Speaker role recognition using question detection and characterization

Bazillon, Thierry; Maza, Benjamin; Rouvier, Michael; Béchet, Frédéric; Nasr, Alexis

doi:10.21437/interspeech.2011-442

Cited by 3 publications

(3 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The lexical information was initially captured by n-gram features which were passed to boosting algorithms or maximum entropy classifiers [36]. Additional language-based features that have been proposed include the types of questions posed by different speakers [41], as well as the errors identified in ASR transcripts [42]. Deep learning approaches have been explored in [39] where SRR relies on word embeddings and Convolutional Neural Networks (CNNs).…”

Section: Speaker Role Recognitionmentioning

confidence: 99%

Linguistically Aided Speaker Diarization Using Speaker Role Information

Flemotomos,

Georgiou,

Narayanan

2019

Preprint

View full text Add to dashboard Cite

Speaker diarization relies on the assumption that speech segments corresponding to a particular speaker are concentrated in a specific region of the speaker space; a region which represents that speaker's identity. These identities are not known a priori, so a clustering algorithm is typically employed, which is traditionally based solely on audio. Under noisy conditions, however, such an approach poses the risk of generating unreliable speaker clusters. In this work we aim to utilize linguistic information as a supplemental modality to identify the various speakers in a more robust way. We are focused on conversational scenarios where the speakers assume distinct roles and are expected to follow different linguistic patterns. This distinct linguistic variability can be exploited to help us construct the speaker identities. That way, we are able to boost the diarization performance by converting the clustering task to a classification one. The proposed method is applied in real-world dyadic psychotherapy interactions between a provider and a patient and demonstrated to show improved results.

show abstract

Section: Speaker Role Recognitionmentioning

confidence: 99%

Linguistically Aided Speaker Diarization Using Speaker Role Information

Flemotomos,

Georgiou,

Narayanan

2019

Preprint

View full text Add to dashboard Cite

show abstract

“…Unfortunately, a speech recognition system is likely to fail achieving two goals at a same time: (1) extract text sequences from the input utterances, (2) detect questions. We can think of a question detection system that works independently and unburdens the load of the speech recognition system [15,10,4,14,3]. Later, an annotation of being a question can form a set with the output of the speech recognition system and handed over to the machine translation system.…”

Section: Introductionmentioning

confidence: 99%

“…Previous studies have focused on using hand-designed features and classifiers such as support vector machines (SVMs) [10,14] or tree-based classifiers [15,4,3]. The classifiers used in these systems are shallow and simple, but there are considerable efforts on designing features based on domain knowledge.…”

Section: Introductionmentioning

confidence: 99%

Detecting Interrogative Utterances with Recurrent Neural Networks

Chung,

Devlin,

Awadalla

2015

Preprint

View full text Add to dashboard Cite

In this paper, we explore different neural network architectures that can predict if a speaker of a given utterance is asking a question or making a statement. We compare the outcomes of regularization methods that are popularly used to train deep neural networks and study how different context functions can affect the classification performance. We also compare the efficacy of gated activation functions that are favorably used in recurrent neural networks and study how to combine multimodal inputs. We evaluate our models on two multimodal datasets: MSR-Skype and CALLHOME.

show abstract