Skip to main content
CARLA HQ
Speech Processing & Acoustic Engineering Operations & Workforce Analytics use casesNACE M72Binary ClassificationTarget: phoneme_sound_classOpenML #1489 ↗

Phoneme

Benchmark dataset comprising 5,404 acoustic observations across 5 spectral harmonic features designed to classify speech vowel phonemes into nasal versus oral sound categories.

✨ Try in CarlaInstall Chrome Extension ↗100% In-Browser WebGPU • Zero Cloud Upload
5,404Records (Rows)
5Predictive Features
6 / 0Numeric / Categorical
0.0%Missing Value Ratio
0.02781 - AUC Error
5.13sWebGPU Mean Split
BUSINESS CONTEXT & OBJECTIVEAcoustic AI & Voice Recognition Systems (M72)

Business Objective: Phoneme

01Business Context

Analytical speech recognition pipelines and automated voice assistants require robust, speaker-independent phonetic feature extraction to accurately transcribe spoken language in real time under noisy conditions.

02Analytical Objective

Classify acoustic vowel segments as either nasal or oral based on the energy-normalized amplitudes of their first five spectral harmonics.

03Economic & Decision Impact

Accurate phoneme disambiguation prevents cascading phonetic transcription errors in automated speech-to-text and IVR systems, directly reducing downstream semantic misunderstanding and customer call escalation costs.

ML Benchmark Narrative

Standard tree-based ensembles (XGBoost, LightGBM, and Random Forest) routinely achieve between 87% and 90% classification accuracy on this benchmark by capturing complex nonlinear spectral harmonic boundaries. Tabular foundation models (TabICL / Carla) deliver comparable zero-shot classification performance directly in-context without manual hyperparameter optimization or local model training.

Source Origin:ELENA Project / ESPRIT 5516 (ROARS) & KEEL Repository
Creator:Dominique Van Cappel (THOMSON-SINTRA) and the ELENA Project Consortium (1993)
License:Open Data / Public Domain
FEATURE SPECIFICATION & DATA DICTIONARY

Target & Feature Column Definitions

Exhaustive business definitions, measurement units, roles, and target variables across all 6 columns.

ColVariable / Feature NameRoleTypeUnit / ScaleDescription & Business Meaning
Aphoneme_sound_classPhoneme Sound ClassTARGETNUM
Target phonetic articulation category representing nasal vowel sounds (Class 1) versus oral vowel sounds (Class 2).
Bharmonic_1_normalized_amplitudeHarmonic 1 Normalized AmplitudeFeatureNUMratio
Normalized amplitude of the 1st spectral harmonic (AH1/Ene), signed positive for local spectral maxima and negative otherwise.
Charmonic_2_normalized_amplitudeHarmonic 2 Normalized AmplitudeFeatureNUMratio
Normalized amplitude of the 2nd spectral harmonic (AH2/Ene), signed positive for local spectral maxima and negative otherwise.
Dharmonic_3_normalized_amplitudeHarmonic 3 Normalized AmplitudeFeatureNUMratio
Normalized amplitude of the 3rd spectral harmonic (AH3/Ene), signed positive for local spectral maxima and negative otherwise.
Eharmonic_4_normalized_amplitudeHarmonic 4 Normalized AmplitudeFeatureNUMratio
Normalized amplitude of the 4th spectral harmonic (AH4/Ene), signed positive for local spectral maxima and negative otherwise.
Fharmonic_5_normalized_amplitudeHarmonic 5 Normalized AmplitudeFeatureNUMratio
Normalized amplitude of the 5th spectral harmonic (AH5/Ene), signed positive for local spectral maxima and negative otherwise.
DATASET PREVIEW

Interactive Data Table

Explore rows, feature values, and target labels for Phoneme.

Loading dataset...
LOCAL EXECUTION TELEMETRY

TabICLv2 WebGPU Benchmark Results

Complete performance metrics on persisted IID or non-IID evaluation splits, executed 100% locally in the browser sandbox.

Mean Split Latency?Mean wall-clock execution time per persisted evaluation split running 100% locally via WebGPU inside Chrome.Explore Latency Guide →
5.13s9 splits • 8 ensembles/split
Total Evaluation Runtime?Total wall-clock duration across all selected splits, including in-context encoding, WebGPU shader execution, and result aggregation.Explore Split Protocol →
46.21sIncludes warmup & sync
Evaluation Protocol?Precomputed row indices ensure the browser and Python runners evaluate identical out-of-sample observations without leakage.Explore Split Protocol →
3×3 SplitsRepeated IID • 9 total
TabICLv2 Score?Primary task performance score (1 - AUC Error) achieved by TabICLv2 across out-of-sample evaluation splits.Explore Metric Formula →
0.02781 - AUC Error
Evaluation SplitTrain RowsTest Rows
Split Duration?Wall-clock inference time taken for this evaluation split running 100% locally via WebGPU.Learn more →
Accuracy?Proportion of correct test predictions across all classes.Google ML Guide & Details →
ROC-AUC?Area under the ROC curve evaluating ranking capability across all thresholds.Google ML Guide & Details →
F1 Score?Harmonic mean of Precision and Recall, robust against class imbalance.Learn more →
Precision?Proportion of predicted positives that were actually positive.Google ML Guide & Details →
Recall?Proportion of actual positive cases successfully captured.Google ML Guide & Details →
R1 / F13,6021,8025.26s91.68%0.97050.85850.85690.8601
R1 / F23,6031,8015.10s92.06%0.97090.86390.86810.8598
R1 / F33,6031,8015.10s92.39%0.97640.86940.87690.8620
R2 / F13,6021,8025.10s90.95%0.97290.84520.84920.8412
R2 / F23,6031,8015.15s92.62%0.97660.87150.88950.8542
R2 / F33,6031,8015.14s92.28%0.97320.86750.87500.8601
R3 / F13,6021,8025.10s92.56%0.97260.86990.89420.8469
R3 / F23,6031,8015.14s90.84%0.96650.84330.84570.8409
R3 / F33,6031,8015.13s90.67%0.96960.83820.85460.8223
Mean ± Std5.13s91.78% ± 0.73%0.972 ± 0.0030.859 ± 0.0120.8680.850
UNIFIED BENCHMARK LEADERBOARD

Foundation Models vs Traditional ML Baselines

Side-by-side evaluation of our in-browser Carla engine (WebGPU), open-source tabular foundation models, and OpenML baselines.

