Extracting possessions from text: Experiments and error analysis

This paper presents a corpus and experiments to mine possession relations from text. Specifically, we target alienable and control possessions and assign temporal anchors indicating when a possession relation holds between the possessor and possessee. We work with intra-sentential possessor and possessees that satisfy lexical and syntactic constraints. We experiment with traditional classifiers and neural networks to automate the task. In addition, we analyze the factors that help to determine possession existence and possession type and common errors made by the best performing classifiers. Experimental results show that determining possession existence relies on the entire sentence, whereas determining possession type primarily relies on the verb, possessor and possessee.

Ambiguity avoidance in reflexive dependencies: Evidence from Romanian

The choice of a referring expression targeting a previously introduced discourse referent is affected by its potential ambiguity in a given context (Fukumura et al., 2013; Hwang, 2020): speakers use fewer pronouns in contexts where they would be ambiguous. In this work we investigate whether this effect extends to reflexive pronouns, whose distribution is typically governed by strict syntactic constraints, i.e. the Binding Theory (Chomsky, 1981; Büring, 2005). To ask this question, we turn to Romanian. Unlike English, the regular Romanian pronouns ea/el ‘her/him’ can corefer with a local referential antecedent (Luna talked about her), and be bound by local quantificational antecedents (Every girl talked about her). However, Romanian also has unambiguous reflexive expressions that may also be used in these contexts. We report two production experiments in Romanian investigating the effect of contextual ambiguity on the choice of referring expression for reflexive dependences with both referential (Experiment 1, e.g. Luna) as well as quantificational antecedents (Experiment 2, e.g. every girl), using a variant of the gender match paradigm used in previous work (Arnold, 2010). We find that, in unambiguous contexts, regular pronouns were the preferred form for reflexive and non reflexive dependencies in both experiments. However, whenever a regular pronoun would be formally ambiguous, speakers chose them less often, preferring instead unambiguous reflexive pronouns. Our results show: (1) like reference to non-local antecedents (Ariel, 1990, 2001; Arnold, 2010), intrasentential reference is also sensitive to discourse considerations, and (2) that potential discourse ambiguity impacts the choice of a referring expression irrespective of whether the dependency is achieved syntactically, i.e. bound variable dependencies, or via discourse computations, i.e. (local) coreference.

Context-Based Interpretation of Subordinating Conjunctions in Communication

While acknowledging that different conjunctions define different relationships between ideas, this study focuses on the interpretation of four subordinating conjunctions, namely, because, since, for and as in causal clauses. Since the meanings of these conjunctions vary according to context, among other things, they usually pose problems to language users leading to misinterpretations. Not only does this emerge from the fine and minute distinctions in usage between them, but also due to the lack of adequate knowledge of the rules of language use in their metatheoretical framework. This kind of knowledge is crucial in interactive communication, speech acts, pragmatics, logical arguments and multidisciplinary debates. The data for this study were compiled from grammar books, articles, and from the British National Corpus (BNC). The data were analyzed not only to identify the discrete features of each conjunction that would render it different from its synonymous counterparts, but also to understand the kind of knowledge required to determine the choice. The findings of the study reveal that, in addition to the syntactic constraints, the degree of the ‘givenness’ or ‘newness’ of the information that the conjunction introduces, context, degree of formality of the register, and lexical density of the utterance that contains the conjunctions emerged to play a role.

