Skip to Main content Skip to Navigation
Preprints, Working Papers, ...

Hierarchical-Task Reservoir for Online Semantic Analysis from Continuous Speech

Luca Pedrelli 1 Xavier Hinaut 1
1 Mnemosyne - Mnemonic Synergy
LaBRI - Laboratoire Bordelais de Recherche en Informatique, Inria Bordeaux - Sud-Ouest, IMN - Institut des Maladies Neurodégénératives [Bordeaux]
Abstract : In this paper, we propose a novel architecture called Hierarchical-Task Reservoir (HTR) suitable for real-time applications for which different levels of abstraction are available. We apply it to semantic role labeling based on continuous speech recognition. Taking inspiration from the brain, that demonstrates hierarchies of representations from perceptive to integrative areas, we consider a hierarchy of four sub-tasks with increasing levels of abstraction (phone, word, part-of-speech and semantic role tags). These tasks are progressively learned by the layers of the HTR architecture. Interestingly, quantitative and qualitative results show that the hierarchical-task approach provides an advantage to improve the prediction. In particular, the qualitative results show that a shallow or a hierarchical reservoir, considered as baselines, do not produce estimations as good as the HTR model would. Moreover, we show that it is possible to further improve the accuracy of the model by designing skip connections and by considering word embedding in the internal representations. Overall, the HTR outperformed the other stateof-the-art reservoir-based approaches and it resulted in extremely efficient w.r.t. typical RNNs in deep learning (e.g. LSTMs). The HTR architecture is proposed as a step toward the modeling of online and hierarchical processes at work in the brain during language comprehension.
Complete list of metadata

https://hal.inria.fr/hal-03031413
Contributor : Xavier Hinaut <>
Submitted on : Tuesday, April 20, 2021 - 4:26:17 PM
Last modification on : Wednesday, April 21, 2021 - 3:33:32 AM

File

PedrelliHinaut2020_preprint_HA...
Files produced by the author(s)

Identifiers

  • HAL Id : hal-03031413, version 2

Collections

Citation

Luca Pedrelli, Xavier Hinaut. Hierarchical-Task Reservoir for Online Semantic Analysis from Continuous Speech. 2021. ⟨hal-03031413v2⟩

Share

Metrics

Record views

15

Files downloads

90