|
Elements of machine learning in the T-parser system of facts extraction
I. M. Adamovich, O. I. Volkov Institute of Informatics Problems, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation
Abstract:
The article focuses on the further development of the system of facts automatic extraction from historical texts T-parser which is a component of the technology of historical and biographical research automation. The article outlines the ways to increase the parsing speed by using machine learning. The chosen forms of machine learning are described and reasoned and the possible problems are formulated. The classification of parsing bifurcations is given. The mechanism of filtering for the precedent database creation based on the methods of statistical quality control on an alternative basis is described and reasoned. The description of the updated parsing algorithm and experimental verification of its effectiveness in comparison with the previous version carried out with real historical texts are adduced. The results of experiments which confirm high efficiency of the updated algorithm and its applicability to the technology of historical and biographical research automation are described. The technology is intended for a broad range of nonprofessional users, which is topical with regard to the increasing public interest to family history.
Keywords:
facts extraction from texts, machine learning, bifurcation, statistical quality control, training set.
Received: 15.05.2018
Citation:
I. M. Adamovich, O. I. Volkov, “Elements of machine learning in the T-parser system of facts extraction”, Sistemy i Sredstva Inform., 28:4 (2018), 145–155
Linking options:
https://www.mathnet.ru/eng/ssi614 https://www.mathnet.ru/eng/ssi/v28/i4/p145
|
Statistics & downloads: |
Abstract page: | 152 | Full-text PDF : | 33 | References: | 21 |
|