Elamur Research⌕
☰
home / papers / 2022 / ANEC: An Amharic Named Entity Corpus and Transformer Based Recognizer

ANEC: An Amharic Named Entity Corpus and Transformer Based Recognizer

Summary

Amharic is a Semitic language written in the Ge'ez-derived Fidäl abugida, and it has had very little named entity recognition (NER) work. Three features make the task hard. The script has no capital letters to mark proper names. Prefixes and suffixes attach to word forms. And the same word can often be spelled several ways. The authors built ANEC, a manually annotated Amharic NER corpus of 8,070 sentences and 182,691 tokens. They also wrote detailed tagging guidelines in an ENAMEX style for persons, organizations and locations. Their recognizer is a BiLSTM network with a CRF output layer. Because entity tokens are rare compared with ordinary words, they applied SMOTE oversampling to reduce class imbalance. The system reaches an F1 of 93%, which the authors report as a new state of the art for Amharic NER. The work matters because it supplies a sizeable labeled resource and a strong baseline for a widely spoken but under-resourced language. NER output also feeds downstream tasks such as QA and machine translation.

Key points

  • The corpus, ANEC, has 8,070 sentences and 182,691 tokens with manually annotated entities, and the authors promise to release it publicly.
  • The model is a BiLSTM-CRF sequence labeler. SMOTE is used to oversample minority entity classes, and the reported F1 is 93%, a claimed new state of the art for Amharic.
  • The paper names three challenges specific to Amharic:
    • there is no capitalization, so ሰላም can be the name "Selam" or the noun "peace";
    • agglutinative affixes can shift an entity's type, so a person name with the prefix ከ can work as a location;
    • spelling variants such as አመት/ዓመት inflate OOV rates.
  • The script has 317 characters in total: 238 base syllable forms, 50 labialized forms, 9 punctuation marks and 20 numerals.
  • The annotation guidelines set out explicit rules:
    • titles such as saint, prophet and imam are left out of PERSON spans;
    • corporate designators and words like "River" or "Mountain" are kept inside the span;
    • place names listed in a row are tagged as separate LOCATIONs;
    • directional words are included only when they belong to an official name.

Interesting ideas

  • I1. The guidelines separate honorific titles from the PERSON span but keep type designators such as "Corporation" or "River" inside ORG and LOC spans, which gives clean and predictable entity boundaries. — For Mova (world): Entity extraction into the world of the text could follow the same split, storing titles as attributes of a person entity while keeping designators as part of the canonical name.
  • I2. The paper shows that morphology, not just context, can change an entity's type: a case-like prefix turns a person name into a place reading. — For Mova (frames): When assigning semantic roles, the case marker or preposition governing a name should be allowed to override its default entity type, for example treating "at Johnson's" as a location.
  • I3. The paper resamples the rare entity classes with SMOTE instead of accepting a heavily skewed label distribution. — For Mova (learn): When training taggers on silver data, rare labels such as uncommon dependency relations or entity types could be oversampled, with gains checked against a negative control.
  • I4. The paper treats orthographic variants of the same word as a direct cause of OOV growth and weaker embeddings. — For Mova (tokens): A normalization map from spelling variants (British/American spellings, informal forms) to a canonical lemma before embedding lookup could reduce sparsity in the tagger.
  • I5. The ambiguity example (Selam versus "peace") shows why surface lookup lists cannot work without capital letters, and why decisions must depend on context. — For Mova (tokens): Lowercased or social-media input could be tested with this kind of paired ambiguous sentence to check that the tagger's proper-noun decisions come from context rather than casing.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Linguistics & languages

Topics: Named entity recognition

Cited by

5 papers of our library cite it (of the 140884 whose bibliographies we read); 1 of them have a page here, 4 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2207.00785 · arXiv 2207.00785

@misc{jibril2022anec,
  title = {{ANEC: An Amharic Named Entity Corpus and Transformer Based Recognizer}},
  author = {Jibril, Ebrahim Chekol and Tantug, A. Cuneyd},
  year = {2022},
  doi = {10.48550/arxiv.2207.00785},
  eprint = {2207.00785},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2207.00785}
}

