Student Presentations
Metaphor in Machine Translation: A Post-Editing Task
Mayra Nas
Literary translation has traditionally been regarded as a creative process requiring a high level of skill, making it seem unsuitable for Machine Translation (MT). However, a small but steadily growing body of research has emerged in the field of Literary Machine Translation (LitMT), covering various genres and languages. Within literary translation, the concept of metaphors presents a significant challenge due to their complexity in both linguistic forms and cultural embeddedness. Building on these findings, we had professional literary translators post-edit machine-translated texts, aiming to explore when, how and why machine-translated metaphors are revised. These texts were translated using three different MT systems. Debriefing interviews followed to assess the translators' focus on the metaphors in the excerpts, without explicitly mentioning it. The next step is to annotate and analyze the post-edited outputs and interviews. Preliminary results indicate that the post-editors often found the MT output too literal, particularly for idiomatic expressions. In this presentation, I will discuss these findings in relation to how MT is currently used for the translation of literary metaphors and how it could be better utilized in the future.
Pitch in the Picture: The Role of Visual and Auditory Feedback in Second Language Acquisition of the Korean Lenis-Aspirated Stop Contrast
Ariëlle Reitsema
This study tested the effectiveness of immediate audio-visual versus auditory-only feedback in second language (L2) acquisition of intonation in Korean. In accentual-phrase initial position, Korean lenis and aspirated stops – which are part of a three-way laryngeal contrast – are mainly contrasted via post-plosive pitch (L and H, respectively). Natively Dutch-speaking learners of Korean typically find this contrast difficult in production due to negative transfer from their non-tonal L1, combined with the limited attention for prosody in Korean L2 education. In this study, 26 Dutch learners of Korean received a computer-assisted pronunciation training using Lorre, an extension to Praat that allowed for instant side-by-side comparisons of L1 example recordings and learners’ productions. Half of the group received an auditory-only training, while the other half received an audio-visual training with visualised f0 contours. Their productions of plosive initial target syllables were recorded a few days before as well as after the training, and were subsequently evaluated by native speaker judges in an online forced-choice identification task. The two trainings were not found to differ meaningfully in how much they helped students to improve their pronunciation, but overall the training did have a positive effect.
Exploring L1 influence in processing L2 English grammatical tense violations
Hajna Williams
This presentation explores second language (L2) processing with a focus on written language comprehension of grammatical errors derived from the use of the incorrect grammatical tense. Proficient Hungarian-English bilinguals will be compared to native English speakers to explore the extent to which first language interference results in differences in processing (Roberts & Liszka, 2013). Much production and offline research on L2 grammatical tense/aspect violations has been carried out, but exploring this through online methods has scarcely been done before (Chan, 2012; O’Reilly, 2018), especially with native speakers of a unique language such as Hungarian.
In English, both tense and aspect are grammaticalized. English is reported to have twelve grammatical tenses (including aspects, as English grammatical tense and aspect are closely related). In contrast, Hungarian has three grammatical tenses (past, present and future). While aspectual perspective is expressed grammatically in English, it is done so lexically in Hungarian (Dosi et al., 2017). ‘Grammatically’ refers to morpho-syntactic properties applied to the verb predicate. While Hungarian does make use of verbal prefixes/particles, they have their own separate meaning and are not simply morpho-syntactic changes to the verb stem. Thus, Hungarian relies primarily on lexical semantic cues to express these different additional temporal distinctions instead of grammatical cues, as English does.
My research questions are whether there is a difference in reading times at the critical regions (the predicative verb) when it is in the correct versus the incorrect tense and whether this difference significantly varies between the Hungarian-English bilingual group and the native English group.
This between-subject experiment includes participants from the student body of the University of Cambridge, UK, who are all young adults (early twenties). One group consists of Hungarian-English bilinguals, who acquired their English L2 in childhood and speak English at a highly proficient level, and the other group consists of native speakers of English. An in-person self-paced reading task is employed consisting of short English texts with the violations included within the text, followed by comprehension questions.
