Category: [:en]Seminars[:fr]Séminaires[:]

Time-frequency fading

Speaker: Marina Kreme Data and place: June 9, 2022, at 10:30 – Hybrid Abstract: We are interested in the problem of attenuating time-frequency regions, for example when a disturbance signal is well localized in the time-frequency plane. We approach this problem from the point of view of time-frequency filtering, by formulating the optimization problem in the signal …

Continue reading

On the impact of normalization strategies in unsupervised adversarial domain adaptation for acoustic scene classification

Speaker: Mauricio Michel Olvera Zambrano Data and place: May 19, 2022, at 10:30 – Hybrid Abstract: Acoustic scene classification systems face performance degradation when trained and tested on data recorded by different devices. Unsupervised domain adaptation methods have been studied to reduce the impact of this mismatch. While they do not assume the availability of labels at …

Continue reading

Multimodal speech animation

Speaker: Louis Abel Data and place: May 12, 2022, at 10:30 – Hybrid Abstract: Multimodal speech animation is the next step to speech synthesis, combining visuals with audio allows the creation of embodied conversational agent (ECA) which can convey more information than a classic text-to-speech approach, several works have been done in the team to progress in …

Continue reading

Expanding the training data for neural network based hate speech classification

Speaker:  Ashwin Geet D’Sa Data and place: April 28, 2022, at 10:30 – Hybrid Abstract: The phenomenal increase in internet usage, catering to the dissemination of knowledge and expression, has also led to an increase in online hate speech. Online hate speech is anti-social communicative behavior, which leads to the threat and violence toward an individual or a group. …

Continue reading

Domain adaptation for accented speech

Speaker: Robin San Roman Data and place: March 31, 2022, at 10:30 – Hybrid Abstract: Automatic speech recognition (ASR) systems tend to have a performance drop with accented speech compared to standard speech. This is due to linguistic differences between the different domains of speech. In this talk, I will present a method to adapt CPC models …

Continue reading

Cross-corpora Hate-speech detection with Dynamically Refined Regularization

Speaker: Tulika Bose Data and place: March 24, 2022, at 10:30 – Hybrid Abstract: Hatespeechclassifiersexhibitsubstantial performance degradation when evaluated on datasets different from the source. This is due to learning spurious correlations between words that are not necessarily relevant to hateful language and hate speech labels from the training corpus. Previous work has attempted to mitigate this problem by regularizing specific terms from pre-defined static dictionaries. While this has been demonstrated to improve the generalizability of classifiers, the coverage of …

Continue reading

Cross-corpora Hate-speech detection with Dynamically Refined Regularization

Speaker: Tulika Bose Data and place: March 24, 2022, at 10:30 – Hybrid Abstract: Hate speech classifiers exhibit substantial performance degradation when evaluated on datasets different from the source. This is due to learning spurious correlations between words that are not necessarily relevant to hateful language and hate speech labels from the training corpus. Previous work has attempted to mitigate this problem by regularizing specific terms from pre-defined static dictionaries. While this has been demonstrated to improve the generalizability of classifiers, the coverage of such methods is …

Continue reading

Identification of Multiword Expressions in Tweets for Hate Speech Detection

Speaker: Nicolas Zampieri Data and place: March 17, 2022, at 10:30 – Hybrid Abstract: Multiword expression (MWE) identification in tweets is a complex task due to the complex linguistic nature of MWEs combined with the non-standard language use on social networks. MWE features were shown to be helpful for hate speech detection (HSD). In this article, we …

Continue reading

Low-resource MT: few-shot learning and historical language normalisation

Speaker: Rachel Bawden Data and place: March 3, 2022, at 10:30 – Hybrid Abstract: Huge progress has been seen in machine translation (MT), spurred on by advances in neural architectures. However, challenges still remain. One of the big challenges is dealing with scenarios that are low-resource (where there is little parallel data available). In this talk, I will …

Continue reading

Privacy-Preserving Speech Representation Learning using Vector Quantization

Speaker: Pierre Champion Data and place: February 3, 2022, at 10:00 – Hybrid Abstract: Speech signals are a rich source of speaker-related information, including sensitive attributes like gender, identity, etc. Those sensitive attributes can be extracted and used for malicious purposes like voice spoofing. Despite the inherent sensitivity of speech signals, more and more services, mainly …

Continue reading