Automation of archival documents meta tagging

D. Yu. Filimonov ; A. V. Svetlov ; O. A. Gorban ; M. V. Kosova ; E. M. Sheptukhina

Matematičeskaâ fizika i kompʹûternoe modelirovanie, Tome 23 (2020) no. 4, pp. 56-68.

Résumé

The main goal of this project is to create a corpus of documents from the “Mikhailovsky stanichny ataman” archival fund. The methods of corpus linguistics seem to be the most optimal in this case, since they involve the processing of a large number of texts in order to solve a wide variety of linguistic problems. Our group joined the team of philologists to provide the technical and software part of the project. The main task for us is to create a document corpus engine, that is, software that solves the tasks of storing a database of marked-up texts, executing queries to this database, and also providing users with a convenient interface for work that does not require special qualifications in the field of information technology. However, it is necessary to prepare documents for inclusion in the corpus: all texts must undergo special markup. There are many types of markup, and in the previous publications [6; 9] our group has already described the solution to the problem of morphological tagging. This article is about meta tagging. Meta tagging refers to the assignment of certain descriptive attributes to the text. In the case of office documents, these are such parameters as the type of document (genre), author (compiler), addressee, date and place of creation. Meta tagging is necessary for the implementation of the corpus search features, so that the researchers can receive text samples with specified external parameters: for example, texts of a certain type, created at a certain period, addressed to a certain addressee, etc. The archives of the “Mikhailovsky stanichny ataman” fund mainly contain documents from the Chanceries of the Don Army from the mid-18$^{\text{th}}$ to the first third of the 19$^{\text{th}}$ century, that's why there are not so many varieties of these documents. Moreover, these are mostly official documents, and they were written up according to certain templates, forms, the parameters of which can be relatively easily extracted from documents through preliminary analysis. This work is also carried out by the team of philologists from VolSU under the guidance of Professor O.A. Gorban. The result of their systematization of documents was the description of special speech markers of genre parameters for all document types in the archive. Thus, in our case, there is no need for heavy methods of statistical analysis or machine learning, it is enough to search for certain markers in the document. Moreover, the main marker in all reviewed documents is a direct indication of their type. Other markers are auxiliary elements of meta tagging. The paper is devoted to the description of the created application for determining the type of a document and its meta tagging by searching the text for certain regular expressions derived from the markers.

Détail

Keywords: automation of linguistic analysis, automation of meta tagging, graphical interface, corpus-based linguistics, regular expressions.