Multimodal Defensive Communication Database (DefComm-DB)

Description: DefComm-DB comprises 261 genuine real-world dialogues between English-speaking individuals in 'real-world' settings that feature one of the defensive behaviours outlined in Birkenbihl's model of communication failures [1]:

Attacking the conversation partner (class Attack): videos that depict individuals actively attacking verbally, blaming the other person, or shifting the other person's attention to themselves.

Withdrawing from the communication (class Flight): videos where people refuse to respond, withdraw from the conversation or change the topic or focus.

Making oneself greater (class Greater): videos that depict individuals boasting, self-justifying in an aggressive manner, denying accusations, exhibiting a sense of dominance or superiority, or expressing indignation.

Making oneself smaller (class Smaller): videos that display individuals engaging in self-deprecation, self-blame, exhibiting a sense of guilt, apologising, and expressing feelings of vulnerability or worthlessness.

[1] Birkenbihl, V. (2013). Kommunikationstraining: Zwischenmenschliche Beziehungen erfolgreich gestalten. Schritte 1–6. : mvg Verlag.

Key statistics on the dataset are provided in Table 1. DefComm-DB features a variety of video topics, including interviews with celebrities and professional athletes, political debates, legal trials, TV shows, and video footage obtained by paparazzi, among others. The situations, number of participants, gender, age, and ethnicity vary from scene to scene.

From each video, we retrieve audio, visual, and textual modalities. In this paper, we focus on the audio modality and the speech transcriptions.

Table 1: Statistics on DefComm-DB: number of video clips, mean duration (μ), standard deviation (σ), minimum, maximum, and total duration of collected videos per class.

Label
# video clips
μ [s]
σ [s]
min [s]
max [s]
Σ duration [s]

Attack
112
8
9
2
46
949

Flight
57
9
8
2
62
494

Greater
45
9
6
2
25
416

Smaller
47
12
8
3
49
556

Total
261
9
8
2
62
2415

If you use DefComm-DB in your research work, you are kindly asked to acknowledge it in your publications.

S. Amiriparian, L. Christ, R. Kushtanova, M. Gerczuk, A. Teynor, and B. Schuller, “Speech-Based Classification of Defensive Communication: A Novel Dataset and Results,” in Proc. INTERSPEECH 2023, Dublin, Ireland, ISCA, 8 2023. 5 pages, to appear.

@inproceedings{Amiriparian23-SCO, 

author = {Shahin Amiriparian and Lukas Christ and Regina Kushtanova and Maurice Gerczuk and Alexandra Teynor and  Björn Schuller}, 
title = {{Speech-Based Classification of Defensive Communication: A Novel Dataset and Results}}, 
booktitle = {{Proceedings INTERSPEECH 2023}}, 
year = {2023}, 
address = {Dublin, Ireland},
publisher = {ISCA},  
month = {8}, 
note = {5 pages, to appear},
}

DefComm-DB 包含261段真实世界的英语使用者对话，场景均为现实生活情境，其中涵盖Birkenbihl沟通失败模型[1]所定义的四类防御性行为：
1. 攻击对话者（Attack）：展示个体主动进行言语攻击、指责他人或将话题焦点转移至自身的视频片段。
2. 退出沟通（Flight）：呈现个体拒绝回应、脱离对话、转换话题或改变沟通焦点的视频片段。
3. 抬高自身（Greater）：包含个体吹嘘、攻击性自我辩解、否认指控、彰显支配欲或优越感，或表达愤慨情绪的视频片段。
4. 贬低自身（Smaller）：展示个体进行自我贬损、自我指责、流露愧疚感、道歉，以及表达脆弱或无价值感的视频片段。

[1] Birkenbihl, V. (2013). *Kommunikationstraining: Zwischenmenschliche Beziehungen erfolgreich gestalten. Schritte 1–6*. mvg Verlag.（德语原书译名为《沟通训练：成功构建人际关系：步骤1至6》，出版社为mvG出版社）

本数据集的关键统计信息详见表1。DefComm-DB涵盖多样的视频主题，包括名人与职业运动员访谈、政治辩论、庭审、电视节目以及狗仔队拍摄的影像素材等。不同场景的情境、参与人数、性别、年龄及种族均存在显著差异。

我们从每段视频中提取音频、视觉与文本三种模态数据。本文聚焦于音频模态及其语音转录文本。

表1：DefComm-DB统计信息：各类别视频片段数量、平均时长（μ）、标准差（σ）、最短时长、最长时长及总时长，具体如下：攻击类（Attack）共112段视频，平均时长8秒，标准差9秒，最短时长2秒，最长时长46秒，总时长949秒；逃避类（Flight）共57段，平均时长9秒，标准差8秒，最短时长2秒，最长时长62秒，总时长494秒；自大型（Greater）共45段，平均时长9秒，标准差6秒，最短时长2秒，最长时长25秒，总时长416秒；自小型（Smaller）共47段，平均时长12秒，标准差8秒，最短时长3秒，最长时长49秒，总时长556秒；数据集总计261段视频，平均时长9秒，标准差8秒，最短时长2秒，最长时长62秒，总时长2415秒。

若您的研究工作中使用DefComm-DB，请在发表成果中予以引用致谢。

引用信息：S. Amiriparian、L. Christ、R. Kushtanova、M. Gerczuk、A. Teynor与B. Schuller，《基于语音的防御性沟通分类：新型数据集与实验结果》，收录于国际语音通信协会年会（INTERSPEECH 2023），爱尔兰都柏林，国际语音通信协会（ISCA），2023年8月，共5页，待刊。

对应的BibTeX引用格式如下：
@inproceedings{Amiriparian23-SCO,
author = {Shahin Amiriparian and Lukas Christ and Regina Kushtanova and Maurice Gerczuk and Alexandra Teynor and Björn Schuller},
title = {{Speech-Based Classification of Defensive Communication: A Novel Dataset and Results}},
booktitle = {{Proceedings INTERSPEECH 2023}},
year = {2023},
address = {Dublin, Ireland},
publisher = {ISCA},
month = {8},
note = {5 pages, to appear},
}