Multiword Expressions Research Articles

The Development of Idiom Knowledge Across the Lifespan

Knowledge of multi-word expressions, such as break the ice, is an important aspect of language proficiency that so far we have known surprisingly little about. For example, it is largely unknown how much variability there is between speakers with respect to the number of different items that they know, or what factors contribute to their acquisition. This lack of knowledge seriously limits the generalizability of experimental studies on the production and comprehension of multi-word expressions (usually idioms) and generally suggests that there still is a sizable unknown territory of language knowledge to explore. Here, we present the results of two familiarity ratings for a large sample of Dutch idioms and a large number of participants that varied in age between 12 and 86 years old. The data show considerable variation between participants and between idioms. Nonlinear mixed-effects regression analyses revealed that the age of participants, but not their education, as well as the frequency and decomposability of the idioms influenced the familiarity scores. Our findings suggest that the knowledge of multiword expressions develops across the lifespan, is acquired from exposure, and - in participants younger than about 40 years of age - varies with item decomposability.

Extracting multiword expressions from texts with the aid of online resources

Abstract This article reports on a classroom intervention where L2 learners were prompted to look for multiword expressions in texts. The participants were two intact classes of Vietnamese learners of English as a foreign language. Over a period of eight weeks, the experimental group (n = 26) looked for expressions in texts, while the comparison group (n = 28) used the same texts for content-related activities. In pairs, students in the experimental group consulted online dictionaries and an online corpus to help them determine which word strings in the texts were common expressions. The students’ worksheets and audio-recorded interactions suggest they were by and large successful at this, but also reveal the students found it hard to identify the boundaries of expressions and occasionally failed to find the dictionary (sub-)entries that matched them. The two groups’ ability to recall the expressions was gauged by comparing their scores on a pre-test and a post-test administered one week after the last class and again five months later. The learning gains were greater in the experimental group, although the difference fell short of significance in the delayed post-test. Students in the experimental group whose proficiency in English was relatively high tended to benefit the most.

Structural and Semantic Classification of Verbal Multi-Word Expressions in Slovene

This paper is an upgraded version of a conference paper presenting the categorization of verbal multi-word expressions (VMWEs) according to the PARSEME COST Action Shared Task 1.1 Guidelines. The categorization is universal but takes into account the characteristics of the individual included languages. It was used to annotate 13,511 sentences of the Slovene ssj500k 2.0 training corpus, which resulted in 3,364 identified VMWEs categorized as inherently reflexive verbs, light verb constructions, inherently adpositional verbs, and verbal idioms. The paper presents both the quantitative and qualitative results of the analysis, compares the suggested categorization system to existing work on VMWEs in Slovene linguistics, and evaluates the use of the proposed system for future work.

Functional and usage-based approaches to aphasia: the grammatical-lexical distinction and the role of frequency

ABSTRACT Background Functional and usage-based theories of language are gaining increasing influence in linguistics. These theories understand language structure as underpinned by domain-general neurocognitive capacities and as shaped by usage patterns and the function of language as a means for communication. Accordingly, they entail an approach to aphasia which differs markedly from established ones based on formal theories. Aims Based on an outline of central claims in functional and usage-based theories, we aim to show how such theories can cast new light on aphasia. Methods & Procedures We focus on two strands of functional and usage-based aphasiological research: 1) research on frequency effects in aphasic speech, 2) and research on the grammatical-lexical distinction and its significance for the description of aphasic speech and the understanding of the causes of aphasia. We review available studies that fall within the two aforementioned strands of research, assessing their strengths and limitations. Outcomes & Results Usage-based methodologies are currently being developed that allow for fast quantification of the degree of formulaicity of a language sample and may thus be helpful in ascertaining the role of fossilized multiword expressions in aphasia. In line with central claims in usage-based linguistics, the first results of studies employing these methodologies have shown that frequency and collocation strength facilitate the retrieval of multiword expressions in a way that resembles the way in which lexical frequency facilitate retrieval of isolated words. A recent functional and usage-based theory understands the grammatical-lexicon distinction as a means for prioritizing parts of complex linguistic messages. Defining grammatical items as items that are discursively secondary (background) and dependent on host items, this theory entails an account of grammatical deficits which bridges the gap between existing structure-oriented and processing-oriented accounts. The theory entails word-class general criteria that allow fine-grained classification of linguistic items as grammatical or lexical. Cross-linguistic studies of verb, pronoun and preposition production show that this classification is significant for the description of aphasic language. Conclusions Functional and usage-based studies of aphasia are still sparse, but show promising results. This approach seems especially qualified for understanding 1) the neurocognitive causes of various types of aphasia, 2) the variability across languages, communicative settings (including tasks and modalities), groups of individuals and individuals, which is characteristic of aphasic speech, and 3) the link between aphasia symptoms and the basic need and challenge for people with aphasia: to remain a social being by communicating with other social beings.

