metadata

license: mit
task_categories:
  - text-generation
language:
  - en
pretty_name: English-translated Akkadian Corpus
size_categories:
  - 1K<n<10K

Akkadian English Corpus

This dataset is a cleaned English-translated Akkadian language dataset. This dataset can and has been used for text generation tasks, for example to fine-tune LLMs.

How it was generated

Please visit my repo on Github which explains the steps that were taken to prepare this dataset for a text generation task.

At a high level, these are steps that were taken:

Sourced a high-quality dataset of English-translated Akkadian by experts
Enforced a minimum line length
Removed duplicate lines
Removed textual notes and other generic notes within parantheses
Inserted translation notes and literal notes in place (preserving grammar and adding clarity to the corpus)

Credit

Credit for the aggregation of the raw data belongs to the Akkademia project. Specifically, the exact data file used as the starting dataset is linked here and was also used to train their SOTA neural machine translation Akkadian->English model as described in their recent paper Gutherz et al. 2023 [1].

Credit for the original source of the raw data belongs to the incredible Open Richly Annotated Cuneiform Corpus (ORACC) project [2]. Specifically, as noted by the Akkademia project above, the RINAP 1, 3, 4, and 5 datasets are the source of the original raw data.

Citations

[1] Gai Gutherz, Shai Gordin, Luis Sáenz, Omer Levy, Jonathan Berant, Translating Akkadian to English with neural machine translation, PNAS Nexus, Volume 2, Issue 5, May 2023, pgad096, https://doi.org/10.1093/pnasnexus/pgad096
[2] Jamie Novotny, Eleanor Robson, Steve Tinney, Niek Veldhuis, et al. Open Richly Annotated Cuneiform Corpus, http://oracc.org