References

70 references, 13 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Adelani, D. I., Abbott, J., Neubig, G., D’souza, D., Kreutzer, J., Lignos, C., ... & Osei, S. (2021). MasakhaNER: Named entity recognition for African languages. Transactions of the Association for Computational Linguistics, 9, 1116-1131. find our card
  2. Akbani, R., Kwek, S., & Japkowicz, N. (2004, September). Applying support vector machines to imbalanced datasets. In European conference on machine learning (pp. 39-50). Springer, Berlin, Heidelberg. find
  3. Alemu, B. (2013). A Named Entity Recognition for Amharic. MSc., Computer Science, Addis Ababa University, Addis Ababa. find
  4. Argaw, A. A. and L. Asker (2007). An Amharic stemmer: Reducing words to their citation forms. Proceedings of the 2007 workshop on computational approaches to Semitic languages: Common issues and resources, Association for Computational Linguistics. find
  5. Babych, B. and A. Hartley (2003). Improving machine translation quality with automatic named entity recognition. Proceedings of the 7th International EAMT workshop on MT and other Language Technology Tools, Improving MT through other Language Technology Tools: Resources and Tools for Building MT, Association for Computational Linguistics. find
  6. Belay, B., Habtegebrial, T., Liwicki, M., Belay, G., & Stricker, D. (2019, September). Factored convolutional neural network for Amharic character image recognition. In 2019 IEEE International Conference on Image Processing (ICIP) (pp. 2906-2910). IEEE. find
  7. Belay, B., Habtegebrial, T., Meshesha, M., Liwicki, M., Belay, G., & Stricker, D. (2020). Amharic OCR: an end-to-end learning. Applied Sciences, 10(3), 1117. find
  8. Belay, M. T. (2014). Amharic Named Entity Recognition Using A Hybrid Approach. MSc., Computer Science, Addis Ababa University, Addis Ababa. find
  9. Chawla, N. V., Bowyer, K. W., Hall, L. O., & Kegelmeyer, W. P. (2002). SMOTE: synthetic minority over-sampling technique. Journal of artificial intelligence research, 16, 321-357. find our card
  10. Chinchor, N. and B. Sundheim (1993). MUC-5 evaluation metrics. In Proceedings of the Fourth Message Understanding Conference (MUC-5): 22–29. find
  11. Chiu, J. P. and E. Nichols (2016). Named entity recognition with bidirectional LSTM-CNNs. Transactions of the Association for Computational Linguistics 4: 357-370. find our card
  12. Collobert, R., Weston, J., Bottou, L., Karlen, M., Kavukcuoglu, K., & Kuksa, P. (2011). Natural language processing (almost) from scratch. Journal of machine learning research, 12(ARTICLE), 2493-2537. find our card
  13. Demeke, G. A. and M. Getachew (2006). Manual annotation of Amharic news items with part-of-speech tags and its challenges. Ethiopian Languages Research Center Working Papers 2: 1-16. find
  14. Demissie, D. (2017). Amharic Named Entity Recognition Using Neural Word Embedding as a Feature. MSc., Computer Science, Addis Ababa University, Addis Ababa. find
  15. DeRouin, E., Brown, J., Fausett, L., & Schneider, M. (1991). Neural Network Training on Unequally Represented Classes. In Intelligent Engineering Systems Through Artificial Neural Networks, pp. 135–141 New York. ASME Press. find
  16. Devlin, J., Chang, M. W., Lee, K., & Toutanova, K. (2018). Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805. ↗
  17. Dai, A. M., & Le, Q. V. (2015). Semi-supervised sequence learning. Advances in neural information processing systems, 28, 3079-3087. find
  18. Dong, C., et al. (2016). Character-based LSTM-CRF with radical-level features for Chinese named entity recognition. Natural Language Understanding and Intelligent Applications, Springer: 239-250. find
  19. Dyer, C., Ballesteros, M., Ling, W., Matthews, A., & Smith, N. A. (2015). Transition-based dependency parsing with stack long short-term memory. arXiv preprint arXiv:1505.08075. ↗ our card
  20. Eberhard, David M.; Simons, Gary F.; Fennig, Charles D., eds. (2021). Amharic. Ethnologue: Languages of the World (Eighteenth ed.). Dallas, Texas: SIL International. Retrieved 2 March 2021. find
  21. Ertekin, S., Huang, J., Bottou, L., & Giles, L. (2007, November). Learning on the border: active learning in imbalanced data classification. In Proceedings of the sixteenth ACM conference on Conference on information and knowledge management (pp. 127-136). find
  22. Fawcett, T., and Provost, F. (1996). Combining Data Mining and Machine Learning for Effective User Profile. In Proceedings of the 2nd International Conference on Knowledge Discovery and Data Mining, pp. 8–13 Portland, OR. AAAI. find
