Welcome to TIGRIS Virtual Lab Project Area

e-ŠNUNNA

Preliminary Results


The project "e-ŠNUNNA" offers a significant example of integration between Humanities (Assyriology) and Computer Science, adopting sofware application for Text Analysis (TaLTaC2) and Data Mining tools (ASTEC). Moreover, ENEA-GRID infrastructure and the computational power of ENEA-CRESCO HPS systems provide a perfect environment to execute experiments and data analysis tools, and offer also the possibility of ubiquitous access to our research, by means of the ENEA web portal for TIGRIS Projects.

The corpus of 50 letters from Tell Harmal of the Project e-ŠNUNNA belongs to the period of King Ibal-pi-El II (1779-1765 b.C.). The school of Šaduppum was excellent and these letters show variety of terms and forms. For this reason, this corpus is interesting for a study of the EŠnunna old-Babylonian grammar. These letters are also important from a sociological-historical point of view, in order to investigate through the procedures and the hierarchized administrative structure of a town and its organizational structure.
For supporting experts with Document Clustering of assyriological e-texts, ENEA developed ASTEC (ASsyriological TExt Clustering) a tool integrated in ENEA-GRID, able to analyze transliterated cuneiform corpora. EŠnunna e-texts have been processed and then clustering task has been executed, employing the K-means clustering algorithm, wich partitioned the corpus into groups (i.e., clusters: the number of cluster varies from 2 to 15).
For our experiment, we employed two relevance measure for terms, providing different cluster identifications: TF (Term Frequency) and TF-IDF (Term Frequency - Inverse Document Frequency. TF is more sensitive to conjunctions, adverbs and to all terms frequently occurring in the majority of letters. TF-IDF tends to give most importance to highly specific terms, occurring only in few letters, excluding terms non relevant for topic identification.
Texts regarding the same officer are often grouped together by ASTEC tool, and relations between officers and "the Palace" are dominant in many cluster, as expected in a corpus composed by letters referring to activities and actions performed by public officers, in a provincial town in relation to the capital. Furthermore, ASTEC clusters letters related to goods and in particular barley, showing how barley was employed (as food, drink costituent, seeding element, trading element).
Other point of contact are:
- the "Palace" (the EŠnunna palace, i.e., the capital palace, the palace of the King): the recipient of messengers and needs of animals (cattle) of quality;
- fields: as a place to grow barley; as a form of payment; for problems of irrigation;
- officers: busy in solving problems, however, requiring the intervention of "the Palace";
- un unbalanced, but mutual, relationship between the King and the officers, in which the king protection is corresponding to officer obedience;
- and conversely, a balanced mutual relationship among local officers. This general rules applies also in Tell Harmal, as showed by a particularly interesting group of letters, where a relationship of equality and collaboration between sender and receiver is established and expressed in sentence like "so NP1 and NP2, who love you, (speak)", where the 'officer(s) who love(s) is/are usually the sender(s) and the 'loved officer" is the receiver. The presence of the term "athutu" (expressing friendly relationship between people of equal standing) confirms that the relationship is equal.

In conclusion, findings show that document clustering on transliterated corpora is highly-effective in discovering high quality groups and in highlighting interesting relations among data. From clustering results it emerges that the e-Šnunna Corpus provides a significant evidence for the relationships among citizens, local authorities and the Palace, and offers important elements for a study of the EŠnunna old-Babylonian grammar.


(Tablets have been translitered and then some characters have been substitued using some latin characters (see text encoding).


To collaborate or for more information please send an email to "contacts"