The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
eusip/silicone(複製)
データの内容
英語の話し言葉を扱うデータ 10 種を束ねたベンチマーク SILICONE のデータ。日常会話、台本のある場面、共同でのタスク、電話、テレビの対話など、領域の異なる発話が集められ、発話ごとに対話行為や感情などのラベルが付く。1 行が 1 発話で、発話は Utterance、対話の番号は Dialogue_ID に入る。
データの概要
| 項目 | 値 |
|---|---|
| 行数 | 614,723(10 構成・30 分割の合計) |
| 列数 | 5〜9(構成による) |
| 容量 | 49.99 KB(ファイル 3 本) |
| ファイルの形式 | py 1 |
| 言語 | en |
| 公開 | public |
構成と分割
| 構成(config) | 分割(split)ごとの行数 | 列数 |
|---|---|---|
dyda_da |
train 87,170、validation 8,069、test 7,740 |
5 |
dyda_e |
train 87,170、validation 8,069、test 7,740 |
5 |
iemocap |
train 7,213、validation 805、test 2,021 |
6 |
maptask |
train 20,905、validation 2,963、test 2,894 |
5 |
meld_e |
train 9,989、validation 1,109、test 2,610 |
7 |
meld_s |
train 9,989、validation 1,109、test 2,610 |
7 |
mrda |
train 83,943、validation 9,815、test 15,470 |
8 |
oasis |
train 12,076、validation 1,513、test 1,478 |
5 |
sem |
train 4,264、validation 485、test 878 |
8 |
swda |
train 190,709、validation 21,203、test 2,714 |
9 |
列(構成 dyda_da。構成ごとに列が違う)
| 列 | 型 |
|---|---|
Utterance |
string |
Dialogue_Act |
string |
Dialogue_ID |
string |
Label |
struct{class_label: struct{names: struct{…}}} |
Idx |
int32 |
行数と列は元のカードの YAML(dataset_info)の記載。データセットビューアでは読めなかった。全データセットの一覧は HYUKI5/dataset-catalog にある。
複製について
eusip/silicone をリビジョン f9e4b166dced70ccc4ddca59ad5e894b5d822b79 の時点でそのまま複製したもの。中身は変えていない。判断モデル Laya(convaiinnovations/laya)の微調整に使う候補として、調査メモ(Laya 学習用データセット調査)に載せたデータを集めている。
出典
- 元のデータセット: https://huggingface.co/datasets/eusip/silicone
- 複製したリビジョン:
f9e4b166dced70ccc4ddca59ad5e894b5d822b79(2024-01-18 11:15:53+00:00) - 元のライセンス表記:
['cc-by-sa-4.0']
論文
アノテーションの方法
ベンチマークの作成者
Emile Chapuis, Pierre Colombo, Ebenge Usip.
微調整に使うときの注意
- 調査メモの台帳での区分: C(NC・ND・研究専用・評価専用など、商用の利用を妨げうる記載がある)。区分は記載の文言で機械的に付けたもので、法的な判定ではない。学習に入れる前に元のカードとライセンスの原文で確かめる。
- Laya の入力は 2,048 token まで(ModernBERT-large / mmBERT)。長い本文は先頭から黙って切られるので、切れる行は事前に除くか、要約・切り出しをしてから入れる。
- 分割は記事・文書単位で行い、同じ本文から作った別の問いを学習側と較正側に分けない。温度と閾値は学習に使っていない分割で別々に当てる。
調査メモの記述
| メモ | 節 | ライセンス(記載) | 確認 | 注記 |
|---|---|---|---|---|
jev_sources_tasksource_bekko.csv |
CSV(tasksource-jev-typed-decisions) | cc-by-sa-4.0, CC BY-NC-SA 4.0 (DPI) / license_use=non-commercial | 確認済 | 複数/混在 / [Jev 元データ source] source_annotation_converted(deterministic recast; no model r |
jev_sources_tasksource_bekko.csv |
CSV(tasksource-jev-typed-decisions) | cc-by-sa-4.0, CC BY 4.0 (DPI) / license_use=commercial | 確認済 | 複数/混在/license_use=commercial / [Jev 元データ source] source_annotation_converted(deterministic recast; no model r |
jev_sources_tasksource_bekko.csv |
