sumerian-lemmaparser

Analyzes a Sumerian corpus to associate any lemmata it contains with the words that the lemmata rows refer to. For instance, in the corpus, especially with older transliterations, there are lemma comments immediately following each line of transliterated text, e.g.:

4. GAN2 ur-{gesz}gigir nu-banda3 gu4
#lem: iku[unit]; PN; nubanda[overseer]; gud[ox]

These lemmata are very helpful for learning the language, but are also useful for testing natural-language algorithms to test recall. In the above case, the transliterated Sumerian words are matched up with either their stems and a personal name (PN) is identified.

This repository creates a "mini-shell" that downloads and unzips the Sumerian corpus from CDLI, the Cuneiform Digital Library Initiative (http://www.cdli.ucla.edu), unzips it, parses the corpus to associate any lemmata it contains with the rows the lemmata describe, and then provides a shell-like interface for querying the data.

To get started, do a make all to download and prepare the corpus. Then, you can run python tag.py at the shell to begin. The tag.py script will parse the corpus. At the time of writing, there are 53146 tablets out of 85458 that contain at least one lemma. The processing of the corpus may take a couple moments; please be patient. Once completed, the script will present a Sumerian$ prompt.

At this point, you have the following options:

• word <sumerian_word>: This will return the lemmata for the specified word. For instance, to get all lemmata for the word "kiszib3":

  Sumerian$ word kiszib3 ... kiszib3 (19038 lemmatized attestations) kiszib = seal kiszib = sealed tablet dub = tablet kiszib3 = seal kiszib3 = hand kiszib3! (6 lemmatized attestations) kiszib = seal kiszib3!(URUDU) (1 lemmatized attestations) kiszib = seal ...

Similarly, to return lemmata for the partial name "ab-ba-sag10":

Sumerian$ word ab-ba-sag10
	a-ab-ba-sag10 (1 lemmatized attestations)
		PN
	ab-ba-sag10 (132 lemmatized attestations)
		PN
		FN
	ab-ba-sag10# (5 lemmatized attestations)
		PN
		FN
	ab-ba-sag10#? (2 lemmatized attestations)
		PN
		FN
	...

You can enter word by itself to get all lemmata for all words in the corpus, but I wouldn't recommend it.

• line <word> [<word> ...]: This will return the lemmata for all words from all lines that match the specified sequence of words. For every word in every matched line, the available lemmata will be expanded. For instance:

  Sumerian$ line sza3-asz-ru-um{ki} ba-hul ... 4. iti nesag2 mu sza3-asz-ru-um{ki} ba-hul iti (1 lemmatized attestations) itud = month nesag2 (1 lemmatized attestations) nesaj = offering mu (1 lemmatized attestations) mu = year sza3-asz-ru-um{ki} (1 lemmatized attestations) GN ba-hul (1 lemmatized attestations) hulu = destroy 1. mu sza3-asz-ru-um{ki} ba-hul mu (1 lemmatized attestations) mu = year sza3-asz-ru-um{ki} (1 lemmatized attestations) GN ba-hul (1 lemmatized attestations) hulu = destroy 10. mu us2-sa sza3-asz-ru-um{ki} ba-hul mu (1 lemmatized attestations) mu = year us2-sa (1 lemmatized attestations) us = follow sza3-asz-ru-um{ki} (1 lemmatized attestations) GN ba-hul (1 lemmatized attestations) hulu = destroy ...

• You can enter line by itself to get all lemmata for all lines in the corpus, but I really wouldn't recommend it.

• tablet <name>: This will return all lines from the tablet with the specified name. The name of the tablet should include the leading P identifier.

• help: This will return a brief message displaying all available commands.

• You can quit the shell by entering exit or quit.