TY - JOUR
T1 - Computer aided functional style identification and correction in modern russian texts
AU - Savchenko, Elizaveta
AU - Lazebnik, Teddy
N1 - Publisher Copyright:
© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2021.
PY - 2022/3
Y1 - 2022/3
N2 - Functional style (FS) identification is a classification task in linguistics that categorizes unrestricted texts into several categories of linguistic norms. FS is widely used to attain a satisfying outcome in style processing. As such, we train a deep learning attention neural network model on modern-Russian texts, divide them into four FS categories. The model obtained an accuracy of 0.72. In particular, 81.08% and 85.71% accuracy in classifying the artistic and academic FS. The proposed model is able to automate the FS identification process and aids both domain experts and non-domain experts to perform FS correction by highlighting style anomalies concerning a desired style for the text. In particular, we show a 34% and 31% average improvement in the duration of performing the style correction task. Moreover, domain experts and non-domain experts obtain 3% and 9% more accurate results, respectively.
AB - Functional style (FS) identification is a classification task in linguistics that categorizes unrestricted texts into several categories of linguistic norms. FS is widely used to attain a satisfying outcome in style processing. As such, we train a deep learning attention neural network model on modern-Russian texts, divide them into four FS categories. The model obtained an accuracy of 0.72. In particular, 81.08% and 85.71% accuracy in classifying the artistic and academic FS. The proposed model is able to automate the FS identification process and aids both domain experts and non-domain experts to perform FS correction by highlighting style anomalies concerning a desired style for the text. In particular, we show a 34% and 31% average improvement in the duration of performing the style correction task. Moreover, domain experts and non-domain experts obtain 3% and 9% more accurate results, respectively.
KW - Deep learning
KW - Linguistic tasks
KW - Russian NLP
KW - Russian text
KW - Text classification
KW - Text styling
UR - http://www.scopus.com/inward/record.url?scp=85146797727&partnerID=8YFLogxK
U2 - 10.1007/s42488-021-00062-2
DO - 10.1007/s42488-021-00062-2
M3 - ???researchoutput.researchoutputtypes.contributiontojournal.article???
AN - SCOPUS:85146797727
SN - 2524-6356
VL - 4
SP - 25
EP - 32
JO - Journal of Data, Information and Management
JF - Journal of Data, Information and Management
IS - 1
ER -