5Dec 2020
00:00 UTC
#linguistweets
#abralin

Program

5Dec 2020 06:00 UTC
5Dec 2020 03:00 Local Time *

Acquisition of syntactic negation & NC

Sumrah Arshad (University of Goettingen - Germany)

This study focuses on investigating the mismatch in the use of negative elements in standard English between child and adults’ speech. It also shows that NC provides essential support for children to acquire the formal features of negation.
Research question:
This study focuses on investigating what determines how NC provides essential support for children in interpreting negation/negative elements (NMs) in the course of language acquisition. Background: Zeijlstra’s proposed framework predicts that languages may or may not require NegP. Double Negation languages in which a negative marker that is an adverb, serves as the negative operator and is interpreted in the semantics directly, cannot have a NegP. Languages where NMs do not directly correspond to a negative operator, are NC languages and may have NegP. For Zeijlstra, NC is a syntactic agreement and provides evidence for the presence of formal features, and multiple NMs stipulate that one of them carries [uNEG] that must be checked in the syntactic component, this is what motivates a NegP.
Proposal: I will use Zeijlstra’s (2004, 2014) framework to explain the acquisition of negation (AoN) and NC in SE. Children acquiring any language must settle the syntactic status of NMs of their target language. SE is a DN language in which every negative form corresponds to a negative meaning. Consequently, it predicts that children acquiring SE should start acquiring it as a DN language where the negative marker is predominantly an adverb. SE also projects formal negation through NegP, by a Nego n’t. We assume that children pass through a stage where they assume SE is a full NC language.
Data and results: This study provides empirical findings based on the analysis of Corpus data of SE retrieved from CHILDES database (McWhinney 2000). As predicted, based on the speech in (1) we argue that children start acquiring SE as a DN language using extensively the adverbial negation e.g. no and not. Going through stages, children place negation external or internal to the sentence. Besides, the negation in early child English does not mirror the adults’ most frequent form of negation, n’t. Findings also show the gradual and delayed acquisition of n’t, in spite of the fact that children receive substantial input for it. As predicted, it is argued that children do produce NC but not before the age of 3 years, exactly the age when a Nego is projected in their grammar. Examples are for instance: (1) a) Adam pocket no/not. [Adam: 27] (2) a) Helen don’t like her. [Helen:31] (3) a) But you don’t want no more. [MT: 50].
Conclusion: We conclude that children’s initial stages of acquisition of negation are instances of semantic negation. Given the absence of essential linguistic input (NC), their acquisition of n’t is rather delayed. Based on the conflicting input, children formalise negation and assume SE as an NC language and do produce NC. When the linguistic input violates children’s hypothesis that SE is an NC language, only then the production of NC decreases in child speech and they realise SE as their target DN language. Hereby, we conclude that SE is an inherently NC language.
5Dec 2020 06:15 UTC
5Dec 2020 03:15 Local Time *

Kazym Khanty -en: 2SG possessive—>proprial article

Stepan Mikhailov (HSE University - Russia)