CSV(tasksource-jev-typed-decisions) | cc-by-sa-4.0, CC BY-NC-SA 4.0 (DPI) / license_use=non-commercial | 確認済 | 複数/混在 / [Jev 元データ source] source_annotation_converted(deterministic recast; no model r |
jev_sources_tasksource_bekko.csv |
CSV(tasksource-jev-typed-decisions) | cc-by-sa-4.0, CC BY-NC-SA 4.0 (DPI) / license_use=non-commercial | 確認済 | 複数/混在 / [Jev 元データ source] source_annotation_converted(deterministic recast; no model r |
元のカード(日本語訳)
以下は元のデータセットのカードの本文を日本語に訳したもの(機械翻訳)。訳に疑問があるときは、末尾の原文を確かめる。
SILICONE ベンチマークのデータセットカード
目次
- データセットの説明
- データセットの概要
- 対応するタスクとリーダーボード
- 言語
- データセットの構造
- データの例
- データのフィールド
- データの分割
- データセットの作成
- 作成の根拠
- 元データ
- アノテーション
- 個人情報と機微な情報
- データの利用上の考慮事項
- データセットの社会的影響
- 偏りについての議論
- その他の既知の制限
- 追加情報
- データセットの作成者
- ライセンス情報
- 引用情報
- 貢献
データセットの説明
- ホームページ: [N/A]
- リポジトリ: https://github.com/eusip/SILICONE-benchmark
- 論文: https://arxiv.org/abs/2009.11152
- リーダーボード: [N/A]
- 連絡先: Ebenge Usip
データセットの概要
SILICONE(Sequence labellIng evaLuatIon benChmark fOr spoken laNguagE)ベンチマークは、話し言葉に特化して設計された自然言語理解システムの学習・評価・分析のための資源集である。データセットはすべて英語で、日常会話、台本のあるシナリオ、共同でのタスク遂行、電話での会話、テレビの対話など、さまざまな領域を扱う。一部のデータセットは感情やセンチメントのラベルも含む。
対応するタスクとリーダーボード
[More Information Needed]
言語
英語。
データセットの構造
データの例
DailyDialog Act Corpus(Dialogue Act)
dyda_da 設定の場合、データセットからの 1 件の例は次のとおりである。
{
'Utterance': "the taxi drivers are on strike again .",
'Dialogue_Act': 2, # "inform"
'Dialogue_ID': "2"
}
DailyDialog Act Corpus(Emotion)
dyda_e 設定の場合、データセットからの 1 件の例は次のとおりである。
{
'Utterance': "'oh , breaktime flies .'",
'Emotion': 5, # "sadness"
'Dialogue_ID': "997"
}
Interactive Emotional Dyadic Motion Capture (IEMOCAP) データベース
iemocap 設定の場合、データセットからの 1 件の例は次のとおりである。
{
'Dialogue_ID': "Ses04F_script03_2",
'Utterance_ID': "Ses04F_script03_2_F025",
'Utterance': "You're quite insufferable. I expect it's because you're drunk.",
'Emotion': 0, # "ang"
}
HCRC MapTask Corpus
maptask 設定の場合、データセットからの 1 件の例は次のとおりである。
{
'Speaker': "f",
'Utterance': "i think that would bring me over the crevasse",
'Dialogue_Act': 4, # "explain"
}
Multimodal EmotionLines Dataset(Emotion)
meld_e 設定の場合、データセットからの 1 件の例は次のとおりである。
{
'Utterance': "'Push 'em out , push 'em out , harder , harder .'",
'Speaker': "Joey",
'Emotion': 3, # "joy"
'Dialogue_ID': "1",
'Utterance_ID': "2"
}
Multimodal EmotionLines Dataset(Sentiment)
meld_s 設定の場合、データセットからの 1 件の例は次のとおりである。
{
'Utterance': "'Okay , y'know what ? There is no more left , left !'",
'Speaker': "Rachel",
'Sentiment': 0, # "negative"
'Dialogue_ID': "2",
'Utterance_ID': "4"
}
ICSI MRDA Corpus
mrda 設定の場合、データセットからの 1 件の例は次のとおりである。
{
'Utterance_ID': "Bed006-c2_0073656_0076706",
'Dialogue_Act': 0, # "s"
'Channel_ID': "Bed006-c2",
'Speaker': "mn015",
'Dialogue_ID': "Bed006",
'Utterance': "keith is not technically one of us yet ."
