Showing papers on "Head (linguistics) published in 2019"

Rethinking Classification and Localization for Object Detection

Abstract: Two head structures (i.e. fully connected head and convolution head) have been widely used in R-CNN based detectors for classification and localization tasks. However, there is a lack of understanding of how does these two head structures work for these two tasks. To address this issue, we perform a thorough analysis and find an interesting fact that the two head structures have opposite preferences towards the two tasks. Specifically, the fully connected head (fc-head) is more suitable for the classification task, while the convolution head (conv-head) is more suitable for the localization task. Furthermore, we examine the output feature maps of both heads and find that fc-head has more spatial sensitivity than conv-head. Thus, fc-head has more capability to distinguish a complete object from part of an object, but is not robust to regress the whole object. Based upon these findings, we propose a Double-Head method, which has a fully connected head focusing on classification and a convolution head for bounding box regression. Without bells and whistles, our method gains +3.5 and +2.8 AP on MS COCO dataset from Feature Pyramid Network (FPN) baselines with ResNet-50 and ResNet-101 backbones, respectively.

Head-Driven Phrase Structure Grammar Parsing on Penn Treebank

Abstract: Head-driven phrase structure grammar (HPSG) enjoys a uniform formalism representing rich contextual syntactic and even semantic meanings. This paper makes the first attempt to formulate a simplified HPSG by integrating constituent and dependency formal representations into head-driven phrase structure. Then two parsing algorithms are respectively proposed for two converted tree representations, division span and joint span. As HPSG encodes both constituent and dependency structure information, the proposed HPSG parsers may be regarded as a sort of joint decoder for both types of structures and thus are evaluated in terms of extracted or converted constituent and dependency parsing trees. Our parser achieves new state-of-the-art performance for both parsing tasks on Penn Treebank (PTB) and Chinese Penn Treebank, verifying the effectiveness of joint learning constituent and dependency structures. In details, we report 95.84 F1 of constituent parsing and 97.00% UAS of dependency parsing on PTB.

Head-Driven Phrase Structure Grammar Parsing on Penn Treebank

Abstract: Head-driven phrase structure grammar (HPSG) enjoys a uniform formalism representing rich contextual syntactic and even semantic meanings. This paper makes the first attempt to formulate a simplified HPSG by integrating constituent and dependency formal representations into head-driven phrase structure. Then two parsing algorithms are respectively proposed for two converted tree representations, division span and joint span. As HPSG encodes both constituent and dependency structure information, the proposed HPSG parsers may be regarded as a sort of joint decoder for both types of structures and thus are evaluated in terms of extracted or converted constituent and dependency parsing trees. Our parser achieves new state-of-the-art performance for both parsing tasks on Penn Treebank (PTB) and Chinese Penn Treebank, verifying the effectiveness of joint learning constituent and dependency structures. In details, we report 96.33 F1 of constituent parsing and 97.20\% UAS of dependency parsing on PTB.

Phrase-level Prosodic Smothering in Makonde

Abstract: This paper focuses on the issue of 'prosodic idiosyncrasies' as it arises in the Bantu language Makonde [kde]. Recently, Bennett, Harizanov, & Henderson (2018) proposed 'prosodic smothering', whereby prosodic requirements of an outer morpheme override (i.e. 'smother') prosodic properties of inner morphemes. We extend their analysis to phrase-level phonology in Makonde. Previous description has established that whether a nominal modifier forms a single phonological phrase φ with the noun is an idiosyncratic property, e.g. a [noun adjective] phrase maps to 2 phonological phrases φ (n) φ (adj) while a [noun demonstrative] phrase forms a single phonological phrase φ (n dem). Prosodic smothering is seen in [noun adj dem] sequences which form a single φ (n adj dem) phonological phrase, where the adj has been 'entrapped' and its prosody 'smothered'. We highlight three contributions which Makonde makes to understanding smothering: (i) smothering targets the lexical head, (ii) smothering is both inward-oriented (a morphological relation) and leftward-oriented (a linear relation), and (iii) a limited amount of outward smothering is parasitic on the presence of inward smothering. From the smothering facts in Makonde, we conclude that prosody is established at two stages: first, prosodic idiosyncrasies apply at spell-out (i.e. the mapping from syntax to phonology), followed by default prosodification which is established within the phonological module itself.

