Publications

Abstract

Languages contain tens of thousands of words, but these are constructed from a tiny handful of phonetic elements. Consequently, words resemble one another, or can be embedded within one another, a coup stick snot with standing. me process of spoken-word recognition by human listeners involves activation of multiple word candidates consistent with the input, and direct competition between activated candidate words. Further, human listeners are sensitive, at an early, prelexical, stage of speeeh processing, to constraints on what could potentially be a word of the language.

Abstract

Speech signals originate as a sequence of linguistic units selected by speakers, but these units are necessarily realised in the suprasegmental dimensions of time, frequency and amplitude. For this reason prosodic structure has been viewed as a mandatory target of language processing by both speakers and listeners. In apparent contradiction, however, prosody has also been argued to be ancillary rather than core linguistic structure, making processing of prosodic structure essentially optional. In the present tribute to one of the luminaries of prosodic research for the past quarter century, we review evidence from studies of the processing of lexical stress and focal accent which reconciles these views and shows that both claims are, each in their own way, fully true.

Abstract

The phoneme detection task is widely used in spoken word recognition research. Alphabetically literate participants, however, are more used to explicit representations of letters than of phonemes. The present study explored whether phoneme detection is sensitive to how target phonemes are, or may be, orthographically realised. Listeners detected the target sounds [b,m,t,f,s,k] in word-initial position in sequences of isolated English words. Response times were faster to the targets [b,m,t], which have consistent word-initial spelling, than to the targets [f,s,k], which are inconsistently spelled, but only when listeners’ attention was drawn to spelling by the presence in the experiment of many irregularly spelled fillers. Within the inconsistent targets [f,s,k], there was no significant difference between responses to targets in words with majority and minority spellings. We conclude that performance in the phoneme detection task is not necessarily sensitive to orthographic effects, but that salient orthographic manipulation can induce such sensitivity.

Abstract

Two lexical decision experiments in Cantonese are described in which the recognition of spoken target words as a function of phonological similarity to a preceding prime is investigated. Phonological similaritv in first syllables produced inhibition, while similarity in second syllables led to facilitation. Differences between syllables in tonal and segmental structure had generally similar effects.

Abstract

This chapter highlights that most language behavior consists of speaking and listening. The chapter also reveals differences and similarities between speaking and listening. The laboratory study of word production raises formidable problems; ensuring that a particular word is produced may subvert the spontaneous production process. Word production is investigated via slips and tip-of-the-tongue (TOT), primarily via instances of processing failure and via the technique of via the picture-naming task. The methodology of word production is explained in the chapter. The chapter also explains the phenomenon of interaction between various stages of word production and the process of speech recognition. In this context, it explores the difference between sound and meaning and examines whether or not the comparisons are appropriate between the processes of recognition and production of spoken words. It also describes the similarities and differences in the structure of the recognition and production systems. Finally, the chapter highlights the common issues in recognition and production research, which include the nuances of frequency of occurrence, morphological structure, and phonological structure.

Abstract

We can describe the position of one item with respect to another using a number of different ‘frames of reference’. For example, I can use a ‘deictic’ frame that involves the speaker’s viewpoint (The chair is on the far side of the room), or an ‘intrinsic’ frame that involves a feature of one of the items (The chair is at the back of the room). Where more than one frame of reference is available in a language, what motivates the speaker’s choice? This elicitation task is designed to explore when and why people select intrinsic frames of reference, and how these choices interact with non-linguistic problem-solving strategies.

Abstract

Different linguistic structures allow us to highlight distinct aspects of a situation. The aim of this survey is to investigate similarities and differences in the expression of situations or events as “stative” (maintaining a state), “inchoative” (adopting a state) and “agentive” (causing something to be in a state). The questionnaire focuses on the encoding of stative, inchoative and agentive possibilities for the translation equivalents of a set of English verbs.