}
BT OASIS Corpus
oasis 設定の場合、データセットからの 1 件の例は次のとおりである。
{
'Speaker': "b",
'Utterance': "when i rang up um when i rang to find out why she said oh well your card's been declined",
'Dialogue_Act': 21, # "inform"
}
SEMAINE データベース
sem 設定の場合、データセットからの 1 件の例は次のとおりである。
{
'Utterance': "can you think of somebody who is like that ?",
'NbPairInSession': "11",
'Dialogue_ID': "59",
'SpeechTurn': "674",
'Speaker': "Agent",
'Sentiment': 1, # "Neutral"
}
Switchboard Dialog Act (SwDA) Corpus
swda 設定の場合、データセットからの 1 件の例は次のとおりである。
{
'Utterance': "but i 'd probably say that 's roughly right .",
'Dialogue_Act': 33, # "aap_am"
'From_Caller': "1255",
'To_Caller': "1087",
'Topic': "CRIME",
'Dialogue_ID': "818",
'Conv_ID': "sw2836",
}
データのフィールド
dyda_da 設定では、フィールドは次のとおり。
Utterance: 文字列としての発話。Dialogue_Act: 発話の対話行為ラベル。"commissive"(0)、"directive"(1)、"inform"(2)、"question"(3)のいずれか。Dialogue_ID: 対話の識別子(文字列)。
dyda_e 設定では、フィールドは次のとおり。
Utterance: 文字列としての発話。Dialogue_Act: 発話の対話行為ラベル。"anger"(0)、"disgust"(1)、"fear"(2)、"happiness"(3)、"no emotion"(4)、"sadness"(5)、"surprise"(6)のいずれか。Dialogue_ID: 対話の識別子(文字列)。
iemocap 設定では、フィールドは次のとおり。
Dialogue_ID: 対話の識別子(文字列)。Utterance_ID: 発話の識別子(文字列)。Utterance: 文字列としての発話。Emotion: 発話の感情ラベル。"Anger"(0)、"Disgust"(1)、"Excitement"(2)、"Fear"(3)、"Frustration"(4)、"Happiness"(5)、"Neutral"(6)、"Other"(7)、"Sadness"(8)、"Surprise"(9)、"Unknown"(10)のいずれか。
maptask 設定では、フィールドは次のとおり。
Speaker: 話者の識別子(文字列)。Utterance: 文字列としての発話。Dialogue_Act: 発話の対話行為ラベル。"acknowledge"(0)、"align"(1)、"check"(2)、"clarify"(3)、"explain"(4)、"instruct"(5)、"query_w"(6)、"query_yn"(7)、"ready"(8)、"reply_n"(9)、"reply_w"(10)、"reply_y"(11)のいずれか。
meld_e 設定では、フィールドは次のとおり。
Utterance: 文字列としての発話。Speaker: 文字列としての話者。Emotion: 発話の感情ラベル。"anger"(0)、"disgust"(1)、"fear"(2)、"joy"(3)、"neutral"(4)、"sadness"(5)、"surprise"(6)のいずれか。Dialogue_ID: 対話の識別子(文字列)。Utterance_ID: 発話の識別子(文字列)。
meld_s 設定では、フィールドは次のとおり。
Utterance: 文字列としての発話。Speaker: 文字列としての話者。Sentiment: 発話のセンチメントラベル。"negative"(0)、"neutral"(1)、"positive"(2)のいずれか。Dialogue_ID: 対話の識別子(文字列)。Utterance_ID: 発話の識別子(文字列)。
mrda 設定では、フィールドは次のとおり。
Utterance_ID: 発話の識別子(文字列)。Dialogue_Act: 発話の対話行為ラベル。"s"(0)[平叙・主観的陳述]、"d"(1)[平叙の疑問文]、"b"(2)[相槌(バックチャンネル)]、"f"(3)[Follow-me]、"q"(4)[質問] のいずれか。Channel_ID: チャンネルの識別子(文字列)。Speaker: 話者の識別子(文字列)。Dialogue_ID: チャンネルの識別子(文字列)。Utterance: 文字列としての発話。
oasis 設定では、フィールドは次のとおり。
