65, 0.66, 0.67, 0.665, 0.66, 0.655, 0.65]),  # emotion 4
    np.array([0.57, 0.62, 0.65, 0.67, 0.68, 0.69, 0.685, 0.68, 0.675, 0.67]),  # emotion 5
]

ARGS = {
    'dataset': 'dailydialog',
    'classify': 'emotion',
    'valid_losses': valid_losses,
    'valid_fscores': valid_fscores,
    'test_fscores': test_fscores,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
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
load_pretrained_glove Ran
ran on a synthesized input
declare-lab/dialogue-understanding/glove-end-to-end/utils.py
code served (permissive licence) · get_code("84babf071e30d4e6")
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 dict. 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 tempfile
import os
import numpy as np

# Create a temporary GloVe file for testing
temp_file = tempfile.NamedTemporaryFile(mode='w', suffix='.txt', delete=False, encoding='utf-8')
temp_path = temp_file.name

# Write sample GloVe format data to the temporary file
glove_data = [
    "the 0.418 0.24968 -0.41242 0.1217 0.34527",
    "hello 0.50 0.25 -0.40 0.12 0.35",
    "world 0.45 0.20 -0.38 0.10 0.30",
    "python 0.55 0.28 -0.42 0.15 0.38"
]

for line in glove_data:
    temp_file.write(line + '\n')

temp_file.close()

# Define the arguments dictionary
ARGS = {
    'path': temp_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
pad Ran
ran on a synthesized input
declare-lab/dialogue-understanding/glove-end-to-end/model.py
code served (permissive licence) · get_code("f04300c52dc606a9")
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 float, shape (5, 4) as a later run recorded it. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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.autograd import Variable

torch.manual_seed(0)

# Create a small tensor with shape (2, 4) - batch size 2, feature dim 4
tensor = torch.randn(2, 4)

# Create a Variable (for PyTorch backward compatibility)
variable = Variable(torch.randn(2, 4))

# Test case 1: tensor shorter than target length
# tensor has size 2, we want length 5
ARGS = {
    "tensor": tensor,
    "length": 5
}
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 float, shape (5, 4) — shape (5, 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
pretrained_matrix Ran
ran on a synthesized input
declare-lab/dialogue-understanding/glove-end-to-end/utils.py
code served (permissive licence) · get_code("71a427612c1f9dec")
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 ndarray, ndim 2, dtype family float, shape (5, 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 numpy as np
import tempfile
import os

# Create a temporary GloVe-format file
temp_dir = tempfile.mkdtemp()
glove_path = os.path.join(temp_dir, "glove.txt")

# Write a small GloVe file with a few word vectors
glove_content = """the 0.1 0.2 0.3 0.4
hello 0.5 0.6 0.7 0.8
world 0.9 1.0 1.1 1.2
test 1.3 1.4 1.5 1.6
"""

with open(glove_path, 'w', encoding='utf-8') as f:
    f.write(glove_content)

# Create id_to_token mapping
id_to_token = {
    0: "the",
    1: "hello",
    2: "world",
    3: "unknown_word",
    4: "test"
}

ARGS = {
    "path": glove_path,
    "id_to_token": id_to_token
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (5, 4) — shape (5, 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
update_context Ran
ran on a synthesized input
declare-lab/dialogue-understanding/glove-end-to-end/train_context_control.py
code served (permissive licence) · get_code("006d03457eb044c2")
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 list, ndim 2, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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)

# Analyzing the function:
# - conversations: list of lists (each conv is a list that supports indexing and slicing)
# - target_idx: list of tensors (each has .item() method to get scalar index)
# - cc: dict with keys 'past_context', 'future_context', 'past_del', 'future_del'

# Build small stand-in conversations (lists of tokens/strings)
conversations = [
    ['token_a', 'token_b', 'token_c', 'token_d', 'token_e'],
    ['word_1', 'word_2', 'word_3', 'word_4', 'word_5', 'word_6']
]

# Build target indices as tensors with .item() method
target_idx = [
    torch.tensor(2),  # index 2 for first conversation
    torch.tensor(3)   # index 3 for second conversation
]

# Build context config dict
cc = {
    'past_context': 1,
    'future_context': 1,
    'past_del': False,
    'future_del': False
}

ARGS = {
    'conversations': conversations,
    'target_idx': target_idx,
    'cc': cc
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 2, dtype family float — a list output has no array shape of its own; none is stated
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
DialogLoader Not yet run
unverified (harvested, no recorded run)
declare-lab/dialogue-understanding/glove-end-to-end/dataloader.py
code served (permissive licence) · get_code("163651cb2fba95af")
How it ran

no invocation recorded.

train_or_eval_model Not yet run
unverified (harvested, no recorded run)
declare-lab/dialogue-understanding/glove-end-to-end/train_context_control.py
code served (permissive licence) · get_code("4e94472c1b5644ad")
How it ran

no invocation recorded.

train_or_eval_model Not yet run
unverified (harvested, no recorded run)
declare-lab/dialogue-understanding/glove-end-to-end/train_inter_speaker.py
code served (permissive licence) · get_code("ad94662f9e3f223c")
How it ran

no invocation recorded.

train_or_eval_model Not yet run
unverified (harvested, no recorded run)
declare-lab/dialogue-understanding/glove-end-to-end/train_intra_speaker.py
code served (permissive licence) · get_code("13d9788acddaa103")
How it ran

no invocation recorded.

train_or_eval_model Not yet run
unverified (harvested, no recorded run)
declare-lab/dialogue-understanding/glove-end-to-end/train_shuffled_context.py
code served (permissive licence) · get_code("49213a805387fe16")
How it ran

no invocation recorded.

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 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 11 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

The recent abundance of conversational data on the Web and elsewhere calls for effective NLP systems for dialog understanding. Complete utterance-level understanding often requires context understanding, defined by nearby utterances. In recent years, a number of approaches have been proposed for various utterance-level dialogue understanding tasks. Most of these approaches account for the context for effective understanding. In this paper, we explore and quantify the role of context for different aspects of a dialogue, namely emotion, intent, and dialogue act identification, using state-of-the-art dialog understanding methods as baselines. Specifically, we employ various perturbations to distort the context of a given utterance and study its impact on the different tasks and baselines. This provides us with insights into the fundamental contextual controlling factors of different aspects of a dialogue. Such insights can inspire more effective dialogue understanding models, and provide support for future text generation approaches. The implementation pertaining to this work is available at https://github.com/declare-lab/dialogue-understanding.

For agents

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

get_harvested_code_for_paper("2009.13902")
get_code_for_paper("2009.13902")
have("2009.13902")

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

curl https://syntology.ai/api/ran/2009.13902.json

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

[![Syntology run record](https://syntology.ai/api/ran/2009.13902.svg)](https://syntology.ai/paper/2009.13902)

Connect an agent — have() is free.