This talk argues that the Kazym Khanty exponent -en stands for four distinct morphemes: 2SG possessive, salient article, anaphoric article, proprial article. The four differ w. r. t. competition with other possessives, agreement in number with the addressee and other features.
The 2SG possessive suffix of Kazym (dialect of Northern) Khanty -en [POSS.2SG] has functions beyond your run-of-the-mill possessives.
In this talk I discuss the properties of the various uses of the exponent -en and argue that it stands for four distinct morphemes: the possessive, the salient, the anaphoric, and the proprial articles. The data were collected through elicitation with speakers of Kazym Khanty in several fieldtrips.
Note first that in the possessive function POSS.2SG “agrees” with the Addressee in number: if the Addressee is plural, the 2NSG possessive -ən must be used instead. Also, it competes with the 1PL possessive -ew: if POSS.2SG is used on an NP, it is implied that the referent of the NP does not belong to both the Speaker and the Addressee (so the alternative utterance with POSS.1PL is false). Both these observations are expected for any possessive.
The picture gets complicated, when POSS.2SG is used in its extended functions.
These include first the salient article function, when it is obligatorily used with unique objects in commands. Here also, POSS.2NSG must be used if the Addressee is plural.
(1) soχλ-#(en)/-#(ən) mɵŋχ-a-λən
board-POSS.2SG/-POSS.2NSG wipe-IMP-NSG.NSG
{A teacher is telling a student/students.} ‘Wipe the blackboard!’.
Secondly, the marker is obligatorily used as an anaphoric or strong definite article with anaphorically accessible referents (2), and it is impossible with anaphorically inaccessible referents (be they unique or not).
(2) amp-#(en) ma peλ-am-a χurət-ti pit-əs
dog-POSS.2SG I at-POSS.1SG-DAT bark-NFIN.NPST become-PST[3SG]
‘{I was walking along the street when I saw a dog.} The dog started barking at me’.
Thirdly, it is obligatorily used as a proprial article with anthroponyms in argument positions.
(3) wɵntər-*(en) sewr-əs tʉt_jʉχ
A.-POSS.2SG cleave-PST[3SG] fire_wood
‘Andrej cleaved a log’.
However, in the salient and anaphoric uses the marker does not agree with the Addressee: only POSS.2SG is felicitous, even if the Addressee is plural. Moreover, in all of the extended uses — and unlike the proper possessive use — the marker does not compete with POSS.1PL, since the substitution of the former for the latter renders the examples infelicitous.
Based on these and other distinguishing properties I argue that -en stands for four distinct morphemes as listed above.
The possessive -en has the standard possessive semantics (Karvovskaya 2018), denoting a relation to the Addressee. The salient -en denotes a SALIENT_TO relation to the Addressee. Hence, agreement in number in both cases.
The anaphoric -en has the standard strong definite article semantics (Schwarz 2009). The proprial -en has the proprial article semantics (Muñoz 2019), restricted to anthroponyms. There’s no agreement in number with the Addressee for the latter two, since it does not figure in their semantics.
5Dec 2020 06:30 UTC
5Dec 2020 03:30 Local Time *

Cracking stereotypes the ling of discourse markers

Aicha Belkadi (SOAS University of London - UK)

The use of discourse markers in oral speech is often stereotyped in English. Yet DMs are anything but absurd words. My thread will describe some linguistic traits of DMs across languages, with a focus on Berber, and show how central they are to human grammar & communication.
In the English-speaking world, discourse markers are negatively considered and those overusing them regularly mocked, particularly on Twitter. The main aim of my tweet-talk is to show to a general audience that DMs are not absurd words but appear to be in fact central to the grammar of human language. My first tweets will briefly discuss studies focussing on different languages, including English, Italian and Korean. These have shown that DMs (i) strongly contribute to the coherent organisation of discourse, for e.g. by signalling a new turn of events, a shift in topic or topic continuity, (ii) are frequently used to convey a speaker’s stance towards the state of affairs described and/or (iii) may also function as turn taking cues (Schiffrin 1987, Traugott 1995, Brinton 1998, Rhee 2020). Furthermore, while much extensive research remains to be done, particularly on minoritised and lesser-described languages, there appear to be some solid cross-linguistic patterns in DM’s grammatical properties. DMs are thus optional and lack propositional content. They can be of any size -- particles, lexical units, phrases or entire clauses –, are prosodically and syntactically independent from other elements of an utterance and can have scope over single utterances or larger portion of discourse (Traugott 1995, Brinton 1998, Waltereit 2002, Heine 2013, Rhee 2020).
The main body of my tweets will focus on the sources that DMs are known to develop from. I will focus on a lesser known but apparently not rare source for DMs: deictic verbs of motion (Heine 2000, Heine & Kuteva 2002, Bourdin 2008). I will describe my work (Belkadi 2018, 2019a, b) on the discourse uses of the verb uɣal ‘to go back/ to return’ in Taqbaylit Berber, an Afro-Asiatic language indigenous to the Djurjura mountain range in Algeria. In narratives and monologues, speakers predominantly use uɣal in utterances which start new thematic portions of discourse or a new set of (related) events. These utterances relate to prior units of discourse in different ways; the chain of events the former starts may be unexpected, logical consequences, continuations or completely unrelated to the ones expressed by the latter (cf. Bourdin 2008). Rarely, uɣal is used to signal a return to the current thematic topic after an aside. uɣal has many of the formal characteristics of DMs, but it seems to have kept much of its verbal properties. In particular, it agrees with the subject of the utterance in which it occurs and sometimes agrees with previously mentioned discourse participants or topics, making it much more syntactically dependent and complex than canonical DMs.
The talk will conclude that DMs are central to human language, in particular for discourse coherence and information processing. DMs are not trivial; they are universal (Fraser 2006, Rhee 2013) and develop across languages from a range of sources, in complex ways.
5Dec 2020 06:45 UTC
5Dec 2020 03:45 Local Time *