  23. Firdyiwek, Y., & Yaqob, D. (1997). The system for Ethiopic representation in ASCII. find
  24. Garg, Y., & Chatterjee, N. (2014, December). Sentiment analysis of twitter feeds. In International Conference on Big Data Analytics (pp. 33-52). Springer, Cham. find
  25. Gasser, M. (2011, May). HornMorpho: a system for morphological processing of Amharic, Oromo, and Tigrinya. In Conference on Human Language Technology for Development, Alexandria, Egypt. find
  26. Graves, A., & Schmidhuber, J. (2005). Framewise phoneme classification with bidirectional LSTM and other neural network architectures. Neural networks, 18(5-6), 602-610. find
  27. Graves, A., Mohamed, A. R., & Hinton, G. (2013, May). Speech recognition with deep recurrent neural networks. In 2013 IEEE international conference on acoustics, speech and signal processing (pp. 6645-6649). find
  28. Greenwood, M. A. and R. Gaizauskas (2003). Using a named entity tagger to generalise surface matching text patterns for question answering. Proceedings of the Workshop on Natural Language Processing for Question Answering (EACL03), Citeseer. find
  29. Grishman, R. and B. Sundheim (1996). Message understanding conference-6: A brief history. COLING 1996 Volume 1: The 16th International Conference on Computational Linguistics. find
  30. Hochreiter, S., Bengio, Y., Frasconi, P., & Schmidhuber, J. (2001). Gradient flow in recurrent nets: the difficulty of learning long-term dependencies. A field guide to dynamical recurrent neural networks. IEEE Press. find
  31. Hochreiter, S. and J. Schmidhuber (1997). Long short-term memory. Neural computation 9(8): 1735-1780. find
  32. Howard, J., & Ruder, S. (2018). Universal language model fine-tuning for text classification. arXiv preprint arXiv:1801.06146. ↗
  33. Huang, Z., Xu, W., & Yu, K. (2015). Bidirectional LSTM-CRF models for sequence tagging. arXiv preprint arXiv:1508.01991. ↗ our card
  34. Kubat, M., Holte, R., & Matwin, S. (1998). Machine Learning for the Detection of Oil Spills in Satellite Radar Images. Machine Learning, 30, 195–215. find
  35. Lafferty, J., McCallum, A., & Pereira, F. C. (2001). Conditional random fields: Probabilistic models for segmenting and labeling sequence data. find
  36. Lample, G., Ballesteros, M., Subramanian, S., Kawakami, K., & Dyer, C. (2016). Neural architectures for named entity recognition. arXiv preprint arXiv:1603.01360. ↗
  37. Li, Q., Shah, S., Liu, X., & Nourbakhsh, A. (2017, May). Data sets: Word embeddings learned from tweets and general data. In Proceedings of the International AAAI Conference on Web and Social Media (Vol. 11, No. 1). find
  38. Liu, Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., ... & Stoyanov, V. (2019). Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692. ↗
  39. Luo, G., Huang, X., Lin, C. Y., & Nie, Z. (2015, September). Joint entity recognition and disambiguation. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing (pp. 879-888). find
  40. Ma, X. and E. Hovy (2016). End-to-end sequence labeling via bi-directional lstm-cnns-crf. arXiv preprint arXiv:1603.01354. ↗ our card
  41. Maloney, J. and M. Niv (1998). TAGARAB: a fast, accurate Arabic name recognizer using high-precision morphological analysis. Computational Approaches to Semitic Languages. find
  42. McHugh, M. L. (2012). Interrater reliability: the kappa statistic. Biochemia medica, 22(3), 276-282. find
  43. Mehamed, M. A. (2010). Named Entity Recognition For Amharic Language. MSc., Computer Science, Addis Ababa University, Addis Ababa. find
  44. Mekuria, G. T., & Mekuria, G. T. (2018). Amharic text document summarization using parser. International Journal of Pure and Applied Mathematics, 118(24). find
  45. Mikolov, T., Karafiát, M., Burget, L., Cernocký, J., & Khudanpur, S. (2010, September). Recurrent neural network based language model. In Interspeech (Vol. 2, No. 3, pp. 1045-1048). find
  46. Mikolov, T., Yih, W. T., & Zweig, G. (2013). Linguistic regularities in continuous space word representations. In Proceedings of the 2013 conference of the north american chapter of the association for computational linguistics: Human language technologies (pp. 746-751). find our card
  47. Misawa, S., et al. (2017). Character-based Bidirectional LSTM-CRF with words and characters for Japanese Named Entity Recognition. Proceedings of the First Workshop on Subword and Character Level Models in NLP. find
  48. Mulugeta, W. and M. Gasser (2012). Learning morphological rules for Amharic verbs using inductive logic programming. Language Technology for Normalisation of Less-Resourced Languages 7. find
