MooDetecTR: Kelime Vektörleri Vasıtasıyla Türkçe Şarkı Sözleri için Ruh Hali Tespiti

Öz

Çevrimiçi müzik platformlarının kullanımının artmasıyla birlikte, katalog tabanlı aramalar, duygu bazlı aramalara dönüşmüştür. Bu çalışmada, Türkçe şarkıların duygu durum tespiti için kelime vektörlerini kullanan yarı denetimli bir öğrenme çerçevesi olan MooDetecTR önerilmiştir. Bu çerçevede, önce kelime vektörleri Word2Vec ve GloVe algoritmaları ile 2,5 milyondan fazla Türkçe belge içeren geniş bir metinsel veri koleksiyonu kullanılarak oluşturulmuştur. Daha sonra, duygu durum tespiti için seçilen şarkı sözlerindeki kelimelerin, daha önceden eğitilmiş kelime vektörlerinin birleştirilmesiyle şarkı sözleri vektörleri üretilmiştir. Son olarak, oluşturulan bu şarkı sözleri vektörleri, müzik duygu durum tespitinde kullanılmak üzere çeşitli makine öğrenmesi algoritmaları kullanılarak oluşturulan modelleri eğitmek için kullanılmıştır. Türkçe müziklerde duygu durumu tespiti karşılaştırma yapılmak üzere ayrıca, hem TF-IDF ağırlıkları kullanılarak geleneksel kelime çantası modeli ile hem de Doc2Vec algoritması kullanılarak oluşturulan modeller ile gerçekleştirilmiştir. Kelimelerin köklerine ayrıştırılması ve gereksiz kelimelerin kaldırılmasının sonuçlara etkileri de incelenmiştir. Önerilen çerçeve ile elde edilen en iyi mikro-f1 skoru (%54,36), Doc2Vec ve kelime çantası yöntemlerinden elde edilen en iyi skorlardan sırasıyla %3,81 ve %2,92 (%7,54 ve %5,68 nispi iyileştirmeler) daha başarılıdır. Sonuç olarak, elde edilen skorlar, Türkçe metin sınıflandırma uygulamasında büyük metinsel verilerin kullanılması ile oluşturulan kelime vektörlerinin olumlu etkisini artan sınıflandırma başarı performansı ile açıkça göstermektedir.

Anahtar Kelimeler

• Metin sınıflandırması
• Özellik üretimi
• Müzik ruh hali sınıflandırma
• Doğal dil işleme
• Kelime vektörleri

MOODETECTR: MOOD DETECTION FOR TURKISH LYRICS THROUGH WORD VECTORS

Abstract

Along with the increasing use of online music platforms, catalogue-based searches have turned into mood-based seeking. In this study, we propose MooDetecTR, a semi-supervised learning framework that employs word vectors for Turkish song mood detection. In this framework, first word vectors are created through a large collection of textual data, which include more than 2.5 million Turkish documents, by using Word2Vec and GloVe algorithms. Subsequently, lyrics vectors are generated through combining already trained word vectors of the words in the lyrics selected for mood detection. Lastly, lyrics vectors are fed into various machine-learning algorithms as features to create models for music mood detection. For comparison, Turkish music mood detection is performed both via traditional bag-of-words model, with TF-IDF weights, and Doc2Vec algorithm. The effects of stemming of the words and stop-words removal on the results are investigated, as well. The best micro-f1 score (54.36%) obtained by the proposed framework is 3.81%, and 2.92% higher (7.54%, and 5.68% relative improvements) than the best score obtained from Doc2Vec and bag-of-words methods, respectively. Consequently, the results obtained show the effectiveness of incorporating word vectors generated using big textual data into Turkish text classification process, which is clearly illustrated by the improved classification performance.

Keywords

• Text classification
• Feature Generation
• Music mood classification
• Natural language processing
• Word embeddings

References

1. Ali, S. O., & Peynircioǧlu, Z. F. (2006). Songs and emotions: Are lyrics and melodies equal partners? Psychology of Music, 34(4), 511–534. https://doi.org/10.1177/0305735606067168
2. Alparslan, E., Karahoca, A., & Bahşi, H. (2011). Classification of confidential documents by using adaptive neuro-fuzzy inference systems. In Procedia Computer Science (pp. 1412–1417). https://doi.org/10.1016/j.procs.2011.01.023
3. Cakir, M. U., & Guldamlasioglu, S. (2016). Text Mining Analysis in Turkish Language Using Big Data Tools. In 2016 IEEE 40th Annual Computer Software and Applications Conference (COMPSAC) (pp. 614–618). https://doi.org/10.1109/COMPSAC.2016.203
4. Casey, M. A., Veltkamp, R., Goto, M., Leman, M., Rhodes, C., & Slaney, M. (2008). Content-Based Music Information Retrieval: Current Directions and Future Challenges. In Proceedings of the IEEE (Vol. 96, pp. 668–696). https://doi.org/10.1109/JPROC.2008.916370
5. Cohen, J. (1960). A Coefficient of Agreement for Nominal Scales. Educational and Psychological Measurement, 20(1), 37–46. https://doi.org/10.1177/001316446002000104 Coltekin, C. (2010). A Freely Available Morphological Analyzer for Turkish. In Proceedings of the 7th International Conference on Language Resources and Evaluation (pp. 19–28).
6. Danilák, M. (n.d.). Langdetect 1.0.7. Python Package Index.
7. Fell, M., & Sporleder, C. (2014). Lyrics-based Analysis and Classification of Music. In International Conference on Computational Linguistics (pp. 620–631).
8. Fleiss, J. L., Nee, J. C., & Landis, J. R. (1979). Large sample variance of kappa in the case of different sets of raters. Psychological Bulletin, 86(5), 974–977. https://doi.org/10.1037/0033-2909.86.5.974