Lexical classification of Tupí-Guaraní languages

Fabrício Gerardi (Universität Tübingen - Germany)

Co-author(s): Stanislav Reichert (Universität Tübingen - Germany), Tiago Tresoldi (Max Planck Institute)

With two methods employed in genetic analyses -- character and distance-based -- we classify Tupían languages based on lexical items from the Tupían Lexical Database (TuLeD). Results agree with previous classifications and support some linguistic and ethnographic conjectures.
Quantitative and computational methods have made it possible to process a great amount of linguistic data as well as to critically examine long-standing hypotheses in the field. They have often been applied to the genetic classification of languages, allowing not only for a clearer understanding of the evolution of the language families in question but also for deeper insights into migration patterns of the populations who speak their languages. Such classifications have traditionally been based on different methods, yet in the case of the Tupían family, quantitative methods have so far been rarely used. The primary strength of the quantitative methods lies in the comparison of their respective results which often yields valuable insights. Similar results provided by different methods could be an indication of reliability, even more so when these results agree with information from other disciplines. 
In our research we have opted for two distinct genetic methods used in molecular biology to analyze 250 items of 40 Tupí-Guaraní languages; one method is character-based, using cognate classes, and the other one is distance-based, using weighted lexical distances. The trees obtained from both analyses show a significant agreement among them. They also show agreement with recent genetic data for some of the Tupi-Guaraní population and with ethnographic literature reaching back to the sixteenth century. Furthermore, supported by the archaeological record and recent linguistic studies, our findings challenge and contradict a long-standing hypothesis, according to which Tupinambá is more closely related to the Guaraní (southern) group.
5Dec 2020 07:00 UTC
5Dec 2020 04:00 Local Time *

The way Spanish and Basque think about causality

Andrea Ariño-Bizarro (University of Zaragoza - Spain)

Co-author(s): Iraide Ibarretxe-Antuñano (University of Zaragoza - Spain), María Louro Mendiguren (University of Zaragoza - Spain)

This psycholinguistic study explores the role of intention in the way Spanish and Basque speakers think and talk about causal events. Results show that both speakers use similar linguistic structures to encode causality but conceptualise it significantly differently.
Intentionality is defined as the degree of participation of the agent in a causal event (Ibarretxe-Antuñano 2012). Previous work on Spanish and Basque caused-motion events has shown that intention is a crucial aspect in the categorisation and linguistic codification of causality (Ariño-Bizarro & Ibarretxe-Antuñano 2020). Accordingly, the key role of intentionality has been stablished crucial in the variety of encoding options available in these languages (Ibarretxe-Antuñano 2012) and in the way speakers categorise and recall these events (Filipović 2013, Ibarretxe-Antuñano et al. 2016).
Stemming from these studies, this talk addresses this research question: Do Spanish and Basque speakers give the same importance to intention when they conceptualise causal events? Data were collected using Causality Across Languages project’s videoclip stimuli (CAL, Suny-Buffalo), a set of 58 videos of causal actions. Thirty-two native European Spanish speakers (from Aragón) and twenty two native Basque speakers (from Ondarroa) participated in two tasks: (i) a non-verbal categorisation task where participants had to attribute different degrees of responsibility to the result performed in the clips; and (ii) a verbal description task, where participants answered the question “what happened?”.
Results show that Spanish and Basque speakers use similar linguistic structures to explicitly encode the semantic opposition between Intentionality vs. Unintentionality (e.g. Sp. se le cayó [CL.3SG.ACC CL.3SG.DAT fall.3SG.P] / Bq. erori zitzaion [fall AUX.3SG.DAT.3SG.ABS] vs Sp. Lo tiró [CL.ACC THROW.3SG.P] / Bq. bota zuen [throw AUX.3SG.ERG.3SG.ABS]). Nevertheless, the way they conceptualise causality is significantly different. Our data shows that, when informants assign the degree of responsibility in causal events, Basque speakers focus their attention on the agent itself, that is, who performs the action that causes the effect regardless of their intention; whereas Spanish speakers do not care about agent, but on agent’s intention.

