Finnish Dialect Identification: The Effect of Audio and Text

Abstract

Finnish is a language with multiple dialects that not only differ from each other in terms of accent (pronunciation) but also in terms of morphological forms and lexical choice. We present the first approach to automatically detect the dialect of a speaker based on a dialect transcript and transcript with audio recording in a dataset consisting of 23 different dialects. Our results show that the best accuracy is received by combining both of the modalities, as text only reaches to an overall accuracy of 57%, where as text and audio reach to 85%. Our code, models and data have been released openly on Github and Zenodo.

Full text

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 8777–8783
November 7–11, 2021. c
2021 Association for Computational Linguistics
8777
Finnish Dialect Identification: The Effect of Audio and Text
Mika Hämäläinen, Khalid Alnajjar, Niko Partanen and Jack Rueter
Department of Digital Humanities
University of Helsinki & Rootroo Ltd
firstname.lastname@helsinki.fi
Abstract
Finnish is a language with multiple dialects
that not only differ from each other in terms
of accent (pronunciation) but also in terms of
morphological forms and lexical choice. We
present the first approach to automatically de-
tect the dialect of a speaker based on a dialect
transcript and transcript with audio recording
in a dataset consisting of 23 different dialects.
Our results show that the best accuracy is re-
ceived by combining both of the modalities,
as text only reaches to an overall accuracy of
57%, where as text and audio reach to 85%.
Our code, models and data have been released
openly on Github and Zenodo.
1 Introduction
We present an approach for identifying the dialect
of a speaker automatically solely based on text and
on audio and text together. We compare the uni-
modal approach to the bimodal one. There are
no previous dialect identification approaches for
Finnish. There are several situations were a dialect
identification method can be of use. For example,
if we have ASR models fine tuned for specific di-
alects, the dialect identification from audio could be
used as a preprocessing step. The model could also
be used to label recorded materials automatically in
order to create archival metadata. In order to make
our contribution useful for others, we have released
our code, models and processed data openly on
GitHub1and Zenodo2.
Finnish is a large Uralic language that is one of
the official languages of Finland, and is used essen-
tially at all levels of the modern society. There are
approximately five million Finnish speakers. The
language belongs to the Finnic branch of the Uralic
language family, and is very closely related to Kare-
lian, Meänkieli and Kveeni, and is also closely re-
lated to the Estonian language. It is more distantly
1https://github.com/Rootroo-ltd/
FinnishDialectIdentification
2https://zenodo.org/record/5330673
related to numerous Uralic languages spoken in
Russia.
The history of written Finnish starts in the 16th
century. Current orthography is connected to this
written tradition, which developed into the current
form in the late 19th century with a conscious plan-
ning and systematic development of the lexicon.
After this, the changes have been minor (Häkkinen,
1994, 16), and also impacted lexicon, especially
what it comes to the development of the vocabu-
lary of the modern society and traditional agrarian
terminology becoming less known.
The Finnish spoken language, however, is still
largely based on Finnish dialects. In the 20th cen-
tury some of the strongest dialectal features have
been disappearing, but there are still clearly dis-
tinguishable spoken vernacular varieties that are
regionally marked. It has been shown that instead
of clear disappearance of dialects there are vari-
ous features that are spreading, but not at uniform
rate, and reportedly younger speakers use the are-
ally marked features less than the older speakers
(Lappalainen,2001, 92). Finnish vernaculars also
represent historically rather different Finnic vari-
eties, with major split between Eastern and Western
dialects. There are, however, also dialect continu-
ums and traditionally found gradual differentiation
from region to region.
Many of the changes have been lexical due to
technical innovations and modernization of the
society: orthographic spelling conventions have
largely remained the same. Spoken Finnish, on the
other hand, traditionally represents an areally di-
vided dialect continuum, with several sharp bound-
aries, and many regions of gradual differentiation
from one municipality to another municipality.
As mentioned, in the later parts of the 20th cen-
tury relatively strong dialect leveling has been tak-
ing place. Some of the Finnish dialects may already
be concerned endangered, although the complex re-
lationship between contemporary vernaculars and

8778
the most traditional dialectal forms makes this hard
to ascertain. Dialect leveling in itself is a process
known from many parts of Europe (Auer,2018).