"I want to be the most neutral model": Language ideologies within German teacher education programmes at Dutch institutions
Katharina Eder
German is one of the languages with the highest amount of first- and second language speakers in the Netherlands. An investigation of these speakers' attitudes, experiences and beliefs allows deeper insights into the interaction of language and identity in a multi-cultural context.
This talk contains an analysis of three interviews conducted with L1-German lecturers who work in teacher education programmes, training students to become German teachers themselves. The speakers' unique interaction with German in a professional environment casts an additional layer on their linguistic identity.
The interviewees' ideologies relating to language emerged both on an explicit and an implicit level. While they all took a rather descriptive stance initially, various attitudes regarding linguistic phenomena, such as non-standard varieties, anglicisms and cross-linguistic influence became prominent quickly. The data points to a discrepancy between professional and private attitudes: The lecturers consider certain phenomena to be informative for their students but view them negatively on a more personal level. Interestingly, the speakers do not seem to be overly aware of this discrepancy.
Poking the frontiers of referentiality: Proper names in Hebrew construct state
Adam Rapoport-Peleg
Dependent (non-head) elements in compounds tend strongly to be non-referential (or generic) (Haspelmath & Sims 2010: 191; Bauer et al. 2013: 464). This feature can even be used to distinguish nominal compounds from similar constructions (Altakhaineh 2016). As this presentation will show, this is also relevant to the history of the construct state in Hebrew. In Semitic linguistics, the term construct state describes a possessive construction in which two nouns are juxtaposed, sometimes with modification to the head (Waltisberg 2012). It will be argued that the construct state started as a possessive construction in Biblical Hebrew but has been overwhelmingly reanalyzed as a nominal compound in Modern (colloquial) Hebrew, except for a clearly defined set of part-whole constructions (cf. Ravid & Shlesinger 1995 and Borer 2009). Consequently, dependent nouns cannot be as referential as they used to be in construct state. This explains why proper names, especially personal names, can now appear as dependent in construct state only under special conditions (cf. Danon 2018). But these special conditions are perhaps even more interesting than the construction on which they apply. It will be proposed that they are not merely an exception to the non-referentiality of dependent nouns in compounds (cf. Bauer et al. 2013: 464; Altakhaineh 2016). Instead, the possibility will be explored that referentiality involves a special kind of mental representation which is more than just reference to a unique entity.
Reconsidering the Persian Classifier tā
Kowsar Amiri
Persian is hardly on anyone's radar when discussing "classifier languages". This is unsurprising; after all sortal classifiers are scarce and have mostly been treated as completely optional in existing literature. Using data from a corpus of contemporary spoken Persian, I will show that the general classifier tā is obligatory in numeral phrases. I also address the apparent exceptions, such as register differences and nouns that double as classifiers.
An interesting feature of tā is its ungrammaticality with the number yek 'one' and complex numbers that end in -yek. While ungrammaticality with 'one' has been addressed in previous literature, incompatibility with complex numbers remains unexplained. This raises the question: How is a classifier sensitive to the inner makeup of complex numerals? Additionally, what makes 'one' special?
Reanalyzing given notions: how a marker of givenness has always been taken for granted
Giosuè Balocco
Quechuan literature has been consistently describing an enclitic found in many Quechuan languages as a topic marker. Usually though, this enclitic (=qa, =ka or =ga in different variants) does not necessarily attach to topical expressions, raising a question: when and why is it used? Recent studies (Grzech 2016, Bendezú-Araujo 2023) have started to question the given notion of Quechua evidentiality proposing new, more fitting analyses, though mostly maintaining the description of =ka as a topic marker. Grzech (2016:274) suggests that the usage of Tena Kichwa =ga is connected to topicality while also noting that it “… could be analysed as a marker of presupposed content which the speaker wishes to make salient to the addressee. However, further research is needed to spell this analysis out in more detail.”