Can word vectors help corpus linguists?

ABSTRACTTwo recent methods based on distributional semantic models (DSMs) have proved very successful in learning high-quality vector representations of words from large corpora: word2vec and GloVe. Once trained on a very large corpus, these algorithms produce distributed representations for words in the form of vectors. DSMs based on deep learning and neural networks have proved efficient in representing the meaning of individual words. In this paper, I assess to what extent state-of-the-art word-vector semantics can help corpus linguists annotate large datasets for semantic classes. Although word vectors suggest exciting opportunities for resolving semantic annotation issues, there is still room for improvement in terms of the representation of polysemy, homonymy, and multiword expressions.

The effect of frequency on learners’ ability to recall the forms of deliberately learned L2 multiword expressions

AbstractIn incidental learning, open class vocabulary items with high or relatively high objective frequency in input are comparatively likely to be acquired. However, many single words and most multiword expressions (MWEs) occur infrequently in authentic input. It has therefore been argued that learners of school age or older can benefit from episodes of instructed or self-managed deliberate (or intentional) L2 vocabulary learning, especially when L2 is learned in an EFL environment and most especially when productive knowledge is the goal. A relevant question is whether the objective frequency of vocabulary items is an important factor in production-oriented deliberate L2 vocabulary learning. We report three small-scale interim meta-analyses addressing this question with regard to two-word English Adj-Noun and Noun-Noun expressions. The data derive from 8 original studies involving 406 learners and 139 different MWEs. Our results suggest that objective frequency has a weak, possibly negative effect in the deliberate learning of MWE forms.

Design of a Multiword Expressions Database

Abstract The paper proposes design of a generic database for multiword expressions (MWE), based on the requirements for implementation of the lexicon of Czech MWEs. The lexicon is aimed at different goals concerning lexicography, teaching Czech as a foreign language, and theoretical issues of MWEs as entities standing between lexicon and grammar, as well as for NLP tasks such as tagging and parsing, identification and search of MWEs, or word sense and semantic disambiguation. The database is designed to account for flexibility in morphology and word order, syntactic and lexical variants and even creatively used fragments. Current state of implementation is presented together with some emerging issues, problems and solutions.

A usage-based alternative to “lexicalization” in sign language linguistics

The usage-based framework considers linguistic structure to be emergent from how human languages are used, and shaped by domain-general cognitive processes. This paper appeals to the cognitive processes of chunking, entrenchment, and routinization to explore a usage-based alternative to the structuralist notion of lexicalization, as it has traditionally been used in sign language linguistics. This exploration shows that chunking, entrenchment, and routinization are useful for re-contextualizing three “lexicalization” phenomena sign language linguistics: multiword expressions, fingerspelled words, and morphologically complex signs. An advantage of the usage-based approach for linguistic theory and description is that it anticipates the existence of linguistic constructions that exhibit analyzable internal structure and holistic properties simultaneously. This alternate framing alleviates the burden for sign language linguists to determine whether or not linguistic constructions have become “lexicalized”, and instead directs analysts to focus on the degree to which linguistic constructs are established in any language user’s mental representation of their language.

