Metodika pro převod strukturovaných znalostí z oboru dialektologie do strojového učení

The methodology addresses the preparation and utilization of dialect data in
dialectology through modern Machine Learning technologies. It focuses on the
processes of consolidating, standardizing, and structuring audial and textual
materials, which form the foundation for developing automatic speech
transcription tools. The core of the study presents procedures applicable to the
digitization and normalization of textual data and it includes a detailed
description of dialect documentation in the field, emphasizing various
exploratory methods, including digital archiving and cataloging of recordings.
The methodology connects theoretical knowledge on the collection and processing
of dialect material with practical procedures that involve the deployment of
Machine Learning. Emphasis is placed on an interdisciplinary approach that
combines linguistic expertise with technological tools for workflow automation.
The methodology also includes procedures for visualizing dialectological data
using thematic cartography, leading to the creation of interactive sound dialect
maps and web-based atlases. This document serves not only as a practical guide
for preparing specific linguistic material but also as inspiration for other
research teams, both in dialectology and in the broader integration of Machine
Learning into the humanities.