Following this path, in my research, I use interactional and elicited data from the Imbabura Kichwa dialect of Chirihuasi, Ecuador, to support a reanalysis of =ka as a marker referencing content already belonging to the Common Ground, exploring contexts in which it would be hard to analyse it as a topic marker.
A demonstration of incoherence: motivating gradient incoherence discourse structures in literary fiction
Natalie Nichols
What makes a discourse incoherent? How do interlocutors seek to reduce or resolve incoherencies?
Though formal pragmatics has spent recent years examining what makes a discourse coherent, incoherence is often left merely as a negative definition: the absence of ‘coherence’. However, I seek to motivate a more robust definition of incoherence, one with qualities in its own right, through applying segmented discourse representation theory (SDRT) to stream-of-consciousness literary fiction like Virginia Woolf’s The Waves. Fiction offers a unique and exciting opportunity for discourse analysts: incoherence is employed intentionally and frequently to illustrate different ideas, whether aesthetic or philosophical. We can look to these examples to motivate and exemplify a robust definition of gradient incoherence, as well as to distinguish SDRT incoherence definitionally from other kinds of infelicitous discourse and techniques.
In this case study, I build off the coherence relations work of Kehler (2002) and more recently Altshuler (2022; 2024) to suggest that incoherence is created by unlabeled or disconnected complex discourse units. Additionally—and more interestingly—I illustrate that coherence is not static, but rather dynamically gradient. Some coherence relations, specifically parallel, contrast, and narration, are inherently gradient concepts; the rest have yet to be proven gradient. This explains the intuitions behind why some discourses feel more gradiently acceptable than current coherence studies currently allow for: presently, something is either coherent or incoherent. I offer evidence of what types of graphs and topical constraints are required for gradient acceptability within these relations. I also distinguish incoherence and infelicitousness: why infelicitous statements can still be coherent to interlocutors and why incoherence can create such high levels of narrative frustration. Our ability to construct and follow narrative progression lies at the heart of our ability to understand one another; our capacity to parse out incoherency, whether intentional or unintentional, affects our cooperation as discourse participants.
Waves Across the Atlantic: the Genetic Position of Faroese
Simon den Hertog
Faroese, spoken by approximately 70,000 people in the Faroe Islands, Denmark and beyond, is traditionally classified as a West Nordic language alongside Icelandic and Norwegian. Within this group, it is often paired with Icelandic under the “Insular Norse” subclade or left unclassified. Over the past few decades, however, other scholars have proposed closer ties between Faroese and (South)west Norwegian, making the classification of Faroese a contentious yet underexplored issue.
This presentation reexamines the genetic position of Faroese through both existing evidence and newly identified linguistic innovations. The findings reveal a complex network of shared innovations, with 5–6 aligning Faroese with Icelandic and as many as 24 aligning Faroese with Norwegian, particularly the southwestern Voss dialect. The results, further enriched by historical records and new genomic data, suggest that the Faroese maintained stronger social ties with Southwestern Norwegians than with Icelanders, although ties with the latter are also evidenced by several shared innovations.
Given this conclusion, I argue for abandoning the “Insular Nordic” label.However, I do not advocate for the addition of Faroese to the Norwegian or Continental Nordic clades; Faroese co-evolved with both Icelandic and Southwest Norwegian to varying degrees. Instead, I propose that the adoption of a non-cladistic, wave-based framework, such as Historical Glottometry, is required to accurately capture the complex linguistic relationships of the West Nordic languages.
Is Sino-Tibetan Related to Ket and Na-Dené? How Should We Evaluate Macro-Families?