  49. Passos, A., Kumar, V., & McCallum, A. (2014). Lexicon infused phrase embeddings for named entity resolution. arXiv preprint arXiv:1404.5367. ↗ our card
  50. Peters, M. E., Neumann, M., Iyyer, M., Gardner, M., Clark, C., Lee, K., & Zettlemoyer, L. (2018). Deep contextualized word representations. arXiv preprint arXiv:1802.05365. ↗
  51. Provost, F., and Fawcett, T. (2001). Robust Classification for Imprecise Environments. Machine Learning, 42/3, 203–231. find
  52. Rajpurkar, P., Zhang, J., Lopyrev, K., & Liang, P. (2016). Squad: 100,000+ questions for machine comprehension of text. arXiv preprint arXiv:1606.05250. ↗ our card
  53. Ramshaw, L. A., & Marcus, M. P. (1999). Text chunking using transformation-based learning. In Natural language processing using very large corpora (pp. 157-176). Springer, Dordrecht. find our card
  54. Ramyachitra, D., and Manikandan, P. (2014). Imbalanced dataset classification and solutions: a review. International Journal of Computing and Business Research (IJCBR), 5(4), 1-29. find
  55. Ratinov, L. and D. Roth (2009). Design challenges and misconceptions in named entity recognition. Proceedings of the thirteenth conference on computational natural language learning, Association for Computational Linguistics. find our card
  56. Ratnaparkhi, A. (1998). Maximum entropy models for natural language ambiguity resolution. University of Pennsylvania. PhD thesis Computer and Information Science, University of Pennsylvania. find
  57. Sang, E. F., and De Meulder, F. (2003). Introduction to the CoNLL-2003 shared task: Language-independent named entity recognition. arXiv preprint cs/0306050. find
  58. Sang, E. F. and J. Veenstra (1999). Representing text chunks. Proceedings of the ninth conference on European chapter of the Association for Computational Linguistics, Association for Computational Linguistics. find our card
  59. Segura-Bedmar, I., et al. (2013). Semeval-2013 task 9: Extraction of drug-drug interactions from biomedical texts (ddiextraction 2013). Second Joint Conference on Lexical and Computational Semantics (* SEM), Volume 2: Proceedings of the Seventh International Workshop on Semantic Evaluation (SemEval 2013). find
  60. Shaalan, K. and H. Raza (2007). Person name entity recognition for Arabic. Proceedings of the 2007 Workshop on Computational Approaches to Semitic Languages: Common Issues and Resources, Association for Computational Linguistics. find
  61. Sikdar, U. K., & Gambäck, B. (2017, April). Named entity recognition for Amharic using stack-based deep learning. In International Conference on Computational Linguistics and Intelligent Text Processing (pp. 276-287). Springer, Cham. find
  62. Sintayehu, H., & Lehal, G. S. (2021). Named entity recognition: a semi-supervised learning approach. International Journal of Information Technology, 13(4), 1659-1665. find
  63. Sutskever, I., Vinyals, O., & Le, Q. V. (2014). Sequence to sequence learning with neural networks. In Advances in neural information processing systems (pp. 3104-3112). find
  64. The Ethiopian Press Agency. Will Amharic be AU’s Lingua Franca? The Ethiopian Herald. February 2019. Available online: https://www.press.et/english/?p=2654#l (accessed on 26 October 2021). ↗
  65. Tjong Kim Sang, E. F. and F. De Meulder (2003). Introduction to the CoNLL-2003 shared task: Language-independent named entity recognition. Proceedings of the seventh conference on Natural language learning at HLT-NAACL 2003-Volume 4, Association for Computational Linguistics. find
  66. Toda, H. and R. Kataoka (2005). A search result clustering method using informatively named entities. Proceedings of the 7th annual ACM international workshop on Web information and data management, ACM. find
  67. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., ... & Polosukhin, I. (2017). Attention is all you need. In Advances in neural information processing systems (pp. 5998-6008). find
  68. Zeng, D., et al. (2017). "LSTM-CRF for drug-named entity recognition." Entropy 19(6): 283. find
  69. Kingma, D. P., & Ba, J. (2014). Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980. ↗
  70. Nadeau, D., Turney, P. D., & Matwin, S. (2006, June). Unsupervised named-entity recognition: Generating gazetteers and resolving ambiguity. In Conference of the Canadian society for computational studies of intelligence (pp. 266-277). Springer, Berlin, Heidelberg. Acknowledgments At this work, Ebrahim Chekol Jibril was supported by the Turkcell - Istanbul Technical University Researcher Funding Program. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/jibril-2022-anec-amharic-named-entity-corpus/