A functional analysis of text-oriented formulaic expressions in written academic discourse: Multiword sequences vs. single words

Text-oriented formulaic expressions (e.g., in addition to, on the other hand, nevertheless) are important in crafting reader-friendly prose and particularly frequent in academic discourse. By accommodating both multi-word expressions (MWEs) and single-word expressions (SWEs), the present study is among the first attempts to explore how the two types of expressions relate and contribute to the level of formulaicity in academic discourse. Through a manual examination of formulaic expressions in context, the present study provides important insights into the nature of formulaicity in the organisation of written academic discourse produced by mature and developing writers, with pedagogical implications for the training of novice writers in scientific fields.

A survey on contrastive opinion summarisation

Contrastive opinion summarisation (COS) is jointly generating summaries for two entities in order to highlight their differences based on the features. COS comprises of feature extraction, Sentiment prediction and summarisation. Recently, the research focus in COS has been in using semantics associated with words and multi-word expressions to shift from syntactic to semantic level. This survey paper covers different methods used for feature extraction, various similarity measures and different types of summarisation. In addition to these, various datasets and performance measures are also addressed. Finally, future research directions are also suggested.

Academic lexical bundles

Abstract An important component of fluent linguistic production and a key distinguishing feature of particular modes, registers and genres is the multi-word expressions referred to as ‘lexical bundles’. These are extended collocations which appear more frequently than expected by chance, helping to shape meanings and contributing to our sense of coherence and distinctiveness in a text. These strings have been studied extensively, particularly in academic writing in English, but little is known about how they may have changed over time. In this paper we explore changes in their use and frequency over the past 50 years, drawing on a corpus of 2.2 million words taken from top research journals in four disciplines. We find that bundles are not static and invariant markers of research writing but change in response to new conditions and contexts, with the most interesting changes within disciplines. The paper also discusses methodological approaches to studying bundles diachronically.

Vocabulary from 1735 Lithuanian translation of the Bibel in the dictionary of Pilypas Ruigys

This article examines the use of biblical vocabulary in Pilypas Ruigys’ dictionary „Littauiſch=Deutſches und Deutſch=Littauiſches Lexicon“ (Königsberg 1747). The biblical words in this dictionary were derived from the 1735 Lithuanian translation of the Bibel. This fact is clearly stated in the title of the dictionary. Moreover, in the foreword to the dictionary the author highlights that he attempted to pick out as many Lithuanian synonyms from the Holy Script as possible. The exact chapters and verses of the Bible referenced next to the Lithuanian words help to determine which words and multi-word expressions were included in the dictionary. To this day there hasn’t been a statistical and textual analysis, which would show the nature and scale of the Bible references in Ruigys’ dictionary. The analysis has shown that Ruigys refered to the Bible mostly in search of suitable translation of German words and multi-word expressions. In addition, the Bible was a source for expanding the list of lemmata of German-Lithuanian part of the dictionary. The biblical references are in most cases placed next to the single words or two-word (in rare cases three-word) expressions that refer to some kind of Biblical terminology, i.e. name of a person, a thing or an occurence. There are very few longer citations of the Bible. Approximately 3500 words and multi-word expressions were copied from the Old Testament and about 650 – from the New Testament. Book of Genesis was referenced most times (more than 350), more than 200 examples were taken from Book of Exodus, Psalms, Book of Isaiah and Book of Job. The most cited part of the New Testament was the Book of Matthews.