References
Ariño-Bizarro,A. & I.Ibarretxe-Antuñano.2020. La accidentalidad de los eventos causales desde la perspectiva de la tipología semántica. López, F.(coord.) La involuntariedad en español. Frankfurt:Peter Lang.
Comrie,B.1985. Causative verb formation and other verb-deriving morphology. Shopen, T. (ed). Language typology and syntactic description.Vol.III. Cambridge:CUP: 309-347.
Filipović,L.2013. The role of language in legal contexts: a forensic cross-linguistic viewpoint. Law and Language: Current Legal Issues 15(19):328-343.
Ibarretxe-Antuñano,I.2012. Placement and removal events in Basque and Spanish. A.Kopecka & B.Narasimham (eds.) The events of ‘putting’ and ‘taking’. A cross-linguistic perspective. Amsterdam: John Benjamins,123-143.
Ibarretxe-Antuñano,I., T.Cadierno & A.Hijazo-Gascón.2016. The role of force dynamics and intentionality in the reconstruction of L2 verb meanings: A Danish-Spanish bidirectional study. Review of Cognitive Linguistics 14(1):136-160.
5Dec 2020 07:15 UTC
5Dec 2020 04:15 Local Time *

Social influence on negation in Early Modern Dutch

Levi Remijnse (Vrije Universiteit Amsterdam - The Netherlands)

This presentation offers a description of the different grammars that have arisen under the influence of both bottom-up and top-down change in negation in 17th century Dutch, from consistent single negation to consistent negative concord, depending on social class and region.
This presentation offers a description of the different grammars that have arisen under the influence of both bottom-up and top-down change in negation in 17th century Dutch. Prior to this century, Dutch was characterized by the expression of negative concord, i.e. negation expressed by the sum of the preverbal clitic en (NEG) and the adverb niet (not). Until 1638, en gradually started to erode in bottom-up fashion. Then, In 1638, elite writers introduced complete deletion of en in top-down fashion. My dataset contains letters written around 30 years after the top-down change, stemming from authors across regions and social classes. The letters reveal that many authors still insert en in their writings, while others progressively utilize single negation. A dominant set of the letters also displays optional deletion of en.

In the syntactic derivation, the negative clitic moves to a designated syntactic slot. Two different positions in the sentence capture this slot. The first position is in the left periphery, where the slot also creates room for extra constituents, such as double complementizers, e.g. als dat (if that). The second position is the right edge of the sentence, where, in verb clusters, en attracts main verb to the first position, e.g. dat ik niet en hadV1 willenV2 kunnenV3 lopenV4 (literally: that I not NEG had want could walk). Based on these phenomena, my dataset reveals four syntactic systems (one per writer):

1. A designated slot for the negative clitic at the left periphery;
2. A designated slot for the negative clitic at the right edge of the sentence;
3. Designated slots for the negative clitic at both the left periphery and the right edge;
4. No designated slots for the negative clitic.

System 1, 2 or 3 are found in the writings of the authors still using the negative clitic. Moreover, these systems are observed in a subset of the writers utilizing consistent single negation. Thus, en was still present in the process of derivation, and then artificially deleted at the surface. This might be an effect of collision between top-down pressure to drop en, as observed at surface structure, and bottom-up natural development of negation with en still being present in the derivational process. System 4 is the most progressive grammar: en is absent both at surface structure and in the derivation.

Finally, when examining the regions and social classes of the writers, we find that consistent use of the negative clitic is only found in the low class of Zeeland, which is expected to be the last condition affected by top-down change. Simultaneously, authors using consistent single negation both at surface structure and in derivation are from Noord-Holland, which is the region where this change is initiated.
5Dec 2020 07:30 UTC
5Dec 2020 04:30 Local Time *

How to clip words in English

Martin Hilpert (Université de Neuchâtel - Switzerland)

Co-author(s): David Correia Saavedra (Université de Neuchâtel - Switzerland), Rains Jennifer (Switzerland)

How do speakers of English make a short word out of a longer one? We put together a collection of 2400 English clippings and analyzed it with multivariate statistics. We detected patterns and regularities that reflect functional pressures that act on the speaker and the hearer.
In many languages, speakers use shortened word forms in spontaneous language use. Examples from English include lab (<< laboratory), exam (<< examination), or gator (<< alligator). This process is known as clipping. Existing work on clipping (Davy 2000, Durkin 2009, Haspelmath & Sims 2010, Don 2014) characterizes it as highly variable and unpredictable. Recently however, that view has been challenged (Lappe 2007, Jamet 2009, Berg 2011, Alber & Arndt-Lappe 2012, Arndt-Lappe 2018).

