Unsupervised Learning of the Morphology of a Natural Language

By J. Goldsmith. International Conference on Computational Logic, 2001

893 citations recorded by the source corpus.

This study reports the results of using minimum description length (MDL) analysis to model unsupervised learning of the morphological segmentation of European languages, using corpora ranging in size from 5,000 words to 500,000 words. We develop a set of heuristics that rapidly develop a probabilistic morphological grammar, and use MDL as our primary tool to determine whether the modifications proposed by the heuristics will be adopted or not. The resulting grammar matches well the analysis that would be developed by a human morphologist. In the final section, we discuss the relationship of this style of MDL grammatical analysis to the notion of evaluation metric in early generative grammar.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/9f834ee11902ada79b874e7fe5072159d72a0f9f

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

A Bayesian framework for word segmentation: exploring the effects of context. (2009, positive): “MDL is a particular formulation of Bayesian learning that has been used successfully in a number of other areas of language acquisition as well (Creutz & Lagus, 2002; Dowman, 2000; Ellison, 1994; Goldsmith, 2001; Goldwater & Johnson, 2004).”
Morphology-aware statistical machine translation based on morphs induced in an unsupervised manner (2007, background): “This distinguishes Morfes sor from other algorithms that pose harder restrictions on the possible structures of words, such that each word is assumed to consist of one stem optionally followed by a suffix; see, e.g., Goldsmith (2001) .”
Unsupervised Lexicon Discovery from Acoustic Input (2015, background): “…of discovering a compact lexicon which explains the distribution of phoneme sequences in the input (Cartwright and Brent, 1994; Brent, 1999a; Goldsmith, 2001; Argamon et al., 2004; Goldsmith, 2006; Creutz and Lagus, 2007; Goldwater et al., 2009; Mochihashi et al., 2009; Elsner et al., 2013;…”
Unsupervised Morpheme Segmentation and Morphology Induction from Text Corpora Using Morfessor 1.0 (2005, methodology): “This distinguishes Morfessor from resembling models, e.g., Goldsmith (2001), which assume that words consist of one stem possibly followed by a suffix and possibly preceded by a prefix.”
Simplicity: a unifying principle in cognitive science? (2003, methodology): “analysis [ 30 ], segmentation [31,24] and grammar induction [32,33]”
Between words and characters: A Brief History of Open-Vocabulary Modeling and Tokenization in NLP (2021, methodology): “MDL-based approaches were proliﬁcally adapted and expanded for unsupervised morphological segmentation, as in Linguistica (Goldsmith, 2001), and found to generate segmentations with high correlations to mor-pheme boundaries on English and Romance languages (Baroni, 2000; Goldsmith, 2001).”
Unsupervised Learning by Program Synthesis (2015, background): “These include systems that learn the phonological form of morphological rules [1, 13, 24], systems that induce morphemes from corpora [17, 25], and systems that learn the productivity of different rules [26].”
Compositional-ly Derived Representations of Morphologically Complex Words in Distributional Semantics (2013, neutral): “Morphological induction systems use corpus-based methods to decide if two words are morphologically related and/or to segment words into morphemes (Dreyer and Eisner, 2011; Goldsmith, 2001; Goldwater and McClosky, 2005; Goldwater, 2006; Naradowsky and Goldwater, 2009; Wicen-towski, 2004).”
Unsupervised Multilingual Learning for Morphological Segmentation (2008, methodology): “Unsupervised Morphological Segmentation Unsupervised morphology is an active area of research (Schone and Jurafsky, 2000; Goldsmith, 2001; Adler and Elhadad, 2006; Creutz and Lagus, 2007; Dasgupta and Ng, 2007).”
Improving stemming for Arabic information retrieval: light stemming and co-occurrence analysis (2002, methodology): “For example, Goldsmith finds the best set of frequently occurring stems and suffixes using an information theoretic measure [ 20 ].”
Morfessor in the Morpho Challenge (2006, background): “In the following sections, some characteristics of the Morfessor model will be outlined and experimental results obtained in the morpheme segmentation as well as Finnish speech recognition task will be reported and discussed.”