Spanish in contact with Korean: New insights into language switching

‘Code-switching’ (CS) refers to language-mixing where individuals who speak two or more languages switch from one to another, often mid-sentence. Several morpho-syntactic constraints governing when switches happen have been proposed in prior work, mostly on Spanish-English CS (e.g. Timm, 1975; Pfaff, 1979; Poplack, 1980). However, what happens when the languages are typologically different? This is the case with Spanish-Korean CS, which has not been systematically investigated. Korean and Spanish differ in many respects, including clause structure/word order, absence/presence of articles, and morphology (Korean: agglutinative, Spanish: fusional) (Kwon, 2012; Bosque, Demonte, Lázaro, Pavón & Española, 1999). For the present study, balanced Spanish-Korean bilinguals were interviewed to obtain a naturalistic corpus of CS. Strikingly, we find that many constraints proposed for Spanish-English CS do not hold for Spanish-Korean. Specifically, there are three main ways that Spanish-Korean CS violates the constraints proposed for Spanish-English: (i) in contexts involving word order/clause structure, (ii) on the level of nouns and (iii) on the level of morphemes. Crucially, the violations are not random: We suggest that they stem from the typological differences between Korean and Spanish. This work highlights the empirical and theoretical benefits of including typologically diverse language pairs when investigating CS.

SG-Net: Syntax-Guided Machine Reading Comprehension

For machine reading comprehension, the capacity of effectively modeling the linguistic knowledge from the detail-riddled and lengthy passages and getting ride of the noises is essential to improve its performance. Traditional attentive models attend to all words without explicit constraint, which results in inaccurate concentration on some dispensable words. In this work, we propose using syntax to guide the text modeling by incorporating explicit syntactic constraints into attention mechanism for better linguistically motivated word representations. In detail, for self-attention network (SAN) sponsored Transformer-based encoder, we introduce syntactic dependency of interest (SDOI) design into the SAN to form an SDOI-SAN with syntax-guided self-attention. Syntax-guided network (SG-Net) is then composed of this extra SDOI-SAN and the SAN from the original Transformer encoder through a dual contextual architecture for better linguistics inspired representation. To verify its effectiveness, the proposed SG-Net is applied to typical pre-trained language model BERT which is right based on a Transformer encoder. Extensive experiments on popular benchmarks including SQuAD 2.0 and RACE show that the proposed SG-Net design helps achieve substantial performance improvement over strong baselines.

Syntactic relations and syntactic constraints

This chapter continues the discussion of the formal architecture of LFG and of ways to describe and constrain constituent structures and functional structures. The chapter introduces additional relations and constraints on structures, and discusses concepts fundamental to our formal theory, including regular expressions (Section 6.1); sets (Section 6.3); off-path constraints (Section 6.6); templates (Section 6.7); relations between f-structures such as f-command, subsumption, and restriction (Section 6.9); c-structure/f-structure constraints including empty nod rules (Section 6.10); and precedence relations (Section 6.11). For most readers, this chapter best serves as a reference to be consulted for definition and discussion of concepts and relations that are used in the analyses presented in the remainder of the book.

Concession in Chinese proverbs

This article studies concession in Chinese proverbs from three different points of view: syntactic, semantic and pragmatic. It is based on several previous works that focus on concession and proverbs both in French and in Chinese. To our knowledge, no research has been done on the concession in proverbs. Proverbs possess several peculiar linguistic traits that condition the expression of concession. The objective of this paper is to examine the realization of this notion under proverbial syntactic constraints as well as to highlight their logico-semantic basis and to assess their pragmatic effects. Based on our corpus of Chinese proverbs, the article not only concludes with the most common syntactic structures, universal logical formula, but also points out the argumentative force and the politeness present in concessive Chinese proverbs.

Syntactic Features

Syntactic features are formal properties of syntactic objects which determine how they behave with respect to syntactic constraints and operations (such as selection, licensing, agreement, and movement). Syntactic features can be contrasted with properties which are purely phonological, morphological, or semantic, but many features are relevant both to syntax and morphology, or to syntax and semantics, or to all three components. The formal theory of syntactic features builds on the theory of phonological features, and normally takes morphosyntactic features (those expressed in morphology) to be the central case, with other, possibly more abstract features being modeled on the morphosyntactic ones. Many aspects of the formal nature of syntactic features are currently unresolved. Some traditions (such as HPSG) make use of rich feature structures as an analytic tool, while others (such as Minimalism) pursue simplicity in feature structures in the interest of descriptive restrictiveness. Nevertheless, features are essential to all explicit analyses.