However, in the case of Finnish the written stan-
dard has remained relatively far from the spoken
Finnish, besides individual narrow domains such
as news broadcasts were the written form is used
also in speech.
Additionally there have been distinct text col-
lections that include materials from this dialect
archive. These include dialect books specific
regions and municipalities, such as Oulun mur-
rekirja [Dialect Book of Oulu] (Pääkkönen,1994)
or Savonlinnan seudun murrekirja [Dialect book
of Savonlinna region] (Palander,1986). There
have also been more recent larger collections
that contains excerpts from essentially all dialects
(Lyytikäinen et al.,2013).
Especially in the later parts of 21th century the
spoken varieties have been leveling away from very
specific local dialects, and although regional vari-
eties still exist, most of the local varieties have
certainly became endangered. Similar processes of
dialect convergence have been reported from dif-
ferent regions in Europe, although with substantial
variation (Auer,2018). In the case of Finnish this
has not, however, resulted in merging of the written
and spoken standards, but the spoken Finnish has
remained, to our day, very distinct from the written
standard. In a late 1950s, a program was set up to
document extant spoken dialects, with the goal of
recording 30 hours of speech from each municipal-
ity. This work resulted in very large collections of
dialectal recordings (Lyytikäinen,1984, 448-449).
Many of these have been published, and some por-
tion has also been manually normalized. Dataset
used is described in more detail in Section 3 Data.
In Finnish linguistics the dialect identification
has primarily been studied in the context of folk
linguistics. In this line of research the perceptions
of native speakers are investigated (Niedzielski and
Preston,2000). This type of studies have been
done for Finnish, for example, by Mielikäinen and
Palander (2014), Räsänen and Palander (2015) and
Palander (2011). It has been possible to suggest
for individual dialects which features are the most
stable and will remain as local regional markers,
and which seem to be in retention (Räsänen and
Palander,2015, 25). In this study we conduct just
individual experiments and report their results, but
in the further research we hope the results could
be analyzed in more detail in connection with the
earlier dialect perception studies, as we believe the
differences in perceived dialect differences could
be compared to the difficulties and successes the
model has to differentiate individual varieties.
Dialect Short Sentences
Etelä-Häme EH 1860
Etelä-Karjala EK 813
Etelä-Pohjanmaa EP 2684
Etelä-Satakunta ES 848
Etelä-Savo ESa 1744
Eteläinen Keski-Suomi EKS 2168
Inkerinsuomalaismurteet IS 4035
Kaakkois-Häme KH 8026
Kainuu K 3995
Keski-Karjala KK 1640
Keski-Pohjanmaa KP 900
Länsi-Satakunta LS 1288
Länsi-Uusimaa LU 1171
Länsipohja LP 1026
Läntinen Keski-Suomi LKS 857
Peräpohjola P 1913
Pohjoinen Keski-Suomi PKS 733
Pohjoinen Varsinais-Suomi PVS 3885
Pohjois-Häme PH 859
Pohjois-Karjala PK 4292
Pohjois-Pohjanmaa PP 1801
Pohjois-Satakunta PS 2371
Pohjois-Savo PSa 2344
Table 1: Dialects and the number of sentences in each
dialect in the corpus
2 Related work
The current approaches to Finnish dialect have fo-
cused on the textual modality only. Previously, bi-
directional LSTM (long short-term memory) based
models have been used to normalize Finnish di-
alects to standard Finnish (Partanen et al.,2019)
and to adapt standard Finnish text into different
dialectal forms (Hämäläinen et al.,2020). Similar
approach has also been used to normalize historical
Finnish (Hämäläinen et al.,2021;Partanen et al.,
2021).
The closest research to our paper conducted for
Finnish has been detection of foreign accents from
audio. Behravan et al. (2013) have detected for-
eign accents from audio only by using i-vectors.
However, foreign accent detection is a very differ-
ent task to native speaker dialect detection. Many
foreign accents have clear cues through phonemes
that are not part of the Finnish phonotactic system,
where as with dialects, all phonemes are part of
Finnish.
There have been several recent approaches for
Arabic to detect dialect from text (Balaji et al.,

8779
2020;Talafha et al.,2020;Alrifai et al.,2021).
Textual dialect detection has been done also for
German (Jauhiainen et al.,2018), Romanian (Za-
haria et al.,2021) and Low Saxon (Siewert et al.,
2020). The methods used range from traditional
machine learning with features such as n-grams to
neural models with pretrained embeddings, as it is
typically the case in NLP research. None of these
approaches use audio, as they rely on text only.