“HARTA” de noveles: un corpus de español académico

The aim of this review is to account for the process of compilation and codification of the corpus HARTA-Noveles. This corpus was created as part of the research project titled “Corpus-based study of lexical combinations of academic Spanish for the development of a computational tool for academic writing assistance” (HARTA) [1] , under the direction of Margarita Alonso Ramos (University of La Coruna). The corpus consists of representative samples of essays produced by Spanish university students and gathered with the purpose of studying academic lexical combinations (CLA) [2] , i.e., recurrent segments specific to the academic domain, along with collocations, discourse markers and other multiword expressions. Inspired by the BAWE corpus (British Academic Written English), HARTA-No-veles is formed exclusively by final project texts (for the degrees) and dissertations (for the masters) selected from different public repositories of Spanish universities and from various scientific domains. These texts have been annotated with an specific system adapted from that followed by the Royal Academy in CORPES [3] . [1] Acronimo del nombre en espanol del proyecto: “Estudio de las combinaciones lexicas del espanol academico basado en corpus para una Herramienta de Ayuda a la Redaccion de Textos Academicos” (HARTA). [2] En espanol, combinaciones lexicas academicas (CLA). [3] Corpus del Espanol del Siglo XXI (CORPES).

Quanteda: An R package for the quantitative analysis of textual data

quanteda is an R package providing a comprehensive workflow and toolkit for natural language processing tasks such as corpus management, tokenization, analysis, and visualization. It has extensive functions for applying dictionary analysis, exploring texts using keywords-in-context, computing document and feature similarities, and discovering multi-word expressions through collocation scoring. Based entirely on sparse operations,it provides highly efficient methods for compiling document-feature matrices and for manipulating these or using them in further quantitative analysis. Using C++ and multi-threading extensively, quanteda is also considerably faster and more efficient than other R and Python packages in processing large textual data. The package is designed for R users needing to apply natural language processing to texts,from documents to final analysis. Its capabilities match or exceed those provided in many end-user software applications, many of which are expensive and not open source. The package is therefore of great benefit to researchers, students, and other analysts with fewer financial resources. While using quanteda requires R programming knowledge, its API is designed to enable powerful, efficient analysis with a minimum of steps. By emphasizing consistent design, furthermore, quanteda lowers the barriers to learning and using NLP and quantitative text analysis even for proficient R programmers.

Text Recitation in Foreign Language Learning: Its Effectiveness on the Learning of Multi-word Expressions

This paper reports on a small-scale study that explored the extent to which two types of recitation tasks that engaged foreign language learners in memorization of conversational turns prepared in advance would differentially lead to the learning of multi-word expressions. Three groups of Japanese college learners of English participated in this research: a whole-text recitation group, a partial-text recitation group, and a contrast group. Results of the speaking tests administered suggest that whole- and partialtext recitation approaches are both conducive to the learning of language chunks and that there is considerable variance in facilitative effects between the two.

Discovering Continuous Multi-word Expressions in Czech

Multi-word expressions frequently cause incorrect annotations in corpora, since they often contain foreign words or syntactic anomalies. In case of foreign material, the annotation quality depends on whether the correct language of the sequence is detected. In case of inter-lingual homographs, this problem becomes difficult. In the previous work, we created a dataset of Czech continuous multi-word expressions (MWEs). The candidates were discovered automatically from Czech web corpus considering their orthographic variability. The candidates were classified and annotated manually. Afterwards, the dataset was extended automatically by generating all word forms of those MWEs that were annotated as nouns. In this work, we used the dataset as positive examples, we filtered out negative examples from the MWE candidates. We trained a classifier with mean accuracy 92.7%. We have shown that the combined approach slightly outperforms approaches concerning only association measures mainly on MWEs containing inter-lingual homographs and out-of-vocabulary words. The discovery methods can be applied to other languages which encounter orthographic variability in web corpora.

