Online adaptative zero-shot learning spoken language understanding using word-embedding

Ferreira, Emmanuel; Jabaian, Bassam; Lefèvre, Fabrice

doi:10.1109/icassp.2015.7178987

Cited by 16 publications

(26 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The main objective of the work presented in this paper is to extend the initial online adaptation strategy (described in [21]) to also allow concept creation and thus domain extension. In this preliminary study, we adopt a simple strategy based on an Adversarial Bandit algorithm to address the model refinement problem.…”

Section: Online Interactive Refinement Problemmentioning

confidence: 99%

“…According to [21], these user feedbacks are converted into a set U of m tuples U := ((c l , T l , f l )) 1≤l≤m , where (c l , T l ) is a chunk/tag pair proposed to the user and f l is her feedback (1 positive, 0 negative). Given K and U after each interaction, the algorithm presented in [21] is used to update K into K .…”

Section: Static Casementioning

confidence: 99%

“…Indeed, these values, which are in range [0, 1], denote how confident the model is about the fact that a chunk is related to a semantic tag (the greater the more confident the model is). According to [21], these coefficients, with initial values set to 1 for the first set of chunk/DA pairs extracted from the ontology, could be re-evaluated afterwards in light of users positive/negative feedbacks and estimated for newly introduced chunks during the online process. Then, the chunks in K are projected into F , as the sum of their composing words D-dimensional vector representations (embeddings).…”

Section: Introductionmentioning

confidence: 99%

“…Then, the experimental study is presented in Section 4 followed by some concluding remarks and perspectives. The SLU model employed in our study is the ZSSP zero-shot learning model proposed in [21,22]. The latter, depicted in Figure 1, makes use of three main components.…”

Section: Introductionmentioning

confidence: 99%

“…In [21,22] a zero-shot learning method for SLU, the Zero-Shot Semantic Parser (ZSSP), based on word embeddings (word2vec [23]) is applied to generalize an initial knowledge base about the task at hand (e.g. database entries, partial ontological description).…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Adversarial Bandit for online interactive active learning of zero-shot spoken language understanding

Ferreira

Masson

Jabaian

et al. 2016

2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

Many state-of-the-art solutions for the understanding of speech data have in common to be probabilistic and to rely on machine learning algorithms to train their models from large amount of data. The difficulty remains in the cost of collecting and annotating such data. Another point is the time for updating an existing model to a new domain. Recent works showed that a zero-shot learning method allows to bootstrap a model with good initial performance. To do so, this method relies on exploiting both a small-sized ontological description of the target domain and a generic word-embedding semantic space for generalization. Then, this framework has been extended to exploit user feedbacks to refine the zero-shot semantic parser parameters and increase its performance online. In this paper, we propose to drive this online adaptive process with a policy learnt using the Adversarial Bandit algorithm Exp3. We show, on the second Dialog State Tracking Challenge (DSTC2) datasets, that this proposition can optimally balance the cost of gathering valuable user feedbacks and the overall performance of the spoken language understanding module.Index Terms-Spoken language understanding, zero-shot learning, bandit problem, out-of-domain training data, online adaptation.

show abstract

Section: Online Interactive Refinement Problemmentioning

confidence: 99%

Section: Static Casementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Adversarial Bandit for online interactive active learning of zero-shot spoken language understanding

Ferreira

Masson

Jabaian

et al. 2016

2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

show abstract

Lilia, A Showcase for Fast Bootstrap of Conversation-Like Dialogues Based on a Goal-Oriented System

Riou

Jabaian

Huet

et al. 2019

Statistical Language and Speech Processing

Self Cite

View full text Add to dashboard Cite

Recently many works have proposed to cast human-machine interaction in a sentence generation scheme. Neural networks models can learn how to generate a probable sentence based on the user's statement along with a partial view of the dialogue history. While appealing to some extent, these approaches require huge training sets of generalpurpose data and lack a principled way to intertwine language generation with information retrieval from back-end resources to fuel the dialogue with actualised and precise knowledge. As a practical alternative, in this paper, we present Lilia, a showcase for fast bootstrap of conversation-like dialogues based on a goal-oriented system. First, a comparison of goaloriented and conversational system features is led, then a conversion process is described for the fast bootstrap of a new system, finalised with an on-line training of the system's main components. Lilia is dedicated to a chitchat task, where speakers exchange viewpoints on a displayed image while trying collaboratively to derive its author's intention. Evaluations with user trials showed its efficiency in a realistic setup.

show abstract

Hierarchical Dialog State Tracking with Unknown Slot Values

Yang

Wang

Yuan

2018

Neural Process Lett

View full text Add to dashboard Cite

Online adaptative zero-shot learning spoken language understanding using word-embedding

Cited by 16 publications

References 11 publications

Adversarial Bandit for online interactive active learning of zero-shot spoken language understanding

Adversarial Bandit for online interactive active learning of zero-shot spoken language understanding

Lilia, A Showcase for Fast Bootstrap of Conversation-Like Dialogues Based on a Goal-Oriented System

Hierarchical Dialog State Tracking with Unknown Slot Values

Contact Info

Product

Resources

About