The guitar fingering problem involves the automatic generation of left-hand fingerings for guitar pieces. This thesis covers the extension of solutions which map the problem to finding an optimal path in a graph. The presented approach is based on a linear-chain conditional random field (CRF) and can generate fingerings to any guitar piece regardless of its note structure. It is the first approach to incorporate knowledge on the physical dimensions of guitars and guitarists. The generated fingerings were mostly appropriate regarding the choice of fretboard positions whereas the quality of the finger assignments was lacking. Compared to measuring distances on the guitar by a number of frets, the inclusion of the physical dimensions led to a slightly more accurate choice of fretboard positions. Major improvements are expected once higher-order dependencies between more than two subsequent notes are modeled

A Linear-Chain Conditional Random Field Approach to the Guitar Fingering Problem