9. Günal, S. (2012). Hybrid feature selection for text classification. Turkish Journal of Electrical Engineering and Computer Sciences, 20(Sup. 2), 1296–1311. https://doi.org/10.3906/elk-1101-1064
10. Güran, A., Akyokuş, S., Güler, N., & Gürbüz, Z. (2009). Turkish Text Categorization Using N-Gram Words. In Proceedings of the international symposium on innovations in intelligent systems and applications (INISTA) (pp. 369–373).
11. He, H., & Garcia, E. A. (2009). Learning from imbalanced data. IEEE Transactions on Knowledge and Data Engineering, 21(9), 1263–1284. https://doi.org/10.1109/TKDE.2008.239
12. Hu, X., & Downie, J. S. (2010). When Lyrics Outperform Audio for Music Mood Classification: A Feature Analysis. In Proceedings of the 10th International Society for Music Information Retrieval Conference (pp. 619–624). Hu, X., Downie, J. S., & Ehmann, A. F. (2009). Lyric text mining in music mood classification. American Music, 619–624.
13. Kenter, T., & Rijke, M. de. (2015). Short Text Similarity with Word Embeddings Categories and Subject Descriptors. In Proceedings of the 24th ACM International on Conference on Information and Knowledge Management (CIKM 2015). https://doi.org/10.1145/2806416.2806475
14. Le, Q. V., & Mikolov, T. (2014). Distributed Representations of Sentences and Documents. In Proceedings of the 31st International Conference on Machine Learning (ICML-14) (pp. 1188–1196). https://doi.org/10.1145/2740908.2742760
15. Lamere, P. (2008). Social Tagging and Music Information Retrieval. Journal of New Music Research, 37(2), 101–114. https://doi.org/Doi 10.1080/09298210802479284\rPii 906001732
16. Mikolov, T., Corrado, G., Chen, K., & Dean, J. (2013). Efficient Estimation of Word Representations in Vector Space. In Proceedings of the International Conference on Learning Representations (ICLR 2013). https://doi.org/10.1162/153244303322533223
17. Mikolov, T., Sutskever, I., Chen, K., Corrado, G., & Dean, J. (2013). Distributed Representations of Words and Phrases and their Compositionality. Retrieved from https://arxiv.org/pdf/1310.4546.pdf
18. Ouyang, X., Zhou, P., Li, C. H., & Liu, L. (2015). Sentiment Analysis Using Convolutional Neural Network. In 2015 IEEE International Conference on Computer and Information Technology (pp. 2359–2364). https://doi.org/10.1109/CIT/IUCC/DASC/PICOM.2015.349
19. Özgür, L., Güngör, T., & Gürgen, F. (2004). Adaptive anti-spam filtering for agglutinative languages: A special case for Turkish. Pattern Recognition Letters, 25(16), 1819–1831. https://doi.org/10.1016/j.patrec.2004.07.004
20. Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., … Duchesnay, É. (2012). Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research, 12(Oct), 2825–2830. https://doi.org/10.1007/s13398-014-0173-7.2
21. Pennington, J., Socher, R., & Manning, C. (2014). Glove: Global Vectors for Word Representation. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP) (pp. 1532–1543). https://doi.org/10.3115/v1/D14-1162
22. Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and Social Psychology, 39(6), 1161–1178. https://doi.org/10.1037/h0077714
23. Song, Y., Dixon, S., & Pearce, M. (2012). Evaluation of Musical Features for Emotion Classification. In International Society for Music Information Retrieval Conference (ISMIR) (pp. 523–528).
24. Su, F., & Xue, H. (2017). Graph-based multimodal music mood classification in discriminative latent space. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (pp. 152–163). https://doi.org/10.1007/978-3-319-51811-4_13
25. Torunoǧlu, D., Çakirman, E., Ganiz, M. C., Akyokuş, S., & Gürbüz, M. Z. (2011). Analysis of preprocessing methods on classification of Turkish texts. In INISTA 2011 - 2011 International Symposium on INnovations in Intelligent SysTems and Applications. https://doi.org/10.1109/INISTA.2011.5946084
26. Uysal, A. K., & Gunal, S. (2014). The impact of preprocessing on text classification. Information Processing and Management, 50(1), 104–112. https://doi.org/10.1016/j.ipm.2013.08.006
27. Yang, Y.-H., & Chen, H. H. (2012). Machine Recognition of Music Emotion. ACM Transactions on Intelligent Systems and Technology, 3(3), 40. https://doi.org/10.1145/2168752.2168754
28. Zhang, Y., & Wallace, B. (2015). A Sensitivity Analysis of (and Practitioners’ Guide to) Convolutional Neural Networks for Sentence Classification. ArXiv Preprint ArXiv:1510.03820. https://doi.org/10.3115/v1/D14-1181