At the same time, North Sami dialects have been
identified from audio by training several models,
kNNs, SVMs, RFs, CRFs, and LSTM, based on ex-
tracted features (Kakouros et al.,2020). Kethireddy
et al. (2020) use Mel-weighted SFF spectrogram
to detect spoken Arabic dialects. Mel spectograms
are also used by Draghici et al. (2020). All these
approaches are mono-modal and use only audio.
Based on our literature review, the existing ap-
proaches use either text or audio for dialect detec-
tion. We, however, use both modalities and apply
them on a language with no existing dialect detec-
tion models.
3 Data
The Finnish dialects are exceptionally well docu-
mented. In the 1950s the Finnish dialect archive
was formed with the goal of recording 30 hours
of speech from each Finnish municipality. This
goal was reached fast, and exceeded, resulting in a
very large collection of archived materials that is
stored in the Institute for the Languages of Finland
(Lyytikäinen,1984, 448-449), and known as Tape
Archive of the Finnish Language
3
. There have been
numerous publications based on these materials,
although it is hard to estimate into which extent
this covers the entire body of recorded work, which
totals 24,000 hours of audio.
The largest individual publication of these mate-
rials is beyond doubt the Samples of Spoken Finnish
series that was published in 1978–2000 as 50 book-
lets.
4
Each book contained approximately two
hours of transcriptions, from two different speak-
ers, and represents a different municipality. Later
these materials have been digitized and published
as an openly licensed dialect corpus (Institute for
the Languages of Finland,2014). There are also
3https://www.kotus.fi/en/corpora_and_
other_material/spoken_language_corpora
4https://www.kotus.fi/aineistot/
puhutun_kielen_aineistot/
murreaanitteita/suomen_kielen_naytteita_
-sarja
other related corpora, most importantly The Finnish
Dialect Syntax Archive that contains similar record-
ings annotated morphosyntactically (University of
Turku and Institute for the Languages of Finland,
1985). Since 1980s follow-up research has been
done in selected municipalities to track the changes
in the dialects (Lyytikäinen and Yli-Paavola,2010,
413), which is another significant line of research
that complements these older dialect materials.
Later the work on these published materials has
resulted in multiple electronic corpora that are cur-
rently available. Although they represent only a
tiny fraction of the entire recorded material, they
reach remarkable coverage of different dialects and
varieties of spoken Finnish. Some of these corpora
contain various levels of manual annotation, while
others are mainly plain text with associated meta-
data. Materials of this type can be characterized
by an explicit attempt to represent dialects in lin-
guistically accurate manner, having been created
primarily by linguists with formal training in the
field. These transcriptions are usually written with
a transcription systems specific for each research
tradition. The result of this type of work is not
simply a text containing some dialectal features,
but a systematic and scientific transcription of the
dialectal speech.
The corpus we have used in this study is the
above-mentioned Samples of Spoken Finnish cor-
pus (Institute for the Languages of Finland,2014).
The electronic version contains manually annotated
normalization into standard Finnish. The corpus is
almost 700,000 tokens large. The digital version,
including audio, is published with CC-BY license,
and is available in the Language Bank of Finland.
5
We have selected it into this study because of the
open license and large dialectal scope. We have
downloaded the corpus with the original audio files,
and extracted from the audio all utterances that are
shorter than 10 seconds in length. The dialect re-
gion classification is taken directly from the corpus
metadata.
Despite the successful attempt of the authors of
the corpus to include all dialects, the dialects are
not entirely equally represented in the corpus. One
reason for this is certainly the different sizes of
the dialect areas, and the variation introduced by
different speech rates of individual speakers. The
difference in the number of sentences per dialect
can be seen in Table 1. We do not consider this
5http://urn.fi/urn:nbn:fi:lb-201407141

8780
uneven distribution to be a problem, as it is mainly
a feature of this dataset. The data has been tok-
enized and the dialectal transcriptions are aligned
with audio on a sentence level. This makes our task
with the dialect detection model easier as no align-
ment is required. We randomly sort the sentences
in the data and split them into a training (70% of
the sentences), validation (15% of the sentences)
and test (15% of the sentences) sets. This means
that the models are trained and tested on a sentence
level rather than on smaller chunks.
