1 |
Algorithmic advancements in Computational Historical Linguistics ...
|
|
|
|
Abstract:
Computergestützte Methoden in der historischen Linguistik haben in den letzten Jahren einen großen Aufschwung erlebt. Die wachsende Verfügbarkeit maschinenlesbarer Daten förderten diese Entwicklung ebenso wie die zunehmende Leistungsfähigkeit von Computern. Die in dieser Forschung verwendeten Berechnungsmethoden stammen aus verschiedenen wissenschaftlichen Disziplinen, wobei Methoden aus der Bioinformatik sicherlich die Initialzündung gaben. Diese Arbeit, die sich von Fortschritten in angrenzenden Gebieten inspirieren lässt, zielt darauf ab, die bestehenden Berechnungsmethoden in verschiedenen Bereichen der computergestützten historischen Linguistik zu verbessern. Mit Hilfe von Fortschritten aus der Forschung aus dem maschinellen Lernen und der Computerlinguistik wird hier eine neue Trainingsmethode für Algorithmen zur Kognatenerkennung vorgestellt. Diese Methode erreicht an vielen Stellen die besten Ergebnisse im Bereich der Kognatenerkennung. Außerdem kann das neue Trainingsschema die Rechenzeit erheblich ... : The use of computational methods in historical linguistics has seen a large boost in recent years. An increasing availability of machine readable data and the growing power of computers fostered this development. While the computational methods which are used in this research stem from different scientific disciplines, a lot of tools from computational biology have found their way into this research. Drawing inspiration from advancements in related fields, this thesis aims at improving existing computational methods in different disciplines of computational historical linguistics. Using advancements from machine learning and natural language processing research, I present an updated training regime for cognate detection algorithms. Besides achieving state of the art performance in a cognate clustering task, the updated training scheme considerably improved computation time. Following up on these results, I develop a novel combination of tools from bioinformatics and historical linguistics is developed. By ...
|
|
Keyword:
004; 400; Historische Sprachwissenschaft , Hidden-Markov-Modell , Bayes-Inferenz , Markov-Ketten-Monte-Carlo-Verfahren
|
|
URL: https://publikationen.uni-tuebingen.de/xmlui/handle/10900/118701 https://dx.doi.org/10.15496/publikation-60075
|
|
BASE
|
|
Hide details
|
|
4 |
Algorithmic advancements in computational historical linguistics
|
|
|
|
BASE
|
|
Show details
|
|
5 |
Algorithmic advancements in Computational Historical Linguistics
|
|
|
|
BASE
|
|
Show details
|
|
7 |
NorthEuraLex: a wide-coverage lexical database of Northern Eurasia
|
|
|
|
BASE
|
|
Show details
|
|
8 |
NorthEuraLex : a wide-coverage lexical database of Northern Eurasia
|
|
|
|
BASE
|
|
Show details
|
|
9 |
Are Automatic Methods for Cognate Detection Good Enough for Phylogenetic Reconstruction in Historical Linguistics? ...
|
|
|
|
BASE
|
|
Show details
|
|
12 |
Are Automatic Methods for Cognate Detection Good Enough for Phylogenetic Reconstruction in Historical Linguistics? ...
|
|
|
|
BASE
|
|
Show details
|
|
13 |
Are automatic methods for cognate detection good enough for phylogenetic reconstruction in historical linguistics?
|
|
|
|
BASE
|
|
Show details
|
|
14 |
Fast and unsupervised methods for multilingual cognate clustering ...
|
|
|
|
BASE
|
|
Show details
|
|
16 |
Reading corpora as an instrument for studying a relevance-based account of language processing. A case study using a reading corpus of German jurisdictional texts
|
|
|
|
DNB Subject Category Language
|
|
Show details
|
|
20 |
Proceedings of the 6th Conference on Quantitative Investigations in Theoretical Linguistics
|
|
|
|
BASE
|
|
Show details
|
|
|
|