Our presentation will discuss findings that have been obtained on the basis of a newly-compiled large database of English clippings. A collection of 2400 English clippings has been annotated for phonological, morphological, syntactic, and semantic variables, along with corpus-based measurements of type and token frequency, as well as distribution.

Earlier research on clippings has emphasized the importance of factors such as the principle of least effort (Zipf 1949), the recoverability of the source word (Tournier 1985), and phonological factors such as stress and syllable structure (Lappe 2007). We will present a multivariate analysis of factors that reveal systematic patterns that guide the formation of clippings. On the basis of a hierarchical configural frequency analysis, we discuss what underlying factors are implicated in the clipping process and how these factors interact.

The overall conclusion is that clipping has been unjustly characterized as an unpredictable word formation process. Not only are clippings formed on the basis of systematic patterns, but these patterns also reflect functional pressures that act on the speaker and the hearer.
5Dec 2020 07:45 UTC
5Dec 2020 04:45 Local Time *

The QUD in quantity judgments

Kurt Erbach (University of Bonn - Germany)

Questions like “Who has more jewelry?” might seem straightforward to answer, by counting the pieces, but it has been shown that the answer changes with how and when the question is asked. I argue that the answers are based on what the Question Under Discussion is assumed to be.
I argue for a question under discussion (QUD) based analysis of quantity comparison tasks in English which, unlike previous analyses, allows for multiple dimensions of comparison as called for by Scontras et al. (2017), rather than one based solely on one dimension of comparison such as individuation or lack thereof (e.g. Barner and Snedeker 2005). Using a QUD based analysis that incorporates a formalized notion of context, this novel semantic model of quantity comparison tasks accounts for the multiple ways that speakers can judge quantities of different entities, unlike any previously proposed analysis of these linguistic situations.

In their groundbreaking study, Barner and Snedeker (2005) use quantity comparison tasks to show the effect of mass and count encoding on noun interpretation. Participants were shown two quantities of a certain entity and asked to judge who has more of the entity in question. Barner and Snedeker (2005) show that count nouns (e.g. chair) and object mass nouns (e.g. furniture) are judged in terms of cardinality of discrete individuals, while substance mass nouns (e.g. mustard) are judged in terms of volume rather than discrete portions. They argue that quantity comparison depended on whether or not a nouns was encoded as individuated, which is assumed to be true of object mass nouns and plural count nouns, but not substance mass nouns.

Subsequent to Barner and Snedeker's (2005) study, other have shown that speakers of English can use multiple dimensions for comparison of nouns in similar, yet modified, quantity comparison tasks: Grimm and Levin (2017) show that English speakers judge artifact nouns with respect to function of use, and they motivate a formal analysis that captures this type of judgment. Scontras et al. (2017) show that when the question is asked without using the target noun in question, participants give a variety of answers including cardinality, volume, function, value, and utility, though the authors provide no formal account of how these different dimensions might be called upon and used.

In my novel, QUD-baded analysis, I build on the precedent setting work of Rothstein (2010), who introduces a context variable into an analysis of the count--mass distinction that captures the variety of possible interpretations of what counts as one. I show how a context variable in the quantity comparison question "Who has more (N)?" where N is the noun in question, accounts for the variety of interpretations of this question seen in the studies by Grimm and Levin (2017) and Scontras et al. (2017). Not only does this analysis account for a complex linguistic phenomenon, but it also illustrates the inherently intertwined nature of semantics and pragmatics by joining the growing body of literature that shows that the larger discourse context is fundamental to determining semantic content (e.g. Rothstein 2010; Schwarzschild 1996).
5Dec 2020 08:00 UTC
5Dec 2020 05:00 Local Time *

Intensifiers across social media

Tatjana Scheffler (RUB - Germany)

Much linguistics, wow! Intensifiers like 'very', 'so' are typical for informal, spoken language, but also occur in social media. @tschfflr shows individual speaker differences in choice of intensifiers in German blogs and tweets by the same authors.
Intensifiers like 'very', 'so', etc. are typical for informal, spoken language, but they also are sometimes employed in writing. However, when an intensifier is used, it is up to the speaker to choose a more formal one ('very') or an informal one 'totally'. We study intensifiers in German blogs and tweets from the same authors, using a corpus of 62 authors and almost 2.5mio tokens. We show that intensifiers as a whole occur at similar frequencies overall in this data: In both blogs and tweets, intensifiers make up about 0.4% of tokens - about three times as many as in newspaper text.

