Graphemic Units Research Articles

Spelling Development of Korean 1st Graders with and without Developmental Dyslexia

Objectives: This study examined the word writing abilities of Korean 1st graders with and without developmental dyslexia (DD) considering meaning accessibility and the structure of a Korean syllable. Methods: Fifteen typically developing children (TD) and fifteen children with developmental dyslexia were assigned word writing tasks in the first and second semesters of the first year. The word writing task contains ten words and ten non-words containing forty syllable-initial and syllable-medial graphemes, and twenty-two syllable-final graphemes. The analysis of the results was conducted in grapheme units, and the accuracy for each word/nonword and the error rate for each structure of a syllable was determined. Results: The DD group showed significant growth in both words/nonwords writing accuracy by the second semester, and the error rates of syllable-initial, syllable-medial, and syllable-final graphemes significantly decreased from the first to the second semester for the DD group. The TD group showed no significant difference in word/non-words writing, and there was no significant difference in error rates according to the structure of a syllable. Both the DD and TD groups showed higher error rates in final graphemes in both the first and second semesters. Conclusion: The word writing abilities of the DD group showed significant growth between the first and second semesters but still had difficulties, particularly in the final grapheme writing. The TD group developed word writing first, and nonword writing was still in development. But final grapheme writing was difficult even for TD group. Follow-up studies are needed to closely observe the developmental processes comparing with decoding level matched TD children.

How long is long? Word length effects in reading correspond to minimal graphemic units: An MEG study in Bangla.

This paper presents a magnetoencephalography (MEG) study on reading in Bangla, an east Indo-Aryan language predominantly written in an abugida script. The study aims to uncover how visual stimuli are processed and mapped onto abstract linguistic representations in the brain. Specifically, we investigate the neural responses that correspond to word length in Bangla, a language with a unique orthography that introduces multiple ways to measure word length. Our results show that MEG signals localised in the anterior left fusiform gyrus, at around 130ms, are highly correlated with word length when measured in terms of the number of minimal graphemic units in the word rather than independent graphemic units (akśar) or phonemes. Our findings suggest that minimal graphemic units could serve as a suitable metric for measuring word length in non-alphabetic orthographies such as Bangla.

Do letters matter? The influence of spelling on acoustic duration.

The present article describes a modified and extended replication of a corpus study by Brewer (2008. Phonetic reflexes of orthographic characteristics in lexical representation. Tucson, AZ: University of Arizona PhD thesis) which reports differences in the acoustic duration of homophonous but heterographic sounds. The original findings point to a quantity effect of spelling on acoustic duration, i.e.,the more letters are used to spell a sound, the longer the sound's duration. Such a finding would have extensive theoretical implications and necessitate more research on how exactly spelling would come to influence speech production. However, the effects found by Brewer (2008) did not consistently reach statistical significance and the analysis did not include many of the covariates which are known by now to influence segment duration, rendering the robustness of the results at least questionable. Employing a more nuanced operationalization of graphemic units and a more advanced statistical analysis, the current replication fails to find the reported effect of letter quantity. Instead, we find an effect of graphemic complexity. Speakers realize consonants that do not have a visible graphemic correlate with shorter durations: the /s/ in tux is shorter that the /s/ in fuss. The effect presumably resembles orthographic visibility effects found in perception. In addition, our results highlight the need for a more rigorous approach to replicability in linguistics.

What does solid spelling reveal about cognition? Evidence from Middle Low German

Abstract This paper investigates the diachronic evolution of lexically complex graphemic units in Middle Low German – sequences that once occurred written as one word, but from today’s perspective are considered separate linguistic units. Examples are enwolde ‘did not want’ or isset ‘is it’. This phenomenon has received little attention, although it gives direct insight into the word concept of German and its diachronic change. The central question is what favors the perception of multiple words as a unit. Data from the Reference Corpus Middle Low German/Low Rhenish (1200–1650) show that it is mainly function words that occur in lexically complex graphemic units. Moreover, this study shows that besides from prosodic patterns, agreement and government relations reinforce lexical sequences to be perceived as linguistic units.

