Citation Context

...an (habit’s). A segmentation based approach models changed stems as distinct morphemes. In Morpho Challenge 2008, we introduced an unsupervised model for morpheme segmentation and allomorphy learning =-=[10]-=-. In [11], some modifications to the model (now referred to as Allomorfessor Alpha) were suggested. In this paper we describe and evaluate the modified Allomorfessor model (referred to as Allomorfesso...

Citation Context

...oldsmith, 2001; Goldsmith, 2006) and Morfessor (Creutz and Lagus, 2005). Linguistica constitutes the first real attempt to use the concept of MDL (Minimum Description Length) for encoding a complete corpus w.r.t. morphemes using as few bits as possible, thus trying to achieve the best possible affix and stem recognition. In (Creutz and Lagus, 2005), the authors also use the MDL approach without restricting the analysis of a word into only one facultative prefix, only one stem and only one suffix as is the case in (Goldsmith, 2001). Morfessor has later been extended for treating allomorphisms (Kohonen et al., 2009). Later, in (Golenia et al., 2009), MDL is used to pre-select possible stems for given forms; the stems are separated from the rest and the remaining strings considered possible affixes. These possible affixes are then first broken into substrings and then re-assembled according to a metric relying on the number of these substrings’ occurrences. Spiegler et al. (2010), Bernhard (2008) and Keshava (2006) describe methods inspired by the work of Harris (1955) and extensions thereof (Hafer and Weiss, 1974; Dejean, 1998). These approaches focus on transition probabilities and letter successor var...

Citation Context

...over allomorphic variation (e.g. ’lla’ and ’llä’ are different, contextdependent surface forms of a morpheme with roughly the meaning ’on’). In an extension of Morfessor Baseline called Allomorfessor =-=[10]-=- this problem is considered. It is assumed that a linguistic (latent) morpheme can have orthographical variants. The variants are coded using operations (insertions, deletions and replacements) applie...

Citation Context

...hip of the affixes are updated one more time by considering quality of shared root distributions. Once affix list is acquired, stemming of word is started from the longest common affix. Allomorfessor =-=[17]-=-, is a modified version of morfessor [2] to solve problems caused by allomorphy. MAP is done by modeling a lexicon of the words in the corpus instead of modeling the original corpus. A new notion muta...