Optimizing CRFs for SLU tasks in various languages using modified training criteria

Hahn, Stefan; Lehnen, Patrick; Heigold, Georg; Ney, Hermann

doi:10.21437/interspeech.2009-697

Cited by 7 publications

(4 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These models use both word and classes, and a rich set of lexical features such like word prefixes, suffixes, word capitalization information etc. We note that the large gap between these CRF models is due to the fact that the CRF of [45] is trained with an improved margin criterion, similar to the large margin principle of SVM [46,47]. We note also that comparing significance tests published in [43], a difference of 0.1 in CER is already statistically significant.…”

Section: Comparison With the State-of-the-artmentioning

confidence: 99%

Label-Dependencies Aware Recurrent Neural Networks

Dupont

Dinarelli

2018

Computational Linguistics and Intelligent Text Processing

View full text Add to dashboard Cite

In the last few years, Recurrent Neural Networks (RNNs) have proved effective on several NLP tasks. Despite such great success, their ability to model sequence labeling is still limited. This lead research toward solutions where RNNs are combined with models which already proved effective in this domain, such as CRFs. In this work we propose a solution far simpler but very effective: an evolution of the simple Jordan RNN, where labels are re-injected as input into the network, and converted into embeddings, in the same way as words. We compare this RNN variant to all the other RNN models, Elman and Jordan RNN, LSTM and GRU, on two well-known tasks of Spoken Language Understanding (SLU). Thanks to label embeddings and their combination at the hidden layer, the proposed variant, which uses more parameters than Elman and Jordan RNNs, but far fewer than LSTM and GRU, is more effective than other RNNs, but also outperforms sophisticated CRF models.

show abstract

Section: Comparison With the State-of-the-artmentioning

confidence: 99%

Label-Dependencies Aware Recurrent Neural Networks

Dupont

Dinarelli

2018

Computational Linguistics and Intelligent Text Processing

View full text Add to dashboard Cite

show abstract

“…CRFs have been successfully used for SLU tasks [4]. We applied CRFs to the MEDIA corpus using the CRF++ toolkit (http://crfpp.googlecode.com/svn/trunk/doc/index.html).…”

Section: Two Different Slu Techniques Have Been Studied a Generative ...mentioning

confidence: 99%

“…We defined a set of basic features that includes only lexical information, setting a window such as incorporates the two previous and the two posterior words. A more complete set of features could be defined for applying the CRFs to SLU tasks [4], however, in this work we have not done a depth study of the best combination of features.…”

Section: Two Different Slu Techniques Have Been Studied a Generative ...mentioning

confidence: 99%

“…In the last few years, different approaches have been developed for the problem of SLU. As in other speech areas, statistical modelization has been successfully used in SLU [1], [2], [3], and [4]. There are many types of applications for SLU techniques; one of the most interesting is their application in limited-domain spoken dialog systems.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Combining multiple translation systems for Spoken Language Understanding portability

García

Hurtado

Segarra

et al. 2012

2012 IEEE Spoken Language Technology Workshop (SLT)

View full text Add to dashboard Cite

We are interested in the problem of learning Spoken Language Understanding (SLU) models for multiple target languages. Learning such models requires annotated corpora, and porting to different languages would require corpora with parallel text translation and semantic annotations. In this paper we investigate how to learn a SLU model in a target language starting from no target text and no semantic annotation. Our proposed algorithm is based on the idea of exploiting the diversity (with regard to performance and coverage) of multiple translation systems to transfer statistically stable word-toconcept mappings in the case of the romance language pair, French and Spanish. Each translation system performs differently at the lexical level (wrt BLEU). The best translation system performances for the semantic task are gained from their combination at different stages of the portability methodology. We have evaluated the portability algorithms on the French MEDIA corpus, using French as the source language and Spanish as the target language.The experiments show the effectiveness of the proposed methods with respect to the source language SLU baseline.

show abstract