Inertial Sensors Reveal Subtle Motor Deficits When Walking With Horizontal Head Turns After Concussion.

Abstract: Objective To examine whether horizontal head turns while seated or while walking, when instrumented with inertial sensors, were sensitive to the acute effects of concussion and whether horizontal head turns had utility for concussion management Setting Applied field setting, athletic training room Participants Twenty-four collegiate athletes with sports-related concussion and 25 healthy control athletes Design Case-control; longitudinal Main measures Peak head angular velocity and peak head angle (range of motion) when performing head turns toward an auditory cue while seated or walking Gait speed when walking with and without head turns Results Athletes with acute sports-related concussion turned their head slower than healthy control subjects initially (group β = -4947; SE = 1633; P = 003) and gradually recovered to healthy control levels within 10 days postconcussion (group × time β = 480; SE = 141; P Conclusion Inertial sensors can improve traditional clinical assessments by quantifying subtle, nonobservable deficits in people following sports-related concussion

Against the PredP Theory of Small Clauses

Abstract: Since Bowers 1993, it has been accepted that nonverbal small clauses are headed by a functional head, Pred0, whose function is to obligatorily mediate all nonverbal predication. I argue against thi...

Indexing and flagging, and head and dependent marking

Abstract: This paper compares the concept pair indexing/flagging with the well-known concept pair head/dependent marking that is widely used in typology. It shows that a general concept of flagging (comprising case and adpositional marking) is needed, and it sketches the advantages of the indexing concept over the older idea of “person agreement”. It then points out that the notions of head and dependent are hard to define (apart from the two basic domains of clauses and nominals), and that the head/dependent marking typology does not take the function of syntactic relation markers into account. On a functional view, both flags and indexes can be seen as roleidentifiers, as opposed to concordants (attributive agreement markers). After discussing three further issues with the head/dependent marking typology, involving construct markers, concordants, and cross-indexes, I conclude that the concept pair indexing/flagging is more suitable for typological purposes than head/dependent

English Focus Perception by Mandarin Listeners

Abstract: This study compared how well native Mandarin and native English speakers can perceive prosodically marked focus in English echo questions. Twenty-five yes–no echo questions were produced with a sentence focus, a verb focus, and an object focus. After hearing each sentence, they were asked to choose a correct response. Native English listeners were more accurate than native Mandarin on verb and object focus, but not on sentence focus. More importantly, both groups confused object focus with sentence focus and vice versa. However, confusion between object and verb focus, and between object and sentence focus was infrequent. These results suggest that, in some cases, (1) acoustic prominence on the head of a phrase or its internal argument can project to the entire phrase and make the entire phrase focused, and (2) parallel transmission of the two functions of intonation, and cross-linguistic variation in focus marking (prosodically versus syntactically) may contribute to their perceptual ambiguity.

Double-Head RCNN: Rethinking Classification and Localization for Object Detection

Abstract: Two head structures (i.e. fully connected head and convolution head) have been widely used in R-CNN based detectors for classification and localization tasks. However, there is a lack of understanding of how does these two head structures work for these two tasks. To address this issue, we perform a thorough analysis and find an interesting fact that the two head structures have opposite preferences towards the two tasks. Specifically, the fully connected head (fc-head) is more suitable for the classification task, while the convolution head (conv-head) is more suitable for the localization task. Furthermore, we examine the output feature maps of both heads and find that fc-head has more spatial sensitivity than conv-head. Thus, fc-head has more capability to distinguish a complete object from part of an object, but is not robust to regress the whole object. Based upon these findings, we propose a Double-Head method, which has a fully connected head focusing on classification and a convolution head for bounding box regression. Without bells and whistles, our method gains +3.5 and +2.8 AP on MS COCO dataset from Feature Pyramid Network (FPN) baselines with ResNet-50 and ResNet-101 backbones, respectively.