Looking at the type of intensifiers reveals differences. I compare the formal intensifiers 'wirklich', 'sehr', 'absolut' and the informal ones ‘echt’, ‘krass’, ‘extrem’, ‘voll’, ‘völlig’, ‘total’, ‘ordentlich’, and ‘sau’. In blogs, formal intensifiers outnumber informal ones by a factor of 3:1, while formal and informal intensifiers are balanced in tweets. Since the same authors are present in both subcorpora, this allows us to further look at individual variation in the choice of intensifiers. The data shows that one group of authors adapt their intensifier choice to the medium (writing more informally in tweets), whereas another group of authors is approximately consistent in their usage of intensifiers across both media. This opens up interesting questions wrt. other aspects of their linguistic behavior.
5Dec 2020 08:15 UTC
5Dec 2020 05:15 Local Time *

Gaze-selection & syntax in multiperson interaction

Virginia Calabria (KU Leuven - Belgium)

I investigate how gaze relates to the grammatical features of turn-construction in collaborative turns. Gaze-selected speakers can provide syntactically more complex (re)completions, self-selected speakers simpler ones: do gaze-selected speakers display more entitlement to speak?
Holding the floor?
Gaze-selection & syntax in multiperson interaction

This contribution is an exploratory investigation of the relationship between speaker selection by gaze (Rossano, 2013; Auer, 2020) and the syntactical format of the recipient’s turn or that of another self-selected speaker. It contributes to recent research on “what embodied interaction can tell us about grammar” (Keevallik, 2018).
Kendon (1967) and more recently Stivers and Rossano (2010) have shown that a speaker’s gaze can elicit responses from a recipient in other sequential contexts than questions. I, therefore, investigate collaborative turns (cf. Lerner 1987, 1991, 2004; Bolden 2003; Auer 2015), namely cases of co-constructions and other-increments. The interplay between gaze-selection and syntax illustrates the relevance of gaze for turn-taking in multiperson interactions (Schegloff, 1995; Egbert, 1997), where there can be strong competition for taking turns.
The analysis is based on 10.5 hours of video data of interactions in present-day Italian involving 3 to 5 participants, recorded in ordinary (a dinner party) and institutional (two business meetings) settings.
Drawing on Conversation Analysis (Sacks, Schegloff & Jefferson, 1974) and Interactional Linguistics (Couper-Kuhlen & Selting, 2018), I show how a gaze-selected speaker can provide syntactically more complex candidate (re)completion (clauses), while a self-selected speaker can provide simpler (re)completions (phrases). This leads to the following research questions: 1) does this relationship between gaze selection and length/complexity of the talk correlate with the possibility of holding the floor? And is doing so more difficult when self-selected? 2) What actions does a non-addressed speaker accomplish by providing a (re)completion? 3) What patterns occur in the relationship between complex syntax and the local contingencies of social interaction and interactants’ embodied conduct?
The final aim of this research is to present a multimodal analysis of this phenomenon as a grammar-body-package (Kärkkäinnen and Thompson, 2018), that is the recurrent combinations of grammatical constructions with particular embodied conduct.
5Dec 2020 08:30 UTC
5Dec 2020 05:30 Local Time *

Variation in framing of real-world events

Gosse Minema (University of Groningen - Netherlands)

Co-author(s): Levi Remijnse (Vrije Universiteit - Amsterdam)

"Riots" or "demonstrations"? "Terrorism" or "rebellion"? The same event can be conceptualized and framed in many ways using language. We present analysis methods, #FrameNet lexicons and #NLP tools for 🇳🇱🇬🇧🇮🇹 to investigate how this works.
The aim of our project is to investigate how the same real-world situation can be “framed” in different ways using language. We work within the theoretical framework of frame semantics, which describes language meaning in terms of conceptual “frames”: abstract representations of events and situations that can be evoked in texts.

The frames that are used in a text can tell us a lot about the perspective of the text on the event. For example, in a text about a terror attack, they could tell us whether the perpetrators are described as terrorists, criminals or as freedom fighters.

We analyze variation in framing between texts in a data-driven way, integrating a dataset of real-world events with texts referencing these events. We also aim to use FrameNet and associated automatic analysis tools as a starting point for our analysis.

