The viewer is disabled because this dataset repo requires arbitrary Python code execution. Please consider removing the loading script and relying on automated data support (you can use convert_to_parquet from the datasets library). If this is not possible, please open a discussion for direct help.


EVI is a challenging spoken multilingual dataset with 5,506 dialogues in English, Polish, and French that can be used for benchmarking and developing knowledge-based enrolment, identification, and identification for spoken dialogue systems.


EVI can be downloaded and used as follows:

from datasets import load_dataset
evi = load_dataset("PolyAI/evi", "en-GB") # for British English

# to download data from all locales use:
# evi = load_dataset("PolyAI/evi", "all")

# see structure

Dataset Structure

We show detailed information of the example for the en-GB configuration of the dataset. All other configurations have the same structure.

Data Instances

An example of a data instance of the config en-GB looks as follows:

    "language": 0,
    "dialogue_id": "CA0007220161df7be23f4554704c8720f5",
    "speaker_id": "e80e9bdd33eda593f16a1b6f2fb228ff",
    "turn_id": 0,
    "target_profile_id": "en.GB.608",
    "asr_transcription": "w20 a b",
    "asr_nbest'": ["w20 a b", "w20 a bee", "w20 a baby"],
    "path": "audios/en/CA0007220161df7be23f4554704c8720f5/0.wav",
    "audio": {
        "path": "/home/georgios/.cache/huggingface/datasets/downloads/extracted/0335ebc25feace53243133b49ba17ba18e26f0f97cb083ffdf4e73dd7427b443/audios/en/CA0007220161df7be23f4554704c8720f5/0.wav",
        "array": array([ 0.00024414,  0.00024414,  0.00024414, ...,  0.00024414,
       -0.00024414,  0.00024414], dtype=float32),
       "sampling_rate": 8000,

Data Fields

The data fields are the same among all splits.

  • language (int): ID of language
  • dialogue_id (str): the ID of the dialogue
  • speaker_id (str): the ID of the speaker
  • turn_id (int)": the ID of the turn
  • target_profile_id (str): the ID of the target profile
  • asr_transcription (str): ASR transcription of the audio file
  • asr_nbest (list): n-best ASR transcriptions of the audio file
  • path (str): Path to the audio file
  • audio (dict): Audio object including loaded audio array, sampling rate and path of audio

Data Splits

Every config only has the "test" split containing ca. 1,800 dialogues.

Dataset Creation

More Information Needed

Considerations for Using the Data

Social Impact of Dataset

More Information Needed

Discussion of Biases

More Information Needed

Other Known Limitations

More Information Needed

Additional Information

Dataset Curators

More Information Needed

Licensing Information

All datasets are licensed under the Creative Commons license (CC-BY).

Citation Information

    author      = {Georgios P. Spithourakis and
                   Ivan Vuli\'{c} and
                   Micha\l{} Lis and
                   I\~{n}igo Casanueva
                   and Pawe\l{} Budzianowski},
    title       = {{EVI}: Multilingual Spoken Dialogue Tasks and Dataset for Knowledge-Based Enrolment, Verification, and Identification},
    year        = {2022},
    note        = {Data available at},
    url         = {},
    booktitle   = {Findings of NAACL (publication pending)}


Thanks to @polinaeterna for helping with adding this dataset

Downloads last month