Josiah Medin
Since the Yeniseian language family has been known to Western linguists, there have been numerous attempts to unite it with other top-level language families. Of these proposals, one of the oldest and best known is the proposed relationship with Na-Dené of Western North America, which, according to some, is one of the most plausible current linguistic macro-family proposals (cf. Kari and Potter 2010). There is, however, another proposal of comparable age that unites the Yeniseian languages, of which the only extant spoken variety is Ket, with the expansive Sino-Tibetan (Trans-Himalayan) family. This was first proposed 1856 by Anton Schiefner (cf. Vajda 2001: 3), but has received renewed attention in recent years due to the advancement of Sinitic and Tibeto-Burman historical reconstruction and the popularity of the Dené-Yeniseian hypothesis.
For example, sinologist Jingyi Gao (2017) has allegedly identified a number of different shared lexical items between Sinitic and Yeniseian which appear to belong to the most basic strata of vocabulary. These include, for example, Proto-Yeniseian ‘eat’ and Old Chinese 食, which Gao reconstructs as *dʑiək (2017: 56), and Proto-Yeniseian *t’ɨˀs/čɨʔs with Old Chinese 石 *ʑiak/*djaɡ” (Gao 2017: 52). Furthermore, David Bradley (2023: 176) notes the existence of an even larger number of lookalike forms between Na-Dené, Sino-Tibetan, and Yeniseian, and argues that, since it is highly unlikely that contact occurred between the languages in recent times, the similarities may be residual. In particular, he notes similarities in morphology between the three, such as an *s- valency increasing morpheme, a negative or prohibitive marker with an alveolar plosive, and some lexical parallels such as body parts, which may exhibit regular sound correspondences (Bradley 2023: 180-182).
It is all the more striking, then, that, despite the generally good academic reputation of some of its proponents, such as David Bradley, the proposal linking Sino-Tibetan to Na-Dené and Yeniseian continues to be largely ignored by many in the field. Thus, in order to fill this gap, the newer linguistic evidence for a connection between Sino-Tibetan, Yeniseian, and Na-Dené will be evaluated in order to determine whether or not the notion that these families are genetically related looks promising. Then, recommendations for researchers looking to investigate the proposed relationship even further will be given, along with a more general discussion of how to evaluate such macro-family proposals epistemologically.
The Curious Status of Simple Thematic Present Verbs in Anatolian
Caroline M.J. Cambré
The Anatolian language branch was the first to split off from the Proto-Indo-European mother tongue, leaving the Core Indo-European languages behind as a separate proto-language. Simple thematic present verbs form one of the most basic verbal classes within the Core Indo-European languages. These verbs seem to be strikingly scarce within the Anatolian languages. Only some fifteen examples have previously been identified. Complex thematic present verbs, on the other hand, are wide-spread in both the Core Indo-European and the Anatolian languages. This presentation will determine the exact status of the Anatolian simple thematic present verbs, which might shed light on the possible causes of their scarcity.
Parallelism in lexical pre-activation and prediction
Olivia Nutt
Despite significant recent attention, the cognitive mechanisms underpinning anticipatory effects in linguistic processing remain controversial. There is little agreement on whether these phenomena represent inhibition or facilitation relative to a baseline, or a gradient inhibition without much of a meaningful baseline. One major confound in this has been the organisation of the analysis: Surprisal theory (Levy, 2008; Smith & Levy, 2013) literature, with its focus on log-transformed probability, has led to significant discussion about the impact of measures of predictability and research conclusions. However, this discussion remains incomplete, as context constraint has received very little attention, and the log-linear debate is often poorly related to issues with the concrete implementation of parallel processing. This study is a re-analysis of human cloze and eyetracking data in the Provo Corpus (Luke & Christianson, 2016, 2018) to provide a more transparent comparison of analysis schemas in predictability and constraint. For the former, this includes a log-linear comparison, as well as a new factor, ordinal rank, aiming to address how many competitor continuations are available in parallel. For the latter, Shannon entropy is contrasted with modal continuation probability, as these are reported to be correlated but have trivially different implications for the status of parallel competitors.