4 Dialect detection
In this section, we describe the two different mod-
els we used to detect dialect automatically in the
corpus. The first method is based on text only and
the second method uses text and audio. Both of the
methods used the same training, validation and test
splits.
4.1 Text only model
We train a dialect classification model using a bi-
directional long short-term memory (LSTM) based
model (Hochreiter and Schmidhuber,1997) by us-
ing OpenNMT-py (Klein et al.,2017) with the de-
fault settings except for the encoder where we use
a BRNN (bi-directional recurrent neural network)
(Schuster and Paliwal,1997) instead of the de-
fault RNN (recurrent neural network), since BRNN
based models have been shown to provide better
results in a variety of tasks.
We use the default of two layers for both the
encoder and the decoder and the default attention
model, which is the general global attention pre-
sented by Luong et al. (2015). The models are
trained for the default of 100,000 steps. The model
receives dialectal text
6
as input and predicts a di-
alect name as an output.
4.2 Text and audio model
Our multi-modal model makes use of the dialectal
text and audio. The model combines BERT (De-
vlin et al.,2019) and XLSR-Wav2Vec2 (Baevski
et al.,2020) neural models trained on Finnish
data. We utilize the uncased Finnish BERT model
7
(Virtanen et al.,2019). The multilingual XLSR-
Wav2Vec2 model released by Facebook does not
6
We also experimented with a character-level model using
the same neural network structure, but the accuracy remained
low, only 37%
7https://huggingface.co/TurkuNLP/
bert-base- finnish-uncased- v1
support Finnish. Therefore we use a Finnish XLSR-
Wav2Vec2 model
8
that is fine-tuned using readily
available Finnish audio datasets: Finnish Common
Voice (Ardila et al.,2020), CSS10 Finnish (Park
and Mulc,2019) and Finnish parliament session
2
9
for 30 epochs. All audio input is resampled to
16kHz.
Our multi-modal model follows a siamese neu-
ral network architecture, where one side of the
network is dedicated to text and the other to au-
dio. We ensure that both sides produce an equal
size of features by 1) setting a fixed input length
to BERT where padding and truncating is applied
where necessary and 2) having two average pooling
layers following the output of each side. For the
textual output, a global average pooling is applied,
whereas an adaptive average pooling is applied to
the audio output. Afterwards, the pooled output
is concatenated and followed by a dropout layer
with a probability of 20%. Lastly, a fully connected
dense layer is employed as the classification layer.
In total, the network has 439 million trainable pa-
rameters and we fine-tuned it for 3 full epochs with
a learning rate of 1e-4.
5 Results
The results of the two models can be seen in Table 2.
These results were calculated using scikit-learn
10
(Pedregosa et al.,2011). It is clear from the results
that the text only model performed worse for every
single dialect than the audio and text model. In
terms of overall accuracies, the text based model
reached only an accuracy of 57%, where as the
text and audio based model reach to an accuracy
of 85%. This indicates that the audio has classifi-
catory features that are not represented in the text
version alone, although the text is in a transcription
system that accurately captures various dialectal
phenomena.
When comparing the per dialect performance
of the better model with the amount of data avail-
able for each dialect, we can make an interesting
observation that a high amount of data does not
equal to a high F1-score. Out of the 10 dialects
with the largest amount of samples in the data, only
3, Kaakkois-Häme,Inkerinsuomalaismurteet and
Kainuu, reached to an F1-score of at least 0.90.
