Automated Phonological Transcription of Akkadian Cuneiform Text

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 3528–3534 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 3528 Automated Phonological Transcription of Akkadian Cuneiform Text Aleksi Sahala1, Miikka Silfverberg1,3, Antti Arppe2, Krister Lind´en1 1University of Helsinki, 2University of Alberta, 3 University of British Columbia {aleksi.sahala, krister.linden}@helsinki.fi, msilfver@mail.ubc.ca,arppe@ualberta.ca Abstract Akkadian was an East-Semitic language spoken in ancient Mesopotamia. The language is attested on hundreds of thousands of cuneiform clay tablets. Several Akkadian text corpora contain only the transliterated text. In this paper, we investigate automated phonological transcription of the transliterated corpora. The phonological transcription provides a linguistically appealing form to represent Akkadian, because the transcription is normalized according to the grammatical description of a given dialect and explicitly shows the Akkadian renderings for Sumerian logograms. Because cuneiform text does not mark the inflection for logograms, the inflected form needs to be inferred from the sentence context. To the best of our knowledge, this is the first documented attempt to automatically transcribe Akkadian. Using a context-aware neural network model, we are able to automatically transcribe syllabic tokens at near human performance with 96% recall @ 3, while the logogram transcription remains more challenging at 82% recall @ 3. Keywords: transcription, Akkadian, cuneiform script 1. Introduction Akkadian was an East-Semitic language spoken in ancient Mesopotamia. The language is attested from hundreds of thousands of cuneiform clay tablets and their fragments ex- cavated from modern day Iraq and the surrounding regions. Although the first exemplars of Akkadian texts date back to 2400 BCE, two later dialects, i.e. Babylonian (1900 BCE - 100 CE) and Assyrian (1950 – 600 BCE) provide the largest source of available text material. Akkadian was written in the cuneiform script mostly on clay tablets as the example shown in Fig. 1. In modern electronic text cor- pora, cuneiform text is represented in two distict latiniza- tions: (1) sign-to-sign level graphemic transliteration and (2) phonemic transcription based on an approximation of the Akkadian language and its reconstructed sound system (Kouwenberg, 2011). A large number of transliterated Akkadian texts are freely available in digitized corpora, however, there exist far fewer phonologically transcribed texts. The principal reason for this discrepancy is that the transcription process is labor- intensive, and the transcription itself is not considered very useful for scholarly interpetation of the cuneiform texts. It is, however, indispensable for the development of compu- tational resources for Akkadian because cuneiform script often omits phonological elements which provide impor- tant information for language analysis tools. For example, vowel quantities are not marked consistently, and dialects may use slightly different writing conventions (Kouwen- berg, 2011). Moreover, the underlying readings of lo- gograms are not transliterated. This information is added during phonological transcription. In addition to compu- tational modeling, phonological transcriptions also play an important role in teaching of the Akkadian language. In this paper we investigate automated phonological tran- scription of transliterated Akkadian texts. Conceptually, transcription consists of two subtasks. For syllabically spelled words which contain …

Content truncated to 500 words