RankAlgorithm / ModelModel FamilyRuntime
Predictive Accuracy?Proportion of correct test predictions across cross-validation splits.Google ML Guide & Details →
ROC-AUC?Area under the ROC curve evaluating ranking capability.Learn more →
F-Measure?Harmonic mean of precision and recall.Learn more →
Reference / Repo
#1Google TabFM v1.0Tabular Foundation ModelPyTorch/Python92.98%0.97980.8793google-research/tabfm ↗
#2EXAONE TabularTabular Foundation ModelPyTorch/Python91.96%0.97240.8617LGAI-Research/EXAONE-Tabular ↗
#3TabPFN v3Tabular Foundation ModelPyTorch/Python91.95%0.97260.8609PriorLabs/tabpfn ↗
#4Carla Engine (TabICLv2 WebGPU)Tabular Foundation ModelBrowser91.78%0.97210.8586100% In-Browser
#5Streaming nanotabiclTabular Foundation ModelPyTorch/Python91.77%0.97220.8578soda-inria/nanotabicl ↗
#6nanotabicl VanillaTabular Foundation ModelPyTorch/Python91.77%0.97220.8583soda-inria/nanotabicl ↗
#7KStarMachine Learning ModelPython91.73%0.96740.9162OpenML #1672792 ↗
#8Full TabICLv2 (PyTorch Reference)Tabular Foundation ModelPyTorch/Python91.72%0.97220.8573soda-inria/tabicl ↗
#9RandomForestRandom ForestPython91.52%0.96430.9149OpenML #568789 ↗
#10AttributeSelectedClassifier RandomForestRandom ForestPython91.21%0.96220.9118OpenML #569670 ↗
#11Bagging RandomForestRandom ForestPython90.88%0.96230.9085OpenML #1672965 ↗
#12RandomCommittee RandomTreeDecision TreePython90.71%0.95120.9062OpenML #1761214 ↗
#13IB1Machine Learning ModelPython90.43%0.87610.9035OpenML #1665152 ↗
#14AdaBoostM1 IBkAdaBoostPython90.43%0.87630.9035OpenML #1673270 ↗
#15Bagging IBkk-Nearest Neighbors (k-NN)Python90.10%0.92810.9000OpenML #1674041 ↗
#16AttributeSelectedClassifier IBkk-Nearest Neighbors (k-NN)Python90.06%0.87150.8997OpenML #569498 ↗
#17Bagging RandomTreeDecision TreePython90.04%0.94800.8989OpenML #1672373 ↗
STATISTICAL DISTRIBUTIONS

Variable Schema & Summary Distributions

Observed numerical ranges, category cardinalities, missing rates, and sample values across 5,404 rows.

ColVariable NameTypeMissingDistinctSummary Stats / DistributionSample Values
Aphoneme_sound_classTARGETNUM02[1, 2] μ=1.3 σ=0.5111
Bharmonic_1_normalized_amplitudeNUM05,336[-2.9337, 3.8286] μ=0 σ=1.00.489927-0.6412650.870593
Charmonic_2_normalized_amplitudeNUM05,312[-3.0387, 3.6648] μ=0 σ=1.0-0.4515280.109245-0.459862
Dharmonic_3_normalized_amplitudeNUM05,308[-2.7959, 2.6309] μ=0 σ=1.0-1.047990.292130.578159
Eharmonic_4_normalized_amplitudeNUM05,336[-2.4853, 3.0469] μ=0 σ=1.0-0.598693-0.9168040.806634
Fharmonic_5_normalized_amplitudeNUM04,499[-2.3675, 4.5867] μ=0 σ=1.0-0.0204180.2402230.835248
FREQUENTLY ASKED QUESTIONS

Frequently Asked Questions: Phoneme

Common questions regarding the Phoneme dataset, machine learning task formulations, and in-browser tabular inference.

What is the Phoneme dataset used for?

The Phoneme dataset is an acoustic speech processing benchmark used to train and evaluate machine learning models in distinguishing nasal vowel sounds from oral vowel sounds based on five normalized harmonic spectral amplitudes.

Can I run zero-server predictions on this dataset in Google Sheets?

Yes, Carla HQ enables zero-server in-context tabular prediction directly in spreadsheets using TabICL foundation models.

What machine learning models perform best on the Phoneme dataset?

Non-linear gradient boosting algorithms (such as LightGBM and XGBoost), Random Forests, and modern tabular foundation models like TabICL achieve the strongest benchmark scores, typically reaching 87% to 90% classification accuracy.

LIVE EVALUATION IN GOOGLE SHEETS

Test TabICLv2 on Phoneme Yourself

Open the pre-loaded Google Sheet and let Carla configure the target and task for local, zero-cloud tabular machine learning.

Step 1

Launch Carla

Click once to open the spreadsheet and Carla side panel together.

Step 2

Review the Setup

Carla selects the dataset target and task from this page automatically.

Step 3

Evaluate & Predict

Run predictions and compute metrics with zero server uploads.