Speaker: 話者の識別子(文字列)。Utterance: 文字列としての発話。Dialogue_Act: 発話の対話行為ラベル。"accept"(0)、"ackn"(1)、"answ"(2)、"answElab"(3)、"appreciate"(4)、"backch"(5)、"bye"(6)、"complete"(7)、"confirm"(8)、"correct"(9)、"direct"(10)、"directElab"(11)、"echo"(12)、"exclaim"(13)、"expressOpinion"(14)、"expressPossibility"(15)、"expressRegret"(16)、"expressWish"(17)、"greet"(18)、"hold"(19)、"identifySelf"(20)、"inform"(21)、"informCont"(22)、"informDisc"(23)、"informIntent"(24)、"init"(25)、"negate"(26)、"offer"(27)、"pardon"(28)、"raiseIssue"(29)、"refer"(30)、"refuse"(31)、"reqDirect"(32)、"reqInfo"(33)、"reqModal"(34)、"selfTalk"(35)、"suggest"(36)、"thank"(37)、"informIntent-hold"(38)、"correctSelf"(39)、"expressRegret-inform"(40)、または "thank-identifySelf"(41)のいずれか。
sem 設定では、フィールドは次のとおり。
Utterance: 文字列としての発話。NbPairInSession: 対話内の発話ペアの数。Dialogue_ID: 対話の識別子(文字列)。SpeechTurn: 話者ターン(文字列)。Speaker: 話者(文字列)。Sentiment: 発話のセンチメントラベル。"Negative"、"Neutral"、"Positive" のいずれか。
swda 設定では、フィールドは次のとおり。
Utterance: 文字列としての発話。
Dialogue_Act: 発話の対話行為ラベル。"sd"(0)[平叙・非意見]、"b"(1)[相槌(Acknowledge)]、"sv"(2)[平叙・意見]、"%"(3)[解釈不能]、"aa"(4)[同意・受諾]、"ba"(5)[称賛]、"fc"(6)[定型の結び]、"qw"(7)[wh 疑問文]、"nn"(8)[否定の応答]、"bk"(9)[応答の受領]、"h"(10)[ヘッジ]、"qy^d"(11)[平叙の Yes-No 疑問文]、"bh"(12)[疑問形の相槌]、"^q"(13)[引用]、"bf"(14)[要約・言い換え]、'fo_o_fw_"by_bc'(15)[その他]、'fo_o_fw_by_bc"'(16)[その他]、"na"(17)[肯定の非 yes 応答]、"ad"(18)[行為指示]、"^2"(19)[共同補完]、"b^m"(20)[句の繰り返し]、"qo"(21)[開かれた質問]、"qh"(22)[修辞疑問]、"^h"(23)[応答・同意の前の保留]、"ar"(24)[拒否]、"ng"(25)[否定の非 no 応答]、"br"(26)[非理解の表明]、"no"(27)[その他の応答]、"fp"(28)[定型の冒頭]、"qrr"(29)[or 節]、"arp_nd"(30)[非優先の応答]、"t3"(31)[第三者への言及]、"oo_co_cc"(32)[申し出・選択肢・コミット]、"aap_am"(33)[保留・部分的受諾]、"t1"(34)[過小化]、"bd"(35)[独り言]、"^g"(36)[付加疑問文]、"qw^d"(37)[平叙の wh 疑問文]、"fa"(38)[謝罪]、"ft"(39)[感謝]、"+"(40)[不明]、"x"(41)[不明]、"ny"(42)[不明]、"sv_fx"(43)[不明]、"qy_qr"(44)[不明]、または "ba_fe"(45)[不明] のいずれか。
From_Caller: 発信者の識別子(文字列)。
To_Caller: 着信者の識別子(文字列)。
Topic: 話題(文字列)。
Dialogue_ID: 対話の識別子(文字列)。
Conv_ID: 会話の識別子(文字列)。
データの分割
| Dataset name | Train | Valid | Test |
|---|---|---|---|
| dyda_da | 87170 | 8069 | 7740 |
| dyda_e | 87170 | 8069 | 7740 |
| iemocap | 7213 | 805 | 2021 |
| maptask | 20905 | 2963 | 2894 |
| meld_e | 9989 | 1109 | 2610 |
| meld_s | 9989 | 1109 | 2610 |
| mrda | 83944 | 9815 | 15470 |
| oasis | 12076 | 1513 | 1478 |
| sem | 4264 | 485 | 878 |
| swda | 190709 | 21203 | 2714 |
データセットの作成
作成の根拠
[More Information Needed]
元データ
初期のデータ収集と正規化
[More Information Needed]
元の言語の生成者は誰か?
[More Information Needed]
アノテーション
アノテーションの過程
[More Information Needed]
アノテーターは誰か?
[More Information Needed]
個人情報と機微な情報