8https://huggingface.co/aapot/
wav2vec2-large- xlsr-53- finnish
9http://urn.fi/urn:nbn:fi:
lb-2017020201
10sklearn.metrics.classification_report

8781
bi-LSTM on text Audio + BERT
precison recall f1 precision recall f1
EH 0.49 0.48 0.49 0.97 0.89 0.93
EK 0.51 0.44 0.47 0.86 0.57 0.69
EP 0.72 0.67 0.69 0.68 0.93 0.79
ES 0.5 0.53 0.51 0.79 0.82 0.8
Esa 0.38 0.37 0.38 0.6 0.97 0.74
EKS 0.44 0.38 0.41 0.9 0.85 0.87
IS 0.74 0.75 0.75 0.96 0.86 0.91
KH 0.67 0.74 0.7 0.86 0.97 0.92
K 0.53 0.49 0.51 0.97 0.83 0.9
KK 0.57 0.54 0.56 0.92 0.95 0.93
KP 0.46 0.45 0.46 0.81 0.87 0.84
LS 0.47 0.38 0.42 0.98 0.74 0.84
LU 0.56 0.52 0.54 0.97 0.98 0.98
LP 0.34 0.32 0.33 0.94 0.92 0.93
LKS 0.34 0.46 0.39 0.72 0.99 0.83
P 0.55 0.58 0.56 0.71 0.93 0.81
PKS 0.4 0.38 0.39 0.93 0.62 0.75
PVS 0.75 0.72 0.73 0.91 0.74 0.82
PH 0.32 0.31 0.31 0.83 0.63 0.72
PKS 0.6 0.58 0.59 0.92 0.8 0.86
PP 0.4 0.45 0.42 0.74 0.38 0.5
PS 0.5 0.53 0.51 0.9 0.89 0.89
PSa 0.43 0.47 0.45 0.68 0.87 0.76
Table 2: Results for the two models
The F1-score of the dialect with the second highest
number of samples, Pohjoinen Keski-Suomi, was
only 0.86. Other dialects that had an F1-score of
at least 0.9 were the 11th most resourced Etelä-
Häme, the 14th most resourced Keski-Karjala and
the 16th and 17th most resourced Länsi-Uusimaa
and Länsipohja.
The lowest F1-score was 0.5 for Pohjois-
Pohjanmaa. This is interesting as the dialect is
the 12th most resourced one. Even the two least
resourced dialects in our dataset, Etelä-Karjala and
Pohjoinen Keski-Suomi got higher F1-scores, 0.69
and 0.75 respectively. These results are an indi-
cation that some of the dialects are more clearly
marked making them easier to detect even with less
data, while some other dialects may have under-
gone a process of dialect leveling (see Hinskens
1998) making them less distinct from other dialec-
tal forms of Finnish. It is also possible that some di-
alects are already significantly close to one another,
and thereby the model simply cannot distinguish
them accurately. Further error analysis could reveal
important details of this type.
6 Conclusions
We have presented the first model for Finnish di-
alect classification for a relatively large number
of different dialects, 23 in total. Based on our ex-
periments, a text only model is not as effective in
dialect classification as a model with text and audio.
It is clear that the amount of data alone is not the
only variable that constitutes a high performance
of the model for a given dialect, but also how dis-
tinctive a given dialect is from other dialects. Since
the speakers in the test set were not present in the
training, we are confident that the dialect is the
feature that the model has learned to predict.
Using the audio materials offers in itself new
interesting possibilities for dialect clustering and
comparison. Traditional dialect atlases have also
been used in automatic comparison and grouping of
different Finnish dialects (Syrjänen et al.,2016). In
further research we believe also this kind of infor-
mation could be connected to the analysis to show
how the dialect identification exactly interacts with
the dialectal variation and differences at close mu-
nicipality level. At the same time the identifiability
of a dialect must be connected to the degree of di-
alect leveling, linguistic distances and differences
between them, so applying the model into newer
recordings could also yield information about these
processes.
We have made all the data, code and models
openly available on Github
11
and Zenodo
12
. We
believe that this is the only way to ensure this line
of research continues for the Finnish language in
the future as well.
References
Khaled Alrifai, Ghaida Rebdawi, and Nada Ghneim.
2021. Arabic tweeps dialect prediction based on
machine learning approach. International Journal
of Electrical & Computer Engineering (2088-8708),
11(2).
Rosana Ardila, Megan Branson, Kelly Davis, Michael
Kohler, Josh Meyer, Michael Henretty, Reuben
Morais, Lindsay Saunders, Francis Tyers, and Gre-
gor Weber. 2020. Common voice: A massively-
multilingual speech corpus. In Proceedings of the
12th Language Resources and Evaluation Confer-
ence, pages 4218–4222, Marseille, France. Euro-
pean Language Resources Association.
Peter Auer. 2018. Dialect change in europe–leveling
and convergence. The Handbook of Dialectology,
pages 159–76.
Alexei Baevski, Yuhao Zhou, Abdelrahman Mohamed,
and Michael Auli. 2020. wav2vec 2.0: A frame-
work for self-supervised learning of speech represen-
tations. Advances in Neural Information Processing
Systems, 33.