Conversational telephone speech recognition for Lithuanian

The research presented in the paper addresses conversational telephone speech recognition and keyword spotting for the Lithuanian language. Lithuanian can be considered a low e-resourced language as little transcribed audio data, and more generally, only limited linguistic resources are available electronically. Part of this research explores the impact of reducing the amount of linguistic knowledge and manual supervision when developing the transcription system. Since designing a pronunciation dictionary requires language-specific expertise, the need for manual supervision was assessed by comparing phonemic and graphemic units for acoustic modeling. Although the Lithuanian language is generally described in the linguistic literature with 56 phonemes, under low-resourced conditions some phonemes may not be sufficiently observed to be modeled. Therefore different phoneme inventories were explored to assess the effects of explicitly modeling diphthongs, affricates and soft consonants. The impact of using Web data for language modeling and additional untranscribed audio data for semi-supervised training was also measured. Out-of-vocabulary (OOV) keywords are a well-known challenge for keyword search. While word-based keyword search is quite effective for in-vocabulary words, OOV keywords are largely undetected. Morpheme-based subword units are compared with character n-gram-based units for their capacity to detect OOV keywords. Experimental results are reported for two training conditions defined in the IARPA Babel program: the full language pack and the very limited language pack, for which, respectively, 40 h and 3 h of transcribed training data are available. For both conditions, grapheme-based and phoneme-based models are shown to obtain comparable transcription and keyword spotting results. The use of Web texts for language modeling is shown to significantly improve both speech recognition and keyword spotting performance. Combining full-word and subword units leads to the best keyword spotting results.

The Graphic Syllable Writing Systemin Hunminjeongeum

The purpose of this study was to investigate the graphic system in Hunminjeongeum, Korean writing system. Hunminjeongeum is included into ‘alphasyllabary’ according to the type of writing system; on the first level, segmental units, and on the second level, a mixed alphabetic-syllabic type. In Hunminjeongeum, the graphic syllable, a crucial unit, could be defined as a graphemic unit which corresponds to the speech syllable, and it could also be a complex system in which graphemic units can be stacked together to form letter blocks. Moreover, the graphic syllable can facilitate the analysis of speech syllable, and due to the influence of Chinese characters, the speech syllable could be divided into three parts-initial, medial, and final. Also, the structure of graphic syllables has connection with Chinese character. Externally, both graphic syllables and Chinese characters have the same spatial shape, ‘square.’ Internally, the graphic syllable has the same structure; a bilateral symmetry, a top-and-bottom symmetry, and a combination of the two. In future, the distinctive natures of speech and graphic syllables will be sustained.

An investigation of the role of grapheme units in word recognition.

In most current models of word recognition, the word recognition process is assumed to be driven by the activation of letter units (i.e., that letters are the perceptual units in reading). An alternative possibility is that the word recognition process is driven by the activation of grapheme units, that is, that graphemes, rather than letters, are the perceptual units in reading. If so, there must be representational units for multiletter graphemes like CH and PH, which play a key role in this process. We examined this idea in four masked priming experiments. Primes were created by transposing, replacing entirely, or removing one component of either multiletter graphemes or two adjacent letters that each represented a grapheme, using both English and Spanish stimuli. In none of the experiments was there any evidence of differential priming effects depending on whether the two letters being manipulated formed a single grapheme or formed two separate graphemes. These data are most consistent with the idea that multiletter graphemes have no special status at the earliest stages of word processing and, therefore, that word recognition is, indeed, driven by the activation of units for individual letters.