[More Information Needed]
データの利用上の考慮事項
データセットの社会的影響
[More Information Needed]
偏りについての議論
[More Information Needed]
その他の既知の制限
[More Information Needed]
追加情報
ベンチマークの作成者
Emile Chapuis, Pierre Colombo, Ebenge Usip.
ライセンス情報
この作品は Creative Commons Attribution-NonCommercial-ShareAlike 4.0 Unported License のもとでライセンスされている。
引用情報
@inproceedings{chapuis-etal-2020-hierarchical,
title = "Hierarchical Pre-training for Sequence Labelling in Spoken Dialog",
author = "Chapuis, Emile and
Colombo, Pierre and
Manica, Matteo and
Labeau, Matthieu and
Clavel, Chlo{\'e}",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2020",
month = nov,
year = "2020",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://www.aclweb.org/anthology/2020.findings-emnlp.239",
doi = "10.18653/v1/2020.findings-emnlp.239",
pages = "2636--2648",
abstract = "Sequence labelling tasks like Dialog Act and Emotion/Sentiment identification are a key component of spoken dialog systems. In this work, we propose a new approach to learn generic representations adapted to spoken dialog, which we evaluate on a new benchmark we call Sequence labellIng evaLuatIon benChmark fOr spoken laNguagE benchmark (SILICONE). SILICONE is model-agnostic and contains 10 different datasets of various sizes. We obtain our representations with a hierarchical encoder based on transformer architectures, for which we extend two well-known pre-training objectives. Pre-training is performed on OpenSubtitles: a large corpus of spoken dialog containing over 2.3 billion of tokens. We demonstrate how hierarchical encoders achieve competitive results with consistently fewer parameters compared to state-of-the-art models and we show their importance for both pre-training and fine-tuning.",
}
貢献
このデータセットを追加してくれた @eusip と @lhoestq に感謝する。
原文(英語)
Dataset Card for SILICONE Benchmark
Dataset Summary
The Sequence labellIng evaLuatIon benChmark fOr spoken laNguagE (SILICONE) benchmark is a collection of resources for training, evaluating, and analyzing natural language understanding systems specifically designed for spoken language. All datasets are in the English language and covers a variety of domains including daily life, scripted scenarios, joint task completion, phone call conversations, and televsion dialogue. Some datasets additionally include emotion and/or sentimant labels.
Supported Tasks and Leaderboards
[More Information Needed]
Languages
English.