Learning to Predict Novel Noun-Noun Compounds.

Abstract: We introduce temporally and contextually-aware models for the novel task of predicting unseen but plausible concepts, as conveyed by noun-noun compounds in a time-stamped corpus. We train compositional models on observed compounds, more specifically the composed distributed representations of their constituents across a time-stamped corpus, while giving it corrupted instances (where head or modifier are replaced by a random constituent) as negative evidence. The model captures generalisations over this data and learns what combinations give rise to plausible compounds and which ones do not. After training, we query the model for the plausibility of automatically generated novel combinations and verify whether the classifications are accurate. For our best model, we find that in around 85% of the cases, the novel compounds generated are attested in previously unseen data. An additional estimated 5% are plausible despite not being attested in the recent corpus, based on judgments from independent human raters.

Identifying N+N compounding in Modern Standard Arabic and Jordanian Arabic

Abstract: This study aims to identify types of compounds in Modern Standard Arabic (MSA) and Jordanian Arabic (JA) by applying the cross‐linguistic criteria for compoundhood discussed in the relevant literature, with a special focus on English and Hebrew. These criteria – orthographic, phonological, syntactic and semantic in nature – have been proposed to make a distinction between compounds and phrases cross‐linguistically. The analysis reveals that the most reliable cross‐linguistic criteria to distinguish between compounds and phrases in MSA and JA are adjacency and referentiality. With regard to the former criterion, no intervening elements can be inserted between the head and the non‐head of compounds, whilst such insertion is allowed in phrases. With regard to the latter criterion, the non‐head of a compound is normally non‐referential, whereas the non‐head of a phrase is always referential. Other criteria have been found to be partially applicable, e.g. compositionality, possibilities for modification and coordination, and free pluralisation of the non‐head. In this study, I also suggest two reliable criteria that are exclusive to Arabic, or potentially Semitic languages in general. The first criterion is the appearance/absence of the possessive marker li‐/la ‘for/of’ when the first element is definite. The second criterion deals with the appearance/absence of the possessive marker li‐/la ‘for/of’ when the first element is preceded by a cardinal number. The study concludes with recommendations for further research.

A cross-linguistic perspective on the Right-Hand Head Rule: the rule and the exceptions

Abstract: Abstract The study investigates the patterning of headedness in compounding, in particular the patterning of regularities and exceptions to the Right-Hand Head Rule (RHHR). An examination of the grammatical descriptions of compounds in English, French, Italian, German, Arabic, Hebrew, Dutch and Spanish indicates that languages such as German and Dutch are strongly right-headed, other languages such as French and Arabic are strongly left-headed, whereas English, Spanish and Italian tend to be mixed between left-headed and right-headed. Despite the existence of some exceptions to RHHR in some of these languages, the rule remains viable, as these exceptions may have a systematic pattern. While in Romance languages the exceptions seem to be phonologically conditioned, in Germanic languages the exceptions appear to be syntactically conditioned. This study raises the question whether internal headedness in a language could be regarded as a fairly arbitrary property, unconnected to the language’s other characteristics, or not.

Blusukan Cultural as a Political Power in The Regional Head Election of Surakarta

Abstract: Blusukan has existed since the days of the Javanese kings. In the context of political communication, the blusukan carried out by political candidates aims to influence the voice of potential voters or to win the campaign. Attire, ways of speaking, language, posture and other attributes used by political candidates determine people's perception of the candidate's image. In order to attract public attention, political candidates use local cultural identities and attributes, namely Javanese culture to various places. Although we have entered the era of digital technology, but the strength of blusukan can still be felt today through the interpersonal communication between political candidates and the society. This research uses literature study method and documentation study from credible and reliable data. The data collected consists of two types, for instance primary and secondary data.Qualitative analysis is carried out in desk study, namely information retrieval and data collection through literature studies. The results showed: the society felt proud to be able to meet directly with the political candidates, blusukan had been able to influence the behavior of potential voters, and blusukan was very effective during the regional head election campaign.

