Abstract
Mongolian morphological analysis (MMA) includes two subtasks: morphological segmentation and morphological tagging. It is a crucial preprocessing step in many Mongolian NLP applications. Recently, end-to-end neural approaches have achieved excellent results in the MMA task. However, these approaches handle morphological segmentation and morphological tagging independently, and ignore the relationship between the two subtasks. In this paper, we propose a multi-task sequence-to-sequence model for the MMA task that learns Mongolian morphological segmentation and tagging jointly. The proposed neural model introduces a shared morphological feature encoder to learn character-level and context-level word information. Besides, we design a flat joint attention decoder and a hierarchical joint attention decoder to generate Mongolian segmentation and tagging results, respectively. We employ the dynamic weight scheme to optimize and balance the weights between the two subtasks in MMA. We compare the proposed model with the baselines and evaluate the effectiveness of the sub-modules in the experiment. The result suggests that the proposed MMA model outperformed the state-of-the-art baselines.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Kullmann, R., Tserenpil, D.: Mongolian Gramma. ADMON Co., Ltd, Mongolia, pp. 33–72 (2008)
Na, L.: Mongolian Morphological Analysis and Application Research. Inner Mongolia University, Hohhot (2022)
Yatu, J., Lei, S., Yila, S., et al.: A strategy for referential problem in low-resource neural machine translation. In: International Conference of Artificial Neural Networks and Machine Learning (2021)
Rui, L., Sisman., Haizhou, L.: GraphSpeech: syntax-aware graph attention network for neural speech synthesis. In: Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing, pp. 6059–6063 (2021)
Wang, W., Bao, F., Gao, G.: Learning morpheme representation for mongolian named entity recognition. Neural Process. Lett. 50(2), 2647–2664 (2019)
Liu, N., Su, X., Gao, G., Bao, F., Lu, M.: Morphological knowledge guided Mongolian constituent parsing. In: Gedeon, T., Wong, K.W., Lee, M. (eds.) ICONIP 2019. LNCS, vol. 11955, pp. 363–375. Springer, Cham (2019). https://doi.org/10.1007/978-3-030-36718-3_31
Loglo, S.: Design and implementation of Mongolian fixed phrase recognition algorithm. J. Chin. Inf. Process. 31, 85–91 (2017)
Sorokin, A.: Convolutional neural networks for low-resource morpheme segmentation: baseline or state-of-the-art?. In: Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics, pp. 154–159 (2019)
Ertai, Q.: Mongolian Gramma (2010)
Na, L., Xiangdong, S., Guanglai, G., et al.: Mongolian word segmentation based on three character level seq2seq models. In: 25th International Conference. Proceedings of the 25th International Conference, pp. 558–569 (2018)
Shunle, Z.: A neural attention based model for morphological segmentation. Wireless Pers. Commun. 102(4), 2527–2534 (2011)
Chen, Z., Badrinarayanan., V., et al.: GradNorm: gradient normalization for adaptive loss balancing in deep multitask networks (2017)
Shikun, L., Johns, E., Davison, A.: Research on slavic Mongolian word segmentation based on dictionary and rule. J. Chin. Inf. Process. (2019). https://doi.org/10.1109/CVPR.2019.00197
Kendall, A., Gal, Y., Cipolla, R. Multi-task learning using uncertainty to weigh losses for scene geometry and semantics. In: Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 7482–7491 (2018)
Mekala., Dheeraj., Jingbo, S.: Contextualized weak supervision for text classification. In: Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pp. 323–333 (2020)
Lin, X., Huiling, Z., Zhehua, Z., et al.: Pareto multi-task learning (2020)
Luong, M.T., Pham, H., Manning, C.D.: Effective approaches to attention-based neural machine translation. In: Proceedings of the Conference on Empirical Methods in Natural Language Processing (2015)
Shengai, C.: Research on Mongolian Morphological Analysis Using Deep Neural Network. Beijing Jiaotong University, Beijing (2019)
Na, L., Xiangdong, S., Haoran, Z., et al.: Incorporating inner-word and out-word features for Mongolian morphological segmentation. In: Proceedings of the 28th International Conference on Computational Linguistics, pp. 4638–4648 (2020)
Acknowledgments
This work was funded by research program of science and technology at Universities of Inner Mongolia Autonomous Region (Grant No. NJZZ22251), 2022 Inner Mongolia Talent Support Project(DC2300001440), Inner Mongolia Natural Science Foundation (2022MS06013), Universities directly under the autonomous region Funded by the Fundamental Research Fund Project (JY20220122), Program for Young Talents of Science and Technology in Universities of Inner Mongolia Autonomous Region (NJYT23059).
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2023 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Liu, N., Qing-Dao-Er-Ji, R., Su, X., Ji, Y., Aodengbala, Liu, G. (2023). Multi-task Learning for Mongolian Morphological Analysis. In: Iliadis, L., Papaleonidas, A., Angelov, P., Jayne, C. (eds) Artificial Neural Networks and Machine Learning – ICANN 2023. ICANN 2023. Lecture Notes in Computer Science, vol 14262. Springer, Cham. https://doi.org/10.1007/978-3-031-44201-8_6
Download citation
DOI: https://doi.org/10.1007/978-3-031-44201-8_6
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-44200-1
Online ISBN: 978-3-031-44201-8
eBook Packages: Computer ScienceComputer Science (R0)