We propose several ways of reusing subword embeddings and other weights in subwordaware neural language models. The proposed techniques do not benefit a competitive character-aware model, but some of them improve the performance of syllableand morpheme-aware models while showing significant reductions in model sizes. We discover a simple hands-on principle: in a multilayer input embedding model...

Finite State parsing tools are generally optimized for run-time efficiency. But a field linguist needs compile-time efficiency, so that incremental changes can be made quickly as new morphemes are discovered and grammar rules revised. Using an available finite state toolkit, the Xerox xfst program, I show how incremental changes can be rapidly compiled by extracting the set of morphemes which c...

ParaMor automatically learns morphological paradigms from unlabelled text, and uses them to annotate word forms with morpheme boundaries. ParaMor competed in the English and German tracks of Morpho Challenge 2007 (Kurimo et al., 2008). In English, ParaMor’s balanced precision and recall outperform at F1 an already sophisticated baseline induction algorithm, Morfessor (Creutz, 2006). In German, ...

In this paper, we describe the use of either words or morphemes as lexical modeling units and the use of either graphemes or phonemes as phoneticmodeling units for Arabic automatic speech recognition (ASR). We designed four Arabic ASR systems: two word-based systems and two morpheme-based systems. Experimental results using these four systems show that they have comparable state-of-the-art perf...

This paper introduces an unsupervised morphological segmentation algorithm that shows robust performance for four languages with different levels of morphological complexity. In particular, our algorithm outperforms Goldsmith’s Linguistica and Creutz and Lagus’s Morphessor for English and Bengali, and achieves performance that is comparable to the best results for all three PASCAL evaluation da...

persian follows a concatenative morphology, where new morphemes are generated by chaining different morphemes together to form a new compound word.  whenever, two morphemes bind to form a new morpheme, there is a possibility that the syllables at the morpheme boundaries undergo structural change.  this study suggests that these syllabic alterations may be captured using a finite state approach....

the persian morpheme ra has attracted the attention of many linguists including karimi (1989), dabir-moghaddam (1990) and ghomeshi (1996) among others. karimi takes ra as the accusative case marker, the presence of which on subjects and objects of prepositions render the sentence ungrammatical. according to ghomeshi (1996), it marks dps functioning as vp-level topics. dabir-moghaddam (1990) ana...

Abstract In this paper, we aim to account for the distribution and interpretation of a novel class free choice items in Romanian, which refer as additive (ADD-FCIs). We show that internal composition ADD-FCIs, well their distribution, differs from attested other paradigms discussed literature. Morphologically, ADD-FCIs are more complex variant regular universal FCIs, by virtue an additional mor...

This research aims to discover the category of words that are taken from an object speech. The researcher focuses on occurrence frequency and process have been used in applied a descriptive qualitative method. collected some through speech, then continued classify them into morphological analysis, specifically derivational morpheme inflectional morpheme. employed Victoria Fromkin theory analyze...