Variation in Spoken North Sami Language

Jokinen, Kristiina; Trọng, Trung Ngô; Hautamäki, Ville

doi:10.21437/interspeech.2016-1438

Cited by 9 publications

(14 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…[2]). In this respect, our results are thus compatible with those of an earlier analysis of the same corpus presented in [12] but strengthen them in two respects. First, unlike in the earlier analysis, the separation of varieties based on majority language presently emerges without any prior division of varieties into majority-language classes.…”

Section: Discussionsupporting

confidence: 92%

See 1 more Smart Citation

Mapping areal variation and majority language influence in North Sámi using hierarchical prosodic analysis

Hiovain¹,

Suni²,

Simko³

et al. 2018

Speech Prosody 2018

View full text Add to dashboard Cite

We present results of a statistical hierarchical analysis of areal variation in prosody of spoken North Sámi languages. The hierarchical analysis method compares unigram models using cross-entropy measure. The models depict distributions of ∆features of f0 and energy signals decomposed using Continuous Wavelet Transform. These signals are obtained from speech recordings of five areal North Sámi varieties recorded in sites in northern Finland and Norway. We evaluate three potential sources of areal variation in prosodic characteristics of these five areal varieties: (1) traditional dialectal analysis of North Sámi, (2) influence of the relevant majority languages, and (3) geographical distance. Our results show a significant positive correlation between crossentropy distances between models and geographical distances between recording sites, demonstrating a viability of the method for typological analysis. Prosodic characteristics of the areal varieties are also influenced by majority languages, and, to a smaller degree, by differences between the North Sámi dialectal varieties.

show abstract

Section: Discussionsupporting

confidence: 92%

“…The same corpus was analyzed in an earlier work using similarly motivated established language recognition techniques [12]. This previous work used an i-vector representation of speech signal, that combines prosodic (energy) and spectral (MFCC) characteristics of speech signal [13].…”

Section: Introductionmentioning

confidence: 99%

Mapping areal variation and majority language influence in North Sámi using hierarchical prosodic analysis

Hiovain¹,

Suni²,

Simko³

et al. 2018

Speech Prosody 2018

View full text Add to dashboard Cite

show abstract

“…The remaining set of speakers is used for training our classifier. As the scheme used in [8], the results are comparable to previous i-vector approach [8]. It is notable that each repetition is done with a totally different set of speakers in order to preserve the generalizability of the experiments.…”

Section: Results and Analysismentioning

confidence: 62%

“…Finnish, Norwegian, or Swedish. In [8], we have previously shown that the North Sami regional dialects are confounded by multiple hidden factors which significantly influence the performance of an i-vector based system. One of the factors seems to be the variation among the speakers caused by the majority language: Finnish in Finland and Norwegian in Norway.…”

Section: Introductionmentioning

confidence: 99%

Staircase Network: structural language identification via hierarchical attentive units

Trọng¹,

Hautamäki²,

Jokinen³

2018

The Speaker and Language Recognition Workshop (Odyssey 2018)

Self Cite

View full text Add to dashboard Cite

Language recognition system is typically trained directly to optimize classification error on the target language labels, without using the external, or meta-information in the estimation of the model parameters. However labels are not independent of each other, there is a dependency enforced by, for example, the language family, which affects negatively on classification. The other external information sources (e.g. audio encoding, telephony or video speech) can also decrease classification accuracy. In this paper, we attempt to solve these issues by constructing a deep hierarchical neural network, where different levels of meta-information are encapsulated by attentive prediction units and also embedded into the training progress. The proposed method learns auxiliary tasks to obtain robust internal representation and to construct a variant of attentive units within the hierarchical model. The final result is the structural prediction of the target language and a closely related language family. The algorithm reflects a "staircase" way of learning in both its architecture and training, advancing from the fundamental audio encoding to the language family level and finally to the target language level. This process not only improves generalization but also tackles the issues of imbalanced class priors and channel variability in the deep neural network model. Our experimental findings show that the proposed architecture outperforms the state-of-the-art i-vector approaches on both small and big language corpora by a significant margin.

show abstract

“…The data also contain annotations at the word, sentence, and phrase level that were produced with a combination of manual annotations and forced alignment (using WebMAUS Basic, see [21,22]). For a more detailed description of the data collection process and corpus, see [23,17,20].…”

Section: The Extended Digisami Corpusmentioning

confidence: 99%

Dialect Identification of Spoken North Sámi Language Varieties Using Prosodic Features

Kakouros¹,

Hiovain²,

Vainio³

et al. 2020

Speech Prosody 2020

View full text Add to dashboard Cite

This work explores the application of various supervised classification approaches using prosodic information for the identification of spoken North Sámi language varieties. Dialects are language varieties that enclose characteristics specific for a given region or community. These characteristics reflect segmental and suprasegmental (prosodic) differences but also highlevel properties such as lexical and morphosyntactic. One aspect that is of particular interest and that has not been studied extensively is how the differences in prosody may underpin the potential differences among different dialects. To address this, this work focuses on investigating the standard acoustic prosodic features of energy, fundamental frequency, spectral tilt, duration, and their combinations, using sequential and context-independent supervised classification methods, and evaluated separately over two different units in speech: words and syllables. The primary aim of this work is to gain a better understanding on the role of prosody in identifying among the different language varieties. Our results show that prosodic information holds an important role in distinguishing between the five areal varieties of North Sámi where the inclusion of contextual information for all acoustic prosodic features is critical for the identification of dialects for words and syllables.

show abstract

Variation in Spoken North Sami Language

Cited by 9 publications

References 9 publications

Mapping areal variation and majority language influence in North Sámi using hierarchical prosodic analysis

Mapping areal variation and majority language influence in North Sámi using hierarchical prosodic analysis

Staircase Network: structural language identification via hierarchical attentive units

Dialect Identification of Spoken North Sámi Language Varieties Using Prosodic Features

Contact Info

Product

Resources

About