Dataset Structure
Data Instances
DailyDialog Act Corpus (Dialogue Act)
For the dyda_da configuration one example from the dataset is:
{
'Utterance': "the taxi drivers are on strike again .",
'Dialogue_Act': 2, # "inform"
'Dialogue_ID': "2"
}
DailyDialog Act Corpus (Emotion)
For the dyda_e configuration one example from the dataset is:
{
'Utterance': "'oh , breaktime flies .'",
'Emotion': 5, # "sadness"
'Dialogue_ID': "997"
}
Interactive Emotional Dyadic Motion Capture (IEMOCAP) database
For the iemocap configuration one example from the dataset is:
{
'Dialogue_ID': "Ses04F_script03_2",
'Utterance_ID': "Ses04F_script03_2_F025",
'Utterance': "You're quite insufferable. I expect it's because you're drunk.",
'Emotion': 0, # "ang"
}
HCRC MapTask Corpus
For the maptask configuration one example from the dataset is:
{
'Speaker': "f",
'Utterance': "i think that would bring me over the crevasse",
'Dialogue_Act': 4, # "explain"
}
Multimodal EmotionLines Dataset (Emotion)
For the meld_e configuration one example from the dataset is:
{
'Utterance': "'Push 'em out , push 'em out , harder , harder .'",
'Speaker': "Joey",
'Emotion': 3, # "joy"
'Dialogue_ID': "1",
'Utterance_ID': "2"
}
Multimodal EmotionLines Dataset (Sentiment)
For the meld_s configuration one example from the dataset is:
{
'Utterance': "'Okay , y'know what ? There is no more left , left !'",
'Speaker': "Rachel",
'Sentiment': 0, # "negative"
'Dialogue_ID': "2",
'Utterance_ID': "4"
}
ICSI MRDA Corpus
For the mrda configuration one example from the dataset is:
{
'Utterance_ID': "Bed006-c2_0073656_0076706",
'Dialogue_Act': 0, # "s"
'Channel_ID': "Bed006-c2",
'Speaker': "mn015",
'Dialogue_ID': "Bed006",
'Utterance': "keith is not technically one of us yet ."
}
BT OASIS Corpus
For the oasis configuration one example from the dataset is:
{
'Speaker': "b",
'Utterance': "when i rang up um when i rang to find out why she said oh well your card's been declined",
'Dialogue_Act': 21, # "inform"
}
SEMAINE database
For the sem configuration one example from the dataset is:
{
'Utterance': "can you think of somebody who is like that ?",
'NbPairInSession': "11",
'Dialogue_ID': "59",
'SpeechTurn': "674",
'Speaker': "Agent",
'Sentiment': 1, # "Neutral"
}
Switchboard Dialog Act (SwDA) Corpus
For the swda configuration one example from the dataset is:
{
'Utterance': "but i 'd probably say that 's roughly right .",
'Dialogue_Act': 33, # "aap_am"
'From_Caller': "1255",
'To_Caller': "1087",
'Topic': "CRIME",
'Dialogue_ID': "818",
'Conv_ID': "sw2836",
}
Data Fields
For the dyda_da configuration, the different fields are:
Utterance: Utterance as a string.Dialogue_Act: Dialog act label of the utterance. It can be one of "commissive" (0), "directive" (1), "inform" (2) or "question" (3).Dialogue_ID: identifier of the dialogue as a string.
For the dyda_e configuration, the different fields are:
Utterance: Utterance as a string.Dialogue_Act: Dialog act label of the utterance. It can be one of "anger" (0), "disgust" (1), "fear" (2), "happiness" (3), "no emotion" (4), "sadness" (5) or "surprise" (6).Dialogue_ID: identifier of the dialogue as a string.
For the iemocap configuration, the different fields are:
Dialogue_ID: identifier of the dialogue as a string.Utterance_ID: identifier of the utterance as a string.Utterance: Utterance as a string.Emotion: Emotion label of the utterance. It can be one of "Anger" (0), "Disgust" (1), "Excitement" (2), "Fear" (3), "Frustration" (4), "Happiness" (5), "Neutral" (6), "Other" (7), "Sadness" (8), "Surprise" (9) or "Unknown" (10).