11https://github.com/Rootroo- ltd/
FinnishDialectIdentification
12https://zenodo.org/record/5330673

8782
Appiah Balaji, Nitin Nikamanth, and Bharathi B.
2020. Semi-supervised fine-grained approach for
Arabic dialect detection task. In Proceedings of
the Fifth Arabic Natural Language Processing Work-
shop, pages 257–261, Barcelona, Spain (Online).
Association for Computational Linguistics.
Hamid Behravan, Ville Hautamäki, and Tomi Kin-
nunen. 2013. Foreign accent detection from spoken
Finnish using i-vectors. In INTERSPEECH, volume
2013, page 14th.
Jacob Devlin, Ming-Wei Chang, Kenton Lee, and
Kristina Toutanova. 2019. Bert: Pre-training of
deep bidirectional transformers for language under-
standing. In Proceedings of the 2019 Conference of
the North American Chapter of the Association for
Computational Linguistics: Human Language Tech-
nologies, Volume 1 (Long and Short Papers), pages
4171–4186.
Alexandra Draghici, Jakob Abeßer, and Hanna Luka-
shevich. 2020. A study on spoken language iden-
tification using deep neural networks. In Proceed-
ings of the 15th International Conference on Au-
dio Mostly, AM ’20, page 253–256, New York, NY,
USA. Association for Computing Machinery.
Kaisa Häkkinen. 1994. Agricolasta nykykieleen:
suomen kirjakielen historia. Söderström.
Mika Hämäläinen, Niko Partanen, and Khalid Alnaj-
jar. 2021. Lemmatization of historical old literary
Finnish texts in modern orthography. In Actes de la
28e Conférence sur le Traitement Automatique des
Langues Naturelles. Volume 1 : conférence princi-
pale, pages 189–198, Lille, France. ATALA.
Mika Hämäläinen, Niko Partanen, Khalid Alnajjar,
Jack Rueter, and Thierry Poibeau. 2020. Automatic
dialect adaptation in Finnish and its effect on per-
ceived creativity. In 11th International Conference
on Computational Creativity (ICCC’20). Associa-
tion for Computational Creativity.
FLMP Hinskens. 1998. Dialect levelling: a two-
dimensional process. Folia Linguistica Historica,
32:35–51.
Sepp Hochreiter and Jürgen Schmidhuber. 1997.
Long short-term memory. Neural computation,
9(8):1735–1780.
Institute for the Languages of Finland. 2014. Suomen
kielen näytteitä - Samples of Spoken Finnish [online-
corpus], version 1.0. http://urn.fi/urn:nbn:fi:lb-
201407141.
Tommi Sakari Jauhiainen, Heidi Annika Jauhiainen,
Bo Krister Johan Linden, et al. 2018. Heli-based
experiments in swiss german dialect identification.
In Proceedings of the Fifth Workshop on NLP for
Similar Languages, Varieties and Dialects (VarDial
2018). The Association for Computational Linguis-
tics.
Sofoklis Kakouros, Katri Hiovain, Martti Vainio, and
Juraj Šimko. 2020. Dialect identification of spoken
north s\’ami language varieties using prosodic fea-
tures. arXiv preprint arXiv:2003.10183.
Rashmi Kethireddy, Sudarsana Reddy Kadiri, Paavo
Alku, and Suryakanth V. Gangashetty. 2020. Mel-
weighted single frequency filtering spectrogram for
dialect identification.IEEE Access, 8:174871–
174879.
Guillaume Klein, Yoon Kim, Yuntian Deng, JeanSenel-
lart, and Alexander M. Rush. 2017. OpenNMT:
Open-Source Toolkit for Neural Machine Transla-
tion. In Proc. ACL.
Hanna Lappalainen. 2001. Sosiolingvistinen katsaus
suomalaisnuorten nykypuhekieleen ja sen tutkimuk-
seen. Virittäjä, 105(1):74–74.
Minh-Thang Luong, Hieu Pham, and Christopher D
Manning. 2015. Effective approaches to attention-
based neural machine translation. arXiv preprint
arXiv:1508.04025.
Erkki Lyytikäinen. 1984. Suomen kielen nauhoiteark-
iston neljännesvuosisata. Virittäjä, 88(4):448–448.
Erkki Lyytikäinen, Jorma Rekunen, and Jaakko Yli-
Paavola. 2013. Suomen murrekirja. Gaudeamus.