Standard FrameNet analysis is done in a bottom-up way: frames are derived from the lexical meanings of words in the text. However, we work in a top-down fashion, starting from the real-world events and frames that are potentially relevant for referencing the event. These frames are then annotated in the texts; to increase coverage of our analysis, we focus not just on lexical meaning, but also allow frames to be derived through pragmatic inference.

For example, consider the following structured data and text about the 2016 Berlin Christmas market attack:

1) Event: 2016 Berlin truck attack
2) Event type: murder
3) Data: Date=19 December 2016; Location=Breitscheidplatz, Berlin, Germany; Participant=Anis Amri
4) Frames: “Killing” (event in which a Killer or Cause causes the death of the Victim), “Use_firearm” (situation in which an Agent causes a Firearm to discharge)
5) Text 1: “On 19 December 2016, a truck was deliberately driven into the Christmas market next to the Kaiser Wilhelm Memorial Church at Breitscheidplatz in Berlin, leaving 12 people dead and 56 others injured.”
6) Text 2: “We raise funds for the victim of a cowardly act.”

Text 1 clearly describes a “Killing” event: there is an action (“a truck was deliberately driven”) which “leaves 12 people dead”, but there is no single word (e.g. “kill” or “murder”) corresponding to this event. Instead, different parts of the text together allow us to infer that the frame is active. Similarly, text 2 also references this “Killing” event, but there is not enough information in the text itself to infer this. “A cowardly act” could refer to various event types, but our background knowledge tells us that it refers to the truck attack, activating the “Killing” frame. “Inferred frames” like this add useful information to our analysis, but are hard to derive using existing bottom-up methods.

In practical terms, our project will yield a large corpus of structured event types and texts, annotated with frames, as well a pragmatic analysis of inferred frames, and computational tools for automatically analyzing frames in texts in different languages.
5Dec 2020 08:45 UTC
5Dec 2020 05:45 Local Time *

Because-X and because-ellipsis: A comparison

Laura Bailey (UK)

Co-author(s): Eleanor Cook (University of Kent - UK)

We argue that because-X (‘I can’t come because school’) is a (non-elliptical) fragment based on corpus, acceptability, and fill-in-the-blank data. This explains the acceptability of the varied categories of complement as well as the low acceptability of finite verbs and DPs.
Because-X and because-ellipsis: A comparison of two constructions

We argue that the because X construction provides further support for the analysis of some fragments as nonsententials (e.g. Progovac 2006), and that because X structures can be derived via the base-generation of such nonsentential elements or via ellipsis, with the two classes of construction exhibiting distinct characteristics.

The phenomenon is that in (3), where because has a complement other than the clause of traditional because or the DP of the because of complex:

(1) The conference is online because of [social distancing]. (because-of)
(2) The conference is online because [we can’t hold it in person]. (because-clause)
(3) The conference is online because [2020]. (because X)

The construction has been labelled ‘internetese’, although Rehn (2015) and Bergs (2018) claim earlier examples. We show that these earlier examples are a forebear of because X and this construction is still available alongside ‘true’ because X. Because X is not used solely in ‘internet speak’ but in informal written and spoken registers of English (and other languages) more generally.

We present data from grammaticality judgement surveys, fill-in-the-blank tasks and a twitter corpus demonstrating the range of complements that can occur in the because X construction and the implications of this. Because X cannot occur with a DP or TP and we therefore argue that Progovac’s analysis of nonsentential elements as Tenseless and Caseless should be extended to this construction.

This stands alongside ‘telegraphic’ utterances which may include Tense and Case elements and exhibit all the standard features of ellipsis, namely identity, recoverability and consistent meaning: 'Can’t talk right now because in the bath'. Such because ellipsis provides a causal or reason event for the main clause event. ‘True’ because X, on the other hand, gains its meaning from context, world knowledge or pragmatic inference and has unpredictable meaning, and provides a property, state or speaker attitude that explains the lack of a reason for the main clause event: 'Eating green pepper and onion pizza at 9:30 at night because yay' (Rehn 2015: 10).

This analysis allows us to explain the results of an experiment showing that because X lacks the ambiguity normally created by negating a because adverbial. We argue that the phrase in question is ‘too small’ to be negated: there is no structure above the phrase to host a negative element.

It also illustrates the close relationship between pragmatics and syntax in emerging constructions such as this one (Fortin 2007), where theoretically possible PP complements are ruled out not due to syntactic constraints but because they are pragmatically ill-formed (and we demonstrate that with enough context, such examples can be made grammatical).
Program