Dependency Parsing as Sequence Labeling with Head-Based Encoding and Multi-Task Learning

Abstract: Dependency parsing as sequence labeling has recently proved to be a relevant alternative to the traditional transitionand graph-based approaches. It offers a good trade-off between parsing accuracy and speed. However, recent work on dependency parsing as sequence labeling ignore the pre-processing time of Part-of-Speech tagging – which is required for this task – in the evaluation of speed while other studies showed that Part-of-Speech tags are not essential to achieve state-ofthe-art parsing scores. In this paper, we compare the accuracy and speed of shared and stacked multi-task learning strategies – as well as a strategy that combines both – to learn Part-of-Speech tagging and dependency parsing in a single sequence labeling pipeline. In addition, we propose an alternative encoding of the dependencies as labels which does not use Part-of-Speech tags and improves dependency parsing accuracy for most of the languages we evaluate.

Templates from syntax to morphology: affix ordering in Qafar

Abstract: The functional head suggests that verbs acquire their inflectional properties by moving from one head position to the next in the syntactic derivation. A problem arises as affixes’ ordering is not sensitive to syntactic properties, as it is the case in Qafar. This Cushitic language exhibits two verbal classes depending on whether verbs can have prefixes. I argue that the hierarchical structure of template corresponds to the syntactic structure. Phonological constraints on templates formation activate adequate syntactic operations. If we assume that templatic domains lie at the interface between syntax and phonology, we account for some issues of affix ordering, that involve no syntactic property.

Question particles in Basque

Abstract: This article proposes separate analyses for the question particles -a and al in Basque, which occur only in polarity questions, within the framework of Generative Grammar. I will propose that the former one, used in the eastern dialects, is the head of FiniteP and that the latter one, used in the central dialects, occupies the head of Particle Phrase located between TP and the CP field. I provide the following evidence in support of this dual analysis: 1) -a can be used with ote but no other particle can appear at the same time with al; 2) al is compatible with allocutivity but -a is not; 4) and, finally, al can be used in embedded clauses, whereas -a cannot. The fact that -a is not allowed to occur in indirect questions and that it is incompatible with the allocutive verbal paradigm shows that it is in complementary distribution with the head of CP and, therefore, that -a occupies such a head; on the other hand, the impossibility of al to appear with other particles suggest that they must occur in the same position and, since it can appear in embedded questions and with allocutive forms, it does not occupy the head of the CP, but the head of a phrase below.

If Birds Have Sesamoid Bones, Do Blackbirds Have Sesamoid Bones? The Modification Effect With Known Compound Words.

Abstract: Three experiments investigate how people infer properties of compound words from the unmodified head. Concepts license inference of properties true of the concept to instances or sub-types of that concept: Knowing that birds generally fly, one infers that a new type of bird flies. However, different names are also believed to reflect real underlying differences. Hence, a different name creates the expectation that a new bird differs from birds in general, and this might impact property inference. In these experiments, participants were told, Almost all (Some, Almost no) birds have sesamoid bones, and then asked, What percentage of blackbirds (birds) have sesamoid bones? The results indicate both inference and contrast effects. People infer properties as less common of the compound than the head when the property is true of the head, but they infer them as more common of the compound than the head when the property is not true of the head. In addition, inferences about properties true of the head are affected by the semantic similarity between the head and the compound, but properties not true of the head do not show any semantic similarity effect, but only a small, consistent effect of contrast. Finally, the presentation format (Open vs. Closed compounds) affects the pattern of effects only when the spacing suggests the existence of a permanent name.