Erkki Lyytikäinen and Jaakko Yli-Paavola. 2010.
Suomen kielen nauhoitearkisto 50-vuotias. Virittäjä,
114(3).
Aila Mielikäinen and Marjatta Palander. 2014. Miten
suomalaiset puhuvat murteista? — kansanlingvisti-
nen tutkimus metakielestä. Suomalaisen Kirjallisuu-
den Seura.
Nancy A Niedzielski and Dennis R Preston. 2000. Folk
linguistics. De Gruyter Mouton.
Matti Pääkkönen. 1994. Oulun seudun murrekirja.
Suomalaisen Kirjallisuuden Seura.
Marjatta Palander. 1986. Savonlinnan seudun mur-
rekirja. Suomalaisen Kirjallisuuden Seura.
Marjatta Palander. 2011. Itä- ja eteläsuomalaisten mur-
rekäsitykset. Suomalaisen Kirjallisuuden Seura.
Kyubyong Park and Thomas Mulc. 2019. Css10: A
collection of single speaker speech datasets for 10
languages. Proc. Interspeech 2019, pages 1566–
1570.
Niko Partanen, Khalid Alnajjar, Mika Hämäläinen, and
Jack Rueter. 2021. Linguistic change and historical
periodization of old literary Finnish. In Proceed-
ings of the 2nd International Workshop on Compu-
tational Approaches to Historical Language Change
2021, pages 21–27, Online. Association for Compu-
tational Linguistics.

8783
Niko Partanen, Mika Hämäläinen, and Khalid Alnajjar.
2019. Dialect text normalization to normative stan-
dard finnish. In Proceedings of the 5th Workshop
on Noisy User-generated Text (W-NUT 2019), pages
141–146.
F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel,
B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer,
R. Weiss, V. Dubourg, J. Vanderplas, A. Passos,
D. Cournapeau, M. Brucher, M. Perrot, and E. Duch-
esnay. 2011. Scikit-learn: Machine learning in
Python. Journal of Machine Learning Research,
12:2825–2830.
Maaret Räsänen and Marjatta Palander. 2015. Kansan-
dialektologinen testi murrepiirteiden keskinäisestä
murteellisuusjärjestyksestä [a perceptual test on the
hierarchy of Eastern-Finnish dialect features]. Virit-
täjä, 119(1).
Mike Schuster and Kuldip K Paliwal. 1997. Bidirec-
tional recurrent neural networks. IEEE transactions
on Signal Processing, 45(11):2673–2681.
Janine Siewert, Yves Scherrer, Martijn Wieling, and
Jörg Tiedemann. 2020. LSDC - a comprehensive
dataset for low Saxon dialect classification. In Pro-
ceedings of the 7th Workshop on NLP for Simi-
lar Languages, Varieties and Dialects, pages 25–35,
Barcelona, Spain (Online). International Committee
on Computational Linguistics (ICCL).
Kaj Syrjänen, Terhi Honkola, Jyri Lehtinen, Antti
Leino, and Outi Vesakoski. 2016. Applying popu-
lation genetic approaches within languages: Finnish
dialects as linguistic populations. Language Dynam-
ics and Change, 6(2):235–283.
Bashar Talafha, Mohammad Ali, Muhy Eddin Za’ter,
Haitham Seelawi, Ibraheem Tuffaha, Mostafa Samir,
Wael Farhan, and Hussein T Al-Natsheh. 2020.
Multi-dialect arabic bert for country-level dialect
identification. arXiv preprint arXiv:2007.05612.
University of Turku and Institute for the Languages of
Finland. 1985. The Finnish Dialect Syntax Archive.
Antti Virtanen, Jenna Kanerva, Rami Ilo, Jouni Lu-
oma, Juhani Luotolahti, Tapio Salakoski, Filip Gin-
ter, and Sampo Pyysalo. 2019. Multilingual is
not enough: BERT for Finnish. arXiv preprint
arXiv:1912.07076.
George-Eduard Zaharia, Andrei-Marius Avram,
Dumitru-Clementin Cercel, and Traian Rebedea.
2021. Dialect identification through adversarial
learning and knowledge distillation on romanian
bert. In Proceedings of the Eighth Workshop on
NLP for Similar Languages, Varieties and Dialects,
pages 113–119.