A comparison of grapheme and phoneme-based units for Spanish spoken term detection

The ever-increasing volume of audio data available online through the world wide web means that automatic methods for indexing and search are becoming essential. Hidden Markov model (HMM) keyword spotting and lattice search techniques are the two most common approaches used by such systems. In keyword spotting, models or templates are defined for each search term prior to accessing the speech and used to find matches. Lattice search (referred to as spoken term detection), uses a pre-indexing of speech data in terms of word or sub-word units, which can then quickly be searched for arbitrary terms without referring to the original audio. In both cases, the search term can be modelled in terms of sub-word units, typically phonemes. For in-vocabulary words (i.e. words that appear in the pronunciation dictionary), the letter-to-sound conversion systems are accepted to work well. However, for out-of-vocabulary (OOV) search terms, letter-to-sound conversion must be used to generate a pronunciation for the search term. This is usually a hard decision (i.e. not probabilistic and with no possibility of backtracking), and errors introduced at this step are difficult to recover from. We therefore propose the direct use of graphemes (i.e., letter-based sub-word units) for acoustic modelling. This is expected to work particularly well in languages such as Spanish, where despite the letter-to-sound mapping being very regular, the correspondence is not one-to-one, and there will be benefits from avoiding hard decisions at early stages of processing. In this article, we compare three approaches for Spanish keyword spotting or spoken term detection, and within each of these we compare acoustic modelling based on phone and grapheme units. Experiments were performed using the Spanish geographical-domain A lbayzin corpus. Results achieved in the two approaches proposed for spoken term detection show us that trigrapheme units for acoustic modelling match or exceed the performance of phone-based acoustic models. In the method proposed for keyword spotting, the results achieved with each acoustic model are very similar.

Predicting children's word‐spelling difficulty for common English words from measures of orthographic transparency, phonemic and graphemic length and word frequency

Spencer demonstrated that spelling and reading difficulty for English words can be predicted from a number of factors, including word frequency, phonemic length and measures of orthographic depth and complexity. In this study, spelling difficulty of high frequency words was investigated across five year groups (ages 7 to 11 years) with a wide ranging series of data sources from which to determine word frequency values and orthographic depth measures. A regression model accounted for 52-66% of the variance for 7- to 11-year-olds and 72% of the variance for the lowest performing quartile group, irrespective of age. The most influential factor, phonetic difference (being the difference in the number of letters and phonemes in a word, and representing grapheme complexity), links the relative influence of large graphemic units in foundation literacy to a similar phenomenon, the 'whammy' effect, which provides support for serial processing in the dual route cascade model of word recognition in skilled readers (Rastle & Coltheart, 1998). The study supports recent research on European orthographies, which concludes that both orthographic depth and complexity contribute to delayed acquisition of foundation literacy skills.

Becoming literate in Hebrew: the grain size hypothesis and Semitic orthographic systems

Ziegler and Goswami (2006, see also Ziegler & Goswami, 2005) offer a cross-language reading theory that focuses on the grain size of lexical units which are converted into phonological structures during visual word recognition in different orthographic systems. Two axioms provide the basis for the grain size theory: first, that reading for meaning in any orthography primarily entails the recovery of the phonological information that is conveyed by the printed symbols; second, that any cross-language theory of reading should, therefore, focus on the ease or the difficulty of grapheme-to-phoneme computation given the consistency by which graphemic units represent phonological units in a given language. Ziegler and Goswami propose that reading in consistent orthographies involves small linguistic units, whereas reading in inconsistent orthographies requires the use of larger units as well. Hence, the grain size theory seems to present an improved and modern alternative to the Orthographic Depth Hypothesis (e.g. Frost, Katz & Bentin, 1987), which historically emerged from the classical dual-route models. The clear advantage of grain size over dualroute theory is that it examines the size of the computed phonological units and thereby allows the use of a continuous measure rather than a dichotomous concept such as ‘lexical’ or ‘prelexical’ phonology (see also Frost, 1998, for a discussion of skilled reading and grain size, according to the Strong Phonological Theory). Hebrew represents a very intriguing case for examining the theory promoted by Ziegler and Goswami. In Hebrew, the letters mostly represent consonants while most of the vowels can optionally be superimposed on the consonants as diacritical marks (‘points’). However, these diacritical marks are omitted from most adult reading material. Therefore, since different vowels may be inserted into the same string of consonants to form different words or nonwords, Hebrew unpointed print cannot specify a unique phonological unit, and a printed consonant string is phonologically ambiguous, often representing more than one word. English and Hebrew are both considered deep orthographies, but the ‘depth’ of the Hebrew orthography is different in character from that of English orthography. Whereas in English the opaque relations of spelling-to-sound are related to the inconsistency of letter clusters, in Hebrew opaque spelling-tosound connections arise simply from missing phonemic information, mainly vowel information. Note that when the diacritical marks are presented in the text, Hebrew orthography is entirely shallow, as the phonemic structure of the printed word can be easily assembled using simple grapheme-to-phoneme conversion rules. Extensive research has consistently revealed that reading in Hebrew involves an interplay of two computational processes which are defined by the size of the computed units. The first process is characterized by the conversion of units of single letters into consonantal information. The second process involves the use of morphological information, mainly word-pattern morphemes, which provide the missing vowels. Frost (1994) examined the effect of delaying the presentation of vowel marks under unpointed words in a naming task. The results showed that the Hebrew readers were willing to delay naming and wait for the vowel (pointed) information to appear,

