Description of file "BPC_saa_01_05_10_13_15_16_17_18_19_21_28082023.csv":

Manual annotations made for: SAA 01,05,15,21, and small percentages of each of SAA 10,13,16,17,18,19.

Explanation of various header fields in the CSV file:

The fields assume a BPC with components [Verb], [Base Preposition], [Body Part Term], and optionally [Direct Object] and [Oblique Object].
These components are labelled in the headers as follows:

Verb - tword
Base Preposition - rword
Body Part Term - sword
Direct Object - doword
Oblique Object - ooword

The lemmas for each of these component words are labelled with fields like those above but with 'lemma' substituted for 'word'.
E.g. tlemma, rlemma, slemma, etc.

Other explanations of fields:

date - Estimated date of composition of letter according to editors of the relevant SAA volume
dialect - Dialect of Akkadian the letter is written in
script - Script of Akkadian the letter is written in
designation - Volume and letter number the attestation appears in
bad_analysis - if the putative example is actually not a BPC (usually because of syntactic misanalysis by model)
ruler - Assyrian king under which letter was written, if known
etype - Syntactic dependency type (Indirect object or Oblique object) relating compound preposition to verb
pcom - url of the letter that the attestation appears in
PP_type - Whether the compound preposition expresses a required, external argument of the verb (EA) or is an optional adjunct (AJ)
Translation - optional translation of BPC in context
Speaker - Identity of speaker of sentence containing the BPC, which may on occasion be different from the letter's author due to quoted speech.
Note that this field and the one below were checked only for a portion of the BPC's.
Addressee - Identity of addressee of sentence containing the BPC

The remaining fields are either self-evident, or they are described in Ong and Gordin (forthcoming) "Neo-Assyrian Metaphors through the Telescope: Linguistic Patterns involving Body Part Constructions in the State Archives Letter Corpus".

Description of files of the form "akk-mcong-ud*":

These are the conllu and ttl files associated with the SAA volumes.

Description of "sparql_Cxn_BCP.txt":

Contains the sparql query used to find BCP's in the ttl files. Known to work on TriplyDB.

Description of "ak_norm_conllu.zip":

Contains conllu files used to train the spaCy language model, including training, development, and test data

Description of "ak_norm_spacy.zip":

Contains files used to train the language model as in "ak_norm_conllu.zip", converted to spaCy's required binary format

Description of "ak_AkkParser_Norm_1_2_5_8_9_10_13_15_16_17_18_19_21_anzu_barutu_rinap4_tcmaassur-0.0.0.tar.gz":

Contains files defining the spaCy language model used in this project.