The AMI Meeting Corpus: A Pre-announcement

By J. Carletta, S. Ashby, S. Bourban, Mike Flynn, M. Guillemot, Thomas Hain, J. Kadlec, Vasilis Karaiskos, Wessel Kraaij, Melissa Kronenthal and 7 more. 2005

1,098 citations recorded by the source corpus.

Abstract. The AMI Meeting Corpus is a multi-modal data set consisting of 100 hours of meeting recordings. It is being created in the context of a project that is developing meeting browsing technology and will eventually be released publicly. Some of the meetings it contains are naturally occurring, and some are elicited, particularly using a scenario in which the participants play different roles in a design team, taking a design project from kick-o' to completion over the course of a day. The corpus is being recorded using a wide range of devices including close-talking and far-field microphones, individual and room-view video cameras, projection, a whiteboard, and individual pens, all of which produce output signals that are synchronized with each other. It is also being hand-annotated for many different phenomena, including orthographic transcription, discourse properties such as named entities and dialogue acts, summaries, emotions, and some head and hand gestures. We describe the data set, including the rationale behind using elicited material, and explain how the material is being recorded, transcribed and annotated.

Subfields: Artificial Intelligence, Evaluation & Benchmarks, Speech & Audio

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Evaluation & Benchmarks

In Speech & Audio

Citations in the corpus

BigSSL: Exploring the Frontier of Large-Scale Semi-Supervised Learning for Automatic Speech Recognition (2021, methodology): “Data: The SpeechStew [30] dataset is assembled by putting together seven public speech corpora—AMI [74], Common Voice [75], English Broadcast News,2 LibriSpeech [28], Switchboard/Fisher,3 TED-LIUM v3 [76], [77] and Wall Street Journal.”
SpeechStew: Simply Mix All Available Speech Recognition Data to Train One Large Neural Network (2021, neutral): “AMI [31].”
Ego4D: Around the World in 3,000 Hours of Egocentric Video (2021, background): “[28] Jean Carletta, Simone Ashby, Sebastien Bourban, Mike Flynn, Mael Guillemot, Thomas Hain, Jaroslav Kadlec,”
Residual LSTM: Design of a Deep Recurrent Architecture for Distant Speech Recognition (2017, methodology): “The experimental result on the AMI SDM corpus [23] showed 10-layer plain and highway LSTMs had severe degradation from increased depth: 13.”
Content Selection in Deep Learning Models of Summarization (2018, methodology): “AMI The AMI corpus (Carletta et al., 2005) is a collection of real and staged office meetings annotated with text transcriptions, along with abstractive summaries.”
Deep Learning for Environmentally Robust Speech Recognition (2017, methodology): “The reported results on the AMI corpus by using the proposed end-to-end framework is much higher than the results of traditional DBMs and are competitive to the LSTM-RNN-based AM; and the results on Aurora-4 beat any other published results on this database, even without performing any speech and fe”
Abstractive Summarization of Spoken and Written Conversations Based on Phrasal Queries (2014, methodology): “We evaluate our system over GNUe Trafﬁc archive 2 Internet Relay Chat (IRC) logs, AMI meetings corpus (Carletta et al., 2005) and BC3 emails dataset (Ulrich et al., 2008).”
QMSum: A New Benchmark for Query-based Multi-domain Meeting Summarization (2021, background): “Product Meetings AMI 1 (Carletta et al., 2005) is a dataset of meetings about product design in an industrial setting.”
SpeechBrain: A General-Purpose Speech Toolkit (2021, methodology): “Table 5 reports the performance achieved on speaker diarization with the AMI meeting corpus [38] when using the embeddings available in SpeechBrain.”
Lattice-free State-level Minimum Bayes Risk Training of Acoustic Models (2018, methodology): “Experiments on the AMI meeting corpus [13], the corpus of spontaneous Japanese (CSJ) [14], and LibriSpeech [15] showed ∼ 2% relative improvements over LF-MMI AMs by LF-sMBR.”
Restructuring Conversations using Discourse Relations for Zero-shot Abstractive Dialogue Summarization. (2019, background): “, the AMI (Carletta et al., 2005) and ICSI (Janin et al.”
Spot the conversation: speaker diarisation in the wild (2020, background): “Research into speaker diarisation has largely evolved independently for different domains, with broadcast news [20], telephone speech [21], and meetings [22, 23] being the most popular domains.”
Using neural network front-ends on far field multiple microphones based speech recognition (2014, methodology): “The AMI corpus [23] is used in our experiments due to its size, high quality recordings and multiple levels of annotation with metainformation like speakers’ head and body movement.”
Injecting Text in Self-Supervised Speech Pretraining (2021, background): “The ﬁrst two rows correspond to the three public corpora, LibriSpeech [44], LibriLight [45] and AMI [46].”
Large-Scale Pre-Training of End-to-End Multi-Talker ASR for Meeting Transcription with Single Distant Microphone (2021, methodology): “Meeting transcription with a distant microphone has been widely studied as one of the most challenging problems for automatic speech recognition (ASR) [1, 2, 3].”
Acoustic Modeling for Distant Multi-talker Speech Recognition with Single- and Multi-channel Branches (2019, methodology): “As a second experiment, we evaluated our models by using the AMI Meeting Corpus [42].”
Audio-Visual Speaker Diarization Based on Spatiotemporal Bayesian Fusion (2016, background): “Moreover, participants are seated, or are static, and there are often dedicated close-ﬁeld microphones and cameras for each participant e.g. [8].”