Hyphenation can improve reading in acquired phonological dyslexia

Background: We describe JD, a person with severe phonological dyslexia. JD is good at reading words yet is extremely poor at reading nonwords. She shows no effect of word regularity on her reading performance. However, she has only a very mild general phonological deficit. Although it is known that teaching grapheme–phoneme correspondence rules and learning bigraph syllables can improve dyslexic reading, we investigate the possibility that segmenting the input string using hyphens can also improve reading. Parts of this paper were based on work reported by David A. O’Mara in his undergraduate thesis at the University of Dundee. We are indebted to Dr Allan Ward, GP, for his summary of JD’s medical history. Most of all, we are grateful to JD, who was always a most willing and cheerful participant. We thank Lynne Duncan, Nadine Martin, and Tim Shallice for their comments. Aims: We investigate the locus of JD's reading deficit and explore means of improving her reading. We examine the extent to which hyphenation can improve nonword reading. Methods & Procedures: We use a battery of tasks in a single‐case study. We test oral reading, writing and spelling ability, and lexical and semantic knowledge. We assess phonological processing using tests of repetition and phonological awareness. We focus on blending and segmentation, and test whether hyphenating the letter string can improve reading. Outcomes & Results: JD is very good at reading words (overall about 90%), but very poor at reading nonwords (overall about 10%). She makes no semantic errors and shows little effect of word regularity. JD has only a mild concomitant phonological deficit. She also has a very good digit span (8). JD can read affixes in isolation, and can also read nonwords made up of inappropriately affixed morphemes (e.g., “dismove”) if a hyphen is inserted at the affixation point (e.g., “dis‐move”). We found that hyphenation improves nonword reading in general, but particularly if it is the grapheme units that are separated by hyphens. Conclusions: We discuss the possibility that JD's phonological dyslexia arises from impaired graphemic parsing, and that affixes have inherent meaning for her. Hyphenation may be a therapeutic tool worthy of wider consideration for improving reading performance in people with dyslexia.

Word identification strategies in context of a gain and no‐gain remedial reading group

Abstract Two‐hundred seventy‐three responses to test words in context were analyzed to determine which word identification strategies were being used by students who made gains in a remedial reading program. Responses consisting of real and pseudowords came from pre‐ and post‐test situations. Results showed that during the course of the program the gain group became more analytical and used larger grapheme units in producing real word responses. Both groups, however, had internalized the meaningfulness of words in context since too few pseudowords for analysis occurred at post‐test time.

Compound Graphemic Units in Standard Babylonian Cuneiform Writing

1. The sign U81 to which has been attributed the syllabic value uD2 is often followed by the sign A which for grammatical reasons in some cases cannot have the syllabic value a. Therefore U8+A must be considered a compound graphemic unit which, as can be seen from the examples givell below, expresses not only /uDa/ but also /uDi/ and /U'U/.3 Since in some cases U8+A interchanges with U8, the latter can be explained as an abbreviated writing of the former.