For the maptask configuration, the different fields are:
Speaker: identifier of the speaker as a string.Utterance: Utterance as a string.Dialogue_Act: Dialog act label of the utterance. It can be one of "acknowledge" (0), "align" (1), "check" (2), "clarify" (3), "explain" (4), "instruct" (5), "query_w" (6), "query_yn" (7), "ready" (8), "reply_n" (9), "reply_w" (10) or "reply_y" (11).
For the meld_e configuration, the different fields are:
Utterance: Utterance as a string.Speaker: Speaker as a string.Emotion: Emotion label of the utterance. It can be one of "anger" (0), "disgust" (1), "fear" (2), "joy" (3), "neutral" (4), "sadness" (5) or "surprise" (6).Dialogue_ID: identifier of the dialogue as a string.Utterance_ID: identifier of the utterance as a string.
For the meld_s configuration, the different fields are:
Utterance: Utterance as a string.Speaker: Speaker as a string.Sentiment: Sentiment label of the utterance. It can be one of "negative" (0), "neutral" (1) or "positive" (2).Dialogue_ID: identifier of the dialogue as a string.Utterance_ID: identifier of the utterance as a string.
For the mrda configuration, the different fields are:
Utterance_ID: identifier of the utterance as a string.Dialogue_Act: Dialog act label of the utterance. It can be one of "s" (0) [Statement/Subjective Statement], "d" (1) [Declarative Question], "b" (2) [Backchannel], "f" (3) [Follow-me] or "q" (4) [Question].Channel_ID: identifier of the channel as a string.Speaker: identifier of the speaker as a string.Dialogue_ID: identifier of the channel as a string.Utterance: Utterance as a string.
For the oasis configuration, the different fields are:
Speaker: identifier of the speaker as a string.Utterance: Utterance as a string.Dialogue_Act: Dialog act label of the utterance. It can be one of "accept" (0), "ackn" (1), "answ" (2), "answElab" (3), "appreciate" (4), "backch" (5), "bye" (6), "complete" (7), "confirm" (8), "correct" (9), "direct" (10), "directElab" (11), "echo" (12), "exclaim" (13), "expressOpinion"(14), "expressPossibility" (15), "expressRegret" (16), "expressWish" (17), "greet" (18), "hold" (19), "identifySelf" (20), "inform" (21), "informCont" (22), "informDisc" (23), "informIntent" (24), "init" (25), "negate" (26), "offer" (27), "pardon" (28), "raiseIssue" (29), "refer" (30), "refuse" (31), "reqDirect" (32), "reqInfo" (33), "reqModal" (34), "selfTalk" (35), "suggest" (36), "thank" (37), "informIntent-hold" (38), "correctSelf" (39), "expressRegret-inform" (40) or "thank-identifySelf" (41).
For the sem configuration, the different fields are:
Utterance: Utterance as a string.NbPairInSession: number of utterance pairs in a dialogue.Dialogue_ID: identifier of the dialogue as a string.SpeechTurn: SpeakerTurn as a string.Speaker: Speaker as a string.Sentiment: Sentiment label of the utterance. It can be "Negative", "Neutral" or "Positive".
For the swda configuration, the different fields are:
Utterance: Utterance as a string.
