Annotated Corpus of Polish Spoken Dialogues

Mykowiecka, Agnieszka; Marasek, Krzysztof; Marciniak, Małgorzata; Rabiega-Wiśniewska, Joanna; Gubrynowicz, Ryszard

doi:10.1007/978-3-642-04235-5_5

Cited by 6 publications

(5 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The incorporated models include phrase-based models in source-to-target and target-to-source direction, IBM-1 like scores at phrase level, again in source-to-target and target-to-source direction, a target language model, and additional word and phrase penalties. These models are log-linearly combined [27]: (13) Here, represents feature functions (which are the aforementioned statistical models) and the corresponding scaling factors. These factors are optimized using some numerical algorithm in order to maximize translation performance on a development corpus.…”

Section: Statistical Machine Translation (Smt)mentioning

confidence: 99%

“…The data for the Polish corpus are human-human dialogues collected at the Warsaw Transportation call-center [38], [13]. This corpus covers the domain of transportation information like, e.g., transportation routes, itinerary, stops, or fare reductions.…”

Section: B Polish Luna Corpusmentioning

confidence: 99%

“…Two more corpora comparable to the MEDIA corpus in size and detail of annotation have been collected within the EU FP6 LUNA project: the Polish Warsaw transportation corpus [13] and the Italian help-desk corpus [14]. The considered corpora have ontologies of different types and complexity that can be represented in a frame language described in [7].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Comparing Stochastic Approaches to Spoken Language Understanding in Multiple Languages

Hahn

Dinarelli

Raymond

et al. 2011

IEEE Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

One of the first steps in building a spoken language understanding (SLU) module for dialogue systems is the extraction of flat concepts out of a given word sequence, usually provided by an automatic speech recognition (ASR) system. In this paper, six different modeling approaches are investigated to tackle the task of concept tagging. These methods include classical, well-known generative and discriminative methods like Finite State Transducers (FSTs), Statistical Machine Translation (SMT), Maximum Entropy Markov Models (MEMMs), or Support Vector Machines (SVMs) as well as techniques recently applied to natural language processing such as Conditional Random Fields (CRFs) or Dynamic Bayesian Networks (DBNs). Following a detailed description of the models, experimental and comparative results are presented on three corpora in different languages and with different complexity. The French MEDIA corpus has already been exploited during an evaluation campaign and so a direct comparison with existing benchmarks is possible. Recently collected Italian and Polish corpora are used to test the robustness and portability of the modeling approaches. For all tasks, manual transcriptions as well as ASR inputs are considered. Additionally to single systems, methods for system combination are investigated. The best performing model on all tasks is based on conditional random fields. On the MEDIA evaluation corpus, a concept error rate of 12.6% could be achieved. Here, additionally to attribute names, attribute values have been extracted using a combination of a rule-based and a statistical approach. Applying system combination using weighted ROVER with all six systems, the concept error rate (CER) drops to 12.0%.

show abstract

Section: Statistical Machine Translation (Smt)mentioning

confidence: 99%

Section: B Polish Luna Corpusmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Comparing Stochastic Approaches to Spoken Language Understanding in Multiple Languages

Hahn

Dinarelli

Raymond

et al. 2011

IEEE Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

show abstract

“…When non-hierarchical semantic representations are used in such a supervised context, the semantic frame induction task is essentially a supervised sequence labeling task, akin to "concept tagging", in which the words of an utterance are tagged with concepts (slots) from the semantic representation. (Hahn et al, 2011) apply a variety of discriminative and generative techniques to perform concept tagging of transcribed speech corpora (Bonneau-Maynard et al, 2009;Mykowiecka et al, 2009;Dinarelli et al, 2009). Previous work in the ALADIN project similarly applied an exemplar-based supervised concept tagging method, using manually tagged PATCOR (cf.…”

Section: Semantic Frame Inductionmentioning

confidence: 99%

Effective weakly supervised semantic frame induction using expression sharing in hierarchical hidden Markov models

Loo¹,

Gemmeke²,

Pauw³

et al. 2019

Preprint

View full text Add to dashboard Cite

We present a framework for the induction of semantic frames from utterances in the context of an adaptive command-and-control interface. The system is trained on an individual user's utterances and the corresponding semantic frames representing controls. During training, no prior information on the alignment between utterance segments and frame slots and values is available. In addition, semantic frames in the training data can contain information that is not expressed in the utterances. To tackle this weakly supervised classification task, we propose a framework based on Hidden Markov Models (HMMs). Structural modifications, resulting in a hierarchical HMM, and an extension called expression sharing are introduced to minimize the amount of training time and effort required for the user. The dataset used for the present study is patcor, which contains commands uttered in the context of a vocally guided card game, Patience. Experiments were carried out on orthographic and phonetic transcriptions of commands, segmented on different levels of n-gram granularity. The experimental results show positive effects of all the studied system extensions, with some effect differences between the different input representations. Moreover, evaluation experiments on held-out data with the optimal system configuration show that the extended system is able to achieve high accuracies with relatively small amounts of training data.

show abstract

“…Polish The data for the Polish corpus has been collected at the Warsaw Transportation call-center [7]. Also as part of the LUNA project, the manual annotation of these human-human dialogues has been performed [8]. This corpus covers the domain of transportation information like e.g.…”

Section: Corporamentioning

confidence: 99%

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

Hahn¹,

Lehnen²,

Heigold³

et al. 2009

Interspeech 2009

View full text Add to dashboard Cite

In this paper, we present improvements of our state-of-the-art concept tagger based on conditional random fields. Statistical models have been optimized for three tasks of varying complexity in three languages (French, Italian, and Polish). Modified training criteria have been investigated leading to small improvements. The respective corpora as well as parameter optimization results for all models are presented in detail. A comparison of the selected features between languages as well as a close look at the tuning of the regularization parameter is given. The experimental results show in what level the optimizations of the single systems are portable between languages.

show abstract

Annotated Corpus of Polish Spoken Dialogues

Cited by 6 publications

References 4 publications

Comparing Stochastic Approaches to Spoken Language Understanding in Multiple Languages

Comparing Stochastic Approaches to Spoken Language Understanding in Multiple Languages

Effective weakly supervised semantic frame induction using expression sharing in hierarchical hidden Markov models

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

Contact Info

Product

Resources

About