Arabic dialect classification has been an important and challenging problem for Arabic language processing, especially for social media text analysis and machine translation. In this paper we propose an approach to improving Arabic dialect classification with semi-supervised learning: multiple classifiers are trained with weakly supervised, strongly supervised, and unsupervised data. Their comb...

This research applies dialectometric methods to purely syntactic dialect data. It will be shown that there is geographic cohesion in syntactic variation when viewed in the aggregate. The amount of syntactic variation which can be accounted for by geography will be determined. Dialectometric techniques will be used to develop an additive measure of syntactic differences. Multidimensional scaling...

The present study explores the phonetic implementation of accentual H(igh) tone of the Akita dialect of Japanese, specifically its timing control. Lexical accent in the dialect is implemented as an eminence in F0. The conspicuous feature is that F0 does not fall sharply as in Tokyo Japanese. Two hypotheses are examined concerning the extent of the duration of the pitch summit. One is that the H...

In dialectometry several approaches exist to automatically generate dialect maps on the basis of calculated dissimilarities between locations (using survey data). Examples for Dutch are Heeringa (2004), who calculates dialect maps on the basis of aggregated phonetic dissimilarities, and Spruit (2008), who calculates dialect maps on the basis of aggregated syntactic dissimilarities. In both exam...

We investigate national dialect identification, the task of classifying English documents according to their country of origin. We use corpora of known national origin as a proxy for national dialect. In order to identify general (as opposed to corpus-specific) characteristics of national dialects of English, we make use of a variety of corpora of different sources, with inter-corpus variation ...

Research has shown that automatic speech recognition (ASR) performance typically decreases when evaluated on a dialectal variation of the same language that was not used for training its models. Similarly, models simultaneously trained on a group of dialects tend to underperform when compared to dialect-specific models. When trying to decide which dialect-specific model (recognizer) to use to d...

Vocalization, diactritization or diacritics restoration is one of the major challenges in Arabic natural language processing. Algiers dialect is also concerned by this issue. In this paper, we present an automatic diacritization system for standard and dialect Arabic texts based on statistical approach. The idea is to use available tools in statistical machine translation to build such a system...

This study investigates the ability of speakers to imitate features from another dialect of their native language. Speakers from two different Norwegian dialects spoken in Stavanger and Trondheim produced speech material using their own and each other’s dialect. Acoustic analysis involved the occurrence of preaspiration in vowel – voiceless stop contexts. Measurements revealed that preaspiratio...

This study investigates how onset and sentence positioning affect the realization of Tone 1 in two dialects of Taiwan Mandarin. Results showed that the central dialect was higher in register when placed in isolation, but lower when placed in a sentential context. When there was a tonal mismatch, coarticulatory effects were more robust in the northern dialect, implying that speakers of the centr...

The Tarascan or phurépecha language contrasts aspirated stops from non-aspirated word initially, but this contrast was hardly heard in Hiwatzio, a lake dialect (Friedrich 1971:177). We measured the length of the aspirated feature following stops and vowels /i a u/'s formants F1 and F2 so as to identify differences among subjects. The results show that adult men speakers from a mountain dialect ...