Language Form in The Speech of MPBI-UMS Students Who Roled As The Head of Middle School

Abstract: This study aims to describe the linguistic form in the speech of the Muhammadiyah Surakarta University of Indonesia (MPBI-UMS) Master of Indonesian Language students who portray themselves as junior high school principals. The data in this study are in the form of words, phrases, clauses, and sentences in the speeches of students acting as principals. The source of research data is in the form of student speech transcripts. The data collection technique uses the technique to see and note. The data analysis technique uses the equivalent and final method, while the data validity technique uses theory and source validation. The results in the study show that in the speech of Indonesian language education students the graduate school master's program of Muhammadiyah University of Surakarta covers five fields. First, the field of phonology includes (a) 10 forms of pronunciation error, (b) two forms of capital letters misuse, (c) five forms of italics, and (d) six forms of spelling writing errors. Second, morphology is found in five prepositions. Third, the field of syntax includes (a) four forms of pleonasm errors, (b) four forms of conjunction errors, and (c) four forms of misuse of the redundant word. Fourth, the field of pragmatics includes (a) one form of implicature, (b) one form of expressive speech acts, and (c) two forms of directive speech acts. Fifth, the field of sosiolinguistics includes (a) five forms of code switching, and (b) two forms of code mixing. Sixth, nonformal variety fields are found in one form.

Understanding nonmanuality – A study on the actions of the head and body in Finnish Sign Language

Abstract: Puupponen, Anna Understanding nonmanuality – A study on the actions of the head and body in Finnish Sign Language Jyväskylä: University of Jyväskylä, 2019, 114 p. (JYU Dissertations ISSN 2489-9003; 78) ISBN 978-951-39-7761-0 (PDF) This dissertation, consisting of four articles and this Overview, reports a study on nonmanuality, that is, the actions of the face, head and body, in Finnish Sign Language (FinSL). More specifically, the study focuses on a relatively understudied area of nonmanuality: the actions of the signer’s head and body. The study is theoretically rooted in usage-based linguistics, sign language linguistics, gesture studies and semiotics, and the analysis is made and the conclusions are drawn on the basis of corpus narratives and dialogues as well as synchronized Motion Capture and digital video recordings of FinSL. The study investigates the forms and functions of the actions of the head and body, the relationship between the actions of these two body parts, and the connection between a particular type of head movement, a head nod, and the actions of the signer’s hands in FinSL. The study also presents a theoretical view of the signals from the signer’s head and body according to Peircean and post-Peircean semiotics, and discusses the role of nonmanuality in sign languages. The results of the study show that (i) forms and functions of the actions of the head and body form prototypes rather than discrete classes, and that these actions rarely show conventional pairing of one form to one function; (ii) the head and the torso cannot be seen as one articulator, and the co-occurring signals from these two body parts come together into combinations that differ in their degree of complexity both formally and functionally; (iii) systematicity can be found in the co-occurrence of head nods and manual syntactic units in FinSL; (iv) head and body movements involve different proportions of iconicity, indexicality and symbolicity, of which indexicality is generally the most prominent feature; (v) in signed utterances, nonmanual signals are one part of a semiotically complex but communicationally holistic whole; (vi) nevertheless, there are differences in how complementary co-occurring signals are, and in what the central semiotic features of signals from different parts of the signer’s body are; and finally, (vii) that these semiotic centralities can be partly traced back to the physical and anatomical characteristics of different parts of the signer’s body, and that nonmanual signals demonstrate how signification, language and cognition are intrinsically connected to how humans navigate in their physical and social surroundings with their bodies.

The Derivation of Wh-Questions in Najrani Arabic

Abstract: The study explores the derivation of wh-questions in Najrani Arabic and attempts to answer the following questions: (i) Can wh-questions in Najrani Arabic be derived in VSO or SVO or both?, and (ii) How can Najrani Arabic wh-questions be accounted for within Chomsky’s (2001,2005, 2013,2015 ) Phase approach? The objective of the study is to present a unified analysis of the derivation of wh-questions in Najrani Arabic and show the interaction between Najrani Arabic data and Chomsky’s Phase framework. It has been shown that Najrani Arabic allows the derivation of wh-questions from the argument and non-argument positions in VSO word order. Given this, we assume that VSO is the unmarked order for the derivation of wh-questions in Najrani Arabic. In VSO, the subject DP does not raise to Spec-TP because the head T does not have the EPP feature: the latter attracts movement of the former. The verb raises to the head T of TP, while the subject DP remains in-situ in Spec-vP. Moreover, in Najrani Arabic intransitive structures, the phase vP does not have a specifier because it does not have an external thematic argument whereas in transitive constructions the vP has. Concerning case assignment, the phase vP merges with an abstract tense af (fix) on the head T, which agrees with and assigns invisible nominative case to the subject wh-word man ‘who’. We assume that the phase head C is the probe and has the Edge feature which attracts the raising of the subject wh-phrase to Spec-CP. Besides, we argue that the light transitive head v has an Edged feature which attracts the raising of the object wh-phrase aish ‘what’ to be the second (outer) specifier of vP. Being the phase head, the v probes for a local goal and finds the object wh-phrase aish ; the v agrees with and assigns accusative case to the object wh-phrase aish . As the TP merges with a null interrogative head C, the phase head C has an Edge feature that attracts the raising of the object wh-word aish to Spec-CP for feature valuation. Following this, the null copies of the moved entities left after movement receive a null spellout in the phonological level and, hence, cannot be accessed for any further operation.