Sociophonetics of the /r/ consonant in Romanian: When does a speech defect become a sound change?
Horea-Ștefan Marinescu
Many speakers of Romanian often have a dorsal realisation for their /r/ consonant. This phenomenon is understudied and often disregarded as speakers merely being unable to pronounce the consonant "correctly". However, after doing some on-the-ground research on the topic for my BA thesis, I've found not only that allophonic variation of /r/ might be more diverse than most realise, but also that this non-standard way of pronouncing what is typically an alveolar tap are not generally seen as wrong by a college-aged demographic.
Language of the Lower Classes in Scotland: Variation in Relativization in Nineteenth-Century Scottish Pauper Petitions
Dominique van der Ven
Studies on language contact between English and Scots in Scotland and the subsequent anglicisation of the native Scottish language have mainly focused on the upper ranks of society (Corbett, 2013; Dossena, 2002, 2005; Meurman-Solin, 1997). Thanks to the availability of new data from the lower classes of Scotland with the emergence of the Scottish Pauper Petition corpus (ScotPP, 2022), it is possible to now shed light on a previously “unknown” part of society. In my research, I analysed relativization patterns, an important feature relevant for documenting anglicisation in Scotland (Devitt, 1982; Herrmann, 2003; Hillberg, 2015), in nineteenth-century Scots texts from all classes with the help of corpus analysis. Relativization patterns concerned the choice of "that" versus wh-relatives and various factors which may contribute to the choice of relative marker (i.e., restrictiveness, animacy, function) and which may vary between (Standard) English and Scots. My research showed substantial differences between the two classes, with lower-class data showing greater numbers of "that" and other typical spoken Scottish language variants compared to the higher-class data, which was in turn more similar to Standard English written conventions. This research provides a fuller picture of Scottish anglicisation across social classes and most importantly, adds to the growing literature in historical (socio)linguistics concerned with studying language history from below.
I love my body, but: An analysis of semantic fields and the use of the word body in #bodypositivity posts on Instagram
Eva-Magdalena Angerer
With a drastic increase in social media use among young people over the recent decade various social media trends concerning various themes can be identified. One of the movements especially popular with female users on Instagram is the so-called body positivity movement which advocates for a healthy body image and acceptance of all body types online. My research projects investigates this movement and its language use by analyzing 15 posts using #bodypositivity on Instagram. In a first step, the most prominent semantic fields in body positivity posts were identified. This part of the study found that the word ‘body’ itself was the most prominent word in the sample, which is why the second part of the study investigated the lexical environment the word is used in. The results of this project suggest that body positivity is considered a life-long process and that it actively promotes self-love and individuality. Moreover, it tries to create an in-group group identity between influencers and their followers by recounting influencers’ experiences that are shared with their followers.
Isn't It Ironic? Or Sarcastic? Or Both? A Study on the Nature of Irony and Sarcasm in Communication
Zoi Maria Matsouka
Irony and sarcasm have been widely discussed in pragmatics, sociolinguistics, and psycholinguistics, yet their precise nature remains unclear. This study investigates the core elements that define irony and its complex relationship with sarcasm, based on established English-based theoretical frameworks, including (im)politeness, neo-Gricean, echoic mention, and pretence theories, while also considering extralinguistic cues such as intonation, gestures, and emojis. In addition, sarcasm is examined as a subcategory of irony and as a separate verbal technic with particular focus on its role in expressing social critique, mock aggression, or playful teasing.
Building on this theoretical foundation, a questionnaire is designed, targeting Greek native speakers, to explore whether the minimal elements defining irony and sarcasm in English are also relevant in another language and how speakers perceive and differentiate irony from sarcasm, based on these elements. Additionally, the study aims to clarify the distinction between first-order (natural language use) and second-order (theoretical and linguistic) understandings of these terms, contributing to a more comprehensive cross-linguistic perspective on irony and sarcasm.