ParaMor: Finding Paradigms across Morphology

Abstract

Our algorithm, ParaMor, fared well in Morpho Challenge 2007 (Kurimo et al., 2007), a peer operated competition pitting against one another algorithms designed to discover the morphological structure of natural languages from nothing more than raw text. ParaMor constructs sets of affixes closely mimicking the paradigms of a language, and, with these structures in hand, annotates word forms with morpheme boundaries. Of the four language tracks in Morpho Challenge 2007, we entered ParaMor in English and German. Morpho Challenge 2007 evaluated systems on their precision, recall, and balanced F 1 at identifying morphological processes, whether those processes mark derivational morphology or inflectional features. In English, ParaMor's balanced precision and recall outperform at F 1 an already sophisticated baseline induction algorithm, Morfessor (Creutz, 2006). ParaMor placed fourth in English overall. In German, ParaMor suffers from a low morpheme recall. But combining ParaMor's analyses with analyses from Morfessor results in a set of analyses that outperform either algorithm alone, and that place first in F 1 among all algorithms submitted

Making Morphologies the “Easy” Way

“…Such languages have a complex morphology, which cannot be covered by hand-made lexical resources. Some studies aim at solving this problem by learning inflectional paradigms from raw text corpora by clustering word forms in the corpus and analyzing the resulting clusters ( [9,8,3]). Other unsupervised methods applied to morphology induction are that of [15], [6] and [5], the latter using morphemes to encode a corpus by grouping morphemes into structures, called signatures, representing inflectional paradigms.…”

Evaluating an agglutinative segmentation model for ParaMor

“…The baseline ParaMor algorithm which we extend here competed in the English and German tracks of Morpho Challenge 2007 (Monson et al, 2007b). The peer operated competitions of the Morpho Challenge series standardize the evaluation of unsupervised morphology induction algorithms (Kurimo et al, 2007a;.…”

“…Although Goldsmith (2001) and Goldsmith and Hu (2004) discuss ideas for segmenting individual words into more than two morphemes, the implemented Linguistica algorithm, as presented in Goldsmith (2006), permits at most a single morpheme boundary in each word. Second, ParaMor decouples the task of paradigm identification from that of word segmentation (Monson et al, 2007b). In contrast, morphology models in Linguistica inherently encode both a belief about paradigm structure on individual words as well as a segmentation of those words.…”

“…Indeed, in our random sample of 100 schemes, 51 of the 59 schemes with morpheme boundary errors incorrectly hypothesized a boundary stem-internally. For this reason, the baseline ParaMor algorithm already discarded schemes that likely misplace a boundary steminternally (Monson et al, 2007b). Although there are fewer schemes that misplace a morpheme boundary suffix-internally, suffix-internal error schemes contain short suffixes that can generalize to segment a large number of word forms.…”

Using Unsupervised Paradigm Acquisition for Prefixes

“…In addition, there are 7 new papers on this topic by Bernhard (2007), Bordag (2007), Chan (2007), McNamee (2007), Monson et al (2007), Pitler and Keshava (2007), and Tepper (2007). The approach described in this paper is a direct extension of Zeman (2007) and we will frequently refer to him.…”