System and methods for discovering, presenting, and accessing information in a collection of text contents

Abstract: System and methods are disclosed for discovering and presenting prominent information in a collection of text contents by identifying prominent terms in the text contents, and displaying the terms as either category nodes for organizing the contents in the collection, or as topics in the text contents, or as labels or tags for highlighting the contents in the collection, or for searching the contents in the collection. Methods include distinguishing the grammatical attributes associated with the terms, including the grammatical attributes of a subject and non-subject of a sentence, or a multi-word phrase and a sub-phrase, or a head and a modifier in a phrase, and other distributional attributes of the terms.

The Structure Of Kina Rutul Noun Phrase

Abstract: This paper presents a description of the NP in Kina Rutul (Lezgic, East Caucasian). The research combines corpus study and elicitation. The description provides a list of possible noun modifiers, types of syntactic relations between the head and its dependents, and the order of modifiers within a NP. I provide evidence for analyzing the Rutul numeral ‘one’ as an indefinite article and analyze various types of discontinuous noun phrases, which appear to lose phrasal properties.

On the Computational Complexity of Head Movement and Affix Hopping

Abstract: Head movement is a syntactic operation used in most generative syntactic analyses. However, its computational properties have not been extensively studied. [27] formalises head movement in the framework of Minimalist Grammars by extending the item representation to allow for easy extraction of the head. This work shows that Stabler’s representation is in fact suboptimal because it causes higher polynomial parsing complexity. A new algorithm is derived for parsing head movement and affix hopping by changing the kinds of representations that the parser deals with. This algorithm has much better asymptotic worst-case runtime of \(\mathcal {O}(n^{2k+5})\). This result makes parsing head movement and affix hopping computationally as efficient as parsing a single phrase movement.

Complexity, Naturalness, and Explanatory Power: The Case of Seenku Argument-Head Tone Sandhi

Abstract: When faced with complex alternations, linguists are faced with a dilemma: Given that complex phenomena rarely lend themselves to simple analyses, where is the analytical complexity best justified? Is it better supported in the phonology proper, in the morphology, or in some combination of grammatical components? With phrasal alternations, the question becomes more complicated still, as the role of syntax must also be considered. This paper probes the question of analytical complexity using data from argument-head tone sandhi in Seenku (Western Mande, Burkina Faso). These complex tonal changes occur phrasally, between a head and its internal argument, and depend on phonological, morphological, and syntactic factors. I compare a phonological and lexical approach to the data, arguing that the lexical approach is to be preferred on the grounds of naturalness and explanatory power, and it allows the phonological component to remain restrictive. The Seenku case is reminiscent of another famous tone sandhi debate in the literature, the infamous Taiwanese Min tone circle (Chen 1987, Tsay and Myers 1996, Zhang and Lai 2008, etc.). The basic data are shown in (1):

Las oraciones de relativo y la definitud de su antecedente expreso

Abstract: The paper is concerned with the (in)definite reference of a noun phrase which is the head of a relative clause in Spanish. Speaker and hearer do not share any knowledge of the referent on the basis of previous mention (anaphora) or situational uses. There is something about the relative clause which makes a first-mention definite article possible. We take an insight into the contents of the description conveyed by such relatives.