Dialogue_Act: Dialogue act label of the utterance. It can be "sd" (0) [Statement-non-opinion], "b" (1) [Acknowledge (Backchannel)], "sv" (2) [Statement-opinion], "%" (3) [Uninterpretable], "aa" (4) [Agree/Accept], "ba" (5) [Appreciation], "fc" (6) [Conventional-closing], "qw" (7) [Wh-Question], "nn" (8) [No Answers], "bk" (9) [Response Acknowledgement], "h" (10) [Hedge], "qy^d" (11) [Declarative Yes-No-Question], "bh" (12) [Backchannel in Question Form], "^q" (13) [Quotation], "bf" (14) [Summarize/Reformulate], 'fo_o_fw_"by_bc' (15) [Other], 'fo_o_fw_by_bc"' (16) [Other], "na" (17) [Affirmative Non-yes Answers], "ad" (18) [Action-directive], "^2" (19) [Collaborative Completion], "b^m" (20) [Repeat-phrase], "qo" (21) [Open-Question], "qh" (22) [Rhetorical-Question], "^h" (23) [Hold Before Answer/Agreement], "ar" (24) [Reject], "ng" (25) [Negative Non-no Answers], "br" (26) [Signal-non-understanding], "no" (27) [Other Answers], "fp" (28) [Conventional-opening], "qrr" (29) [Or-Clause], "arp_nd" (30) [Dispreferred Answers], "t3" (31) [3rd-party-talk], "oo_co_cc" (32) [Offers, Options Commits], "aap_am" (33) [Maybe/Accept-part], "t1" (34) [Downplayer], "bd" (35) [Self-talk], "^g" (36) [Tag-Question], "qw^d" (37) [Declarative Wh-Question], "fa" (38) [Apology], "ft" (39) [Thanking], "+" (40) [Unknown], "x" (41) [Unknown], "ny" (42) [Unknown], "sv_fx" (43) [Unknown], "qy_qr" (44) [Unknown] or "ba_fe" (45) [Unknown].
From_Caller: identifier of the from caller as a string.
To_Caller: identifier of the to caller as a string.
Topic: Topic as a string.
Dialogue_ID: identifier of the dialogue as a string.
Conv_ID: identifier of the conversation as a string.
Data Splits
| Dataset name | Train | Valid | Test |
|---|---|---|---|
| dyda_da | 87170 | 8069 | 7740 |
| dyda_e | 87170 | 8069 | 7740 |
| iemocap | 7213 | 805 | 2021 |
| maptask | 20905 | 2963 | 2894 |
| meld_e | 9989 | 1109 | 2610 |
| meld_s | 9989 | 1109 | 2610 |
| mrda | 83944 | 9815 | 15470 |
| oasis | 12076 | 1513 | 1478 |
| sem | 4264 | 485 | 878 |
| swda | 190709 | 21203 | 2714 |
Dataset Creation
Curation Rationale
[More Information Needed]
Source Data
Initial Data Collection and Normalization
[More Information Needed]
Who are the source language producers?
[More Information Needed]
Annotations
Annotation process
[More Information Needed]
Who are the annotators?
[More Information Needed]
Personal and Sensitive Information
[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
Benchmark Curators
Emile Chapuis, Pierre Colombo, Ebenge Usip.
Licensing Information
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 Unported License.
Citation Information
@inproceedings{chapuis-etal-2020-hierarchical,
title = "Hierarchical Pre-training for Sequence Labelling in Spoken Dialog",
author = "Chapuis, Emile and
Colombo, Pierre and
Manica, Matteo and
Labeau, Matthieu and
Clavel, Chlo{\'e}",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2020",
month = nov,
year = "2020",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://www.aclweb.org/anthology/2020.findings-emnlp.239",
doi = "10.18653/v1/2020.findings-emnlp.239",
pages = "2636--2648",
abstract = "Sequence labelling tasks like Dialog Act and Emotion/Sentiment identification are a key component of spoken dialog systems. In this work, we propose a new approach to learn generic representations adapted to spoken dialog, which we evaluate on a new benchmark we call Sequence labellIng evaLuatIon benChmark fOr spoken laNguagE benchmark (SILICONE). SILICONE is model-agnostic and contains 10 different datasets of various sizes. We obtain our representations with a hierarchical encoder based on transformer architectures, for which we extend two well-known pre-training objectives. Pre-training is performed on OpenSubtitles: a large corpus of spoken dialog containing over 2.3 billion of tokens. We demonstrate how hierarchical encoders achieve competitive results with consistently fewer parameters compared to state-of-the-art models and we show their importance for both pre-training and fine-tuning.",
}
Contributions
- Downloads last month
- 57