Morfessor demonstration

This demonstration produces a morph segmentation of words you type in. The program has several segmentation models and corpora from which you can choose one. In order to run larger scale experiments on your own data, download the Morfessor software.

Write words to analyze:

Instructions

Select model type and training corpus:

Model type	Corpus
Categories-ML: Probabilistic morph segmentation model that learns morph categories (prefix, stem, and suffix) and sequential dependencies between these. This produces a better segmentation than the baseline model, when evaluated against a morphological "gold-standard" segmentation. Baseline: Morph segmentation model inspired by the Minimum Description Length principle. More on methods	Finnish: 1 million word forms extracted from a 16 million word corpus 68 000 word forms extracted from a 250 000 word corpus English: 110 000 word forms extracted from a 12 million word corpus 21 000 word forms extracted from a 250 000 word corpus Swedish: 90 000 word forms extracted from a 1 million word corpus By varying the size of the corpus, you can observe the effects on the resulting segments (for some languages).

Model type

Corpus

Categories-ML:

Probabilistic morph segmentation model that learns morph categories (prefix, stem, and suffix) and sequential dependencies between these. This produces a better segmentation than the baseline model, when evaluated against a morphological "gold-standard" segmentation.

Baseline:

Morph segmentation model inspired by the Minimum Description Length principle.