Enlarging multiword expression dataset by co-training

In multiword expressions (MWEs), multiple words unite to build a new unit in language. When MWE identification is accepted as a binary classification task, one of the most important factors in performance is to train the classifier with enough number of labelled samples. Since manual labelling is a time-consuming task, the performances of MWE recognition studies are limited with the size of the training sets. In this study, we propose the comparison-based and common-decision co-training approaches in order to enlarge the MWE dataset. In the experiments, the performances of the proposed approaches were compared to those of the standard co-training [1] and manual labelling where statistical and linguistic features are employed as two different views of the MWE dataset [2]. A number of tests with different settings were performed on a Turkish MWE dataset. Ten different classifiers were utilized in the experiments and the best performing classifier pair was observed to be the SMO-SMO pair. The experimental results showed that the common-decision co-training approach is an alternative to hand-labeling of large MWE datasets and both newly proposed approaches outperform the standard co-training [2] when the training set is to be enlarged in MWE classification.

Use of Stance Lexical Bundles by Turkish and Japanese EFL Learners and Native English Speakers in Academic Writing

Linguists have been interested in the sequences of words that tend to occur for a long time. The present study examines a particular type of recurrent chunks called lexical bundles (LB). LBs are multi-word expressions and an important component of the fluent linguistic production depends on the control of them (Hyland, 2008a). The purpose of this study was to find out the most common stance LBs used in argumentative essays written by native English speakers and Turkish and Japanese EFL learners. It also aimed at finding the structural and functional characteristics of these bundles, and to what extent these structures used by the Turkish and Japanese EFL learners are similar to the ones used by the native speakers. To answer these questions, Louvain Corpus of Native English Essays (LOCNESS), Japanese International Corpus of Learner English (JPICLE) and Turkish International Corpus of Learner English (TICLE) were used. The structures of the stance LBs were determined by following Biber’s (2006) classification and the functions were determined by adapting their classification. The concordancing program WordSmith was used to find and determine the 3-4 word stance LBs. In the statistical analysis, Type-token ratio and Log Likelihood were used. The results showed that native speakers use lexical bundles least; whereas, Japanese EFL learners use them most frequently. The functions and the structures of LBs vary in each group. Suggestions regarding how to teach these devices in foreign language education were also given.

Localising memory retrieval and syntactic composition: an fMRI study of naturalistic language comprehension

ABSTRACTThis study examines memory retrieval and syntactic composition using fMRI while participants listen to a book, The Little Prince. These two processes are quantified drawing on methods from computational linguistics. Memory retrieval is quantified via multi-word expressions that are likely to be stored as a unit, rather than built-up compositionally. Syntactic composition is quantified via bottom-up parsing that tracks tree-building work needed in composed syntactic phrases. Regression analyses localise these to spatially-distinct brain regions. Composition mainly correlates with bilateral activity in anterior temporal lobe and inferior frontal gyrus. Retrieval of stored expressions drives right-lateralised activation in the precuneus. Less cohesive expressions activate well-known nodes of the language network implicated in composition. These results help to detail the neuroanatomical bases of two widely-assumed cognitive operations in language processing.

Constructing a Second Language Sociolinguistic Repertoire: A Sociocultural Usage-based Perspective

Abstract This article discusses a sociocultural usage-based perspective on the development of sociolinguistic competence. Previous research has focused on learners’ acquisition and use of alternative ways of ‘saying the same thing’ (i.e. native-like variation) in relation to study abroad, contact with native speakers, and pedagogy. Missing from the literature are studies examining the developmental trajectories of individual learners from a qualitative perspective. This article takes a first step in this direction by documenting the specific lexicogrammatical constructions deployed by one learner of French, Leon, over the course of a 6-week stylistic variation intervention. The findings show that his sociolinguistic repertoire emerged from a single multiword expression, which in combination with Leon’s application of new metalinguistic knowledge and mediation from a teacher expanded to become a more productive schematic template. The research suggests that future work on L2 sociolinguistic development would do well to focus on qualitative accounts of individual developmental trajectories, emphasizing the specific lexicogrammatical constructions learners appropriate, to understand how L2 sociolinguistic repertoires are constructed across time.