Abstract

Expressions of human activities and states are a rich area for cross-linguistic comparison. Some languages of the world treat human posture verbs (e.g., sit, lie, kneel) as a special class of predicates, with distinct formal properties. This survey examines lexical, semantic and grammatical patterns for posture verbs, with special reference to contrasts between “stative” (maintaining a posture), “inchoative” (adopting a posture), and “agentive” (causing something to adopt a posture) constructions. The enquiry is thematically linked to the more general questionnaire 'Intransitive Predicate Form Class Survey'.

Abstract

This article surveys what is currently known about the complex interplay between genetics and the language sciences. It focuses not only on the genetic architecture of language and speech, but also on their interactions on the cultural and evolutionary timescales. Given the complexity of these issues and their current state of flux and high dynamism, this article surveys the main findings and topics of interest while also briefly introducing the main relevant methods, thus allowing the interested reader to fully appreciate and understand them in their proper context. Of course, not all the relevant publications and resources are mentioned, but this article aims to select the most relevant, promising, or accessible for nonspecialists.

Abstract

The Recent Out-of-Africa human evolutionary model seems to be generally accepted. This impression is very prevalent outside palaeoanthropological circles (including studies of language evolution), but proves to be unwarranted. This paper offers a short review of the main challenges facing ROA and concludes that alternative models based on the concept of metapopulation must be also considered. The implications of such a model for language evolution and diversity are briefly reviewed.

Abstract

The papers in this volume focus on the impact of information structure on language acquisition, thereby taking different linguistic approaches into account. They start from an empirical point of view, and examine data from natural first and second language acquisition, which cover a wide range of varieties, from early learner language to native speaker production and from gesture to Creole prototypes. The central theme is the interplay between principles of information structure and linguistic structure and its impact on the functioning and development of the learner's system. The papers examine language-internal explanatory factors and in particular the communicative and structural forces that push and shape the acquisition process, and its outcome. On the theoretical level, the approach adopted appeals both to formal and communicative constraints on a learner’s language in use. Two empirical domains provide a 'testing ground' for the respective weight of grammatical versus functional determinants in the acquisition process: (1) the expression of finiteness and scope relations at the utterance level and (2) the expression of anaphoric relations at the discourse level.

Abstract

Numerous experiments show that space and musical pitch are
closely linked in people's minds. However, the exact nature of
space-pitch associations and their neuronal underpinnings are
not well understood. In an fMRI experiment we investigated
different types of spatial representations that may underlie
musical pitch. Participants judged stimuli that varied in
spatial height in both the visual and tactile modalities, as well
as auditory stimuli that varied in pitch height. In order to
distinguish between unimodal and multimodal spatial bases of
musical pitch, we examined whether pitch activations were
present in modality-specific (visual or tactile) versus
multimodal (visual and tactile) regions active during spatial
height processing. Judgments of musical pitch were found to
activate unimodal visual areas, suggesting that space-pitch
associations may involve modality-specific spatial
representations, supporting a key assumption of embodied
theories of metaphorical mental representation.

Abstract

This paper summarizes the results of a descriptive syntactic category analysis of child English no which reveals that young children use and represent no as a determiner and negatives like no pen as NPs, contra standard analyses.

Abstract

Previous studies have demonstrated that native listeners
modify their interpretation of a speech sound when a talker
produces an ambiguous sound in order to quickly tune into a
speaker, but there is hardly any evidence that non-native
listeners employ a similar mechanism when encountering
ambiguous pronunciations. So far, one study demonstrated
this lexically-guided perceptual learning effect for nonnatives,
using phoneme categories similar in the native
language of the listeners and the non-native language of the
stimulus materials. The present study investigates the question
whether phoneme category retuning is possible in a nonnative
language for a contrast, /l/-/r/, which is phonetically
differently embedded in the native (Dutch) and nonnative
(English) languages involved. Listening experiments indeed
showed a lexically-guided perceptual learning effect.
Assuming that Dutch listeners have different phoneme
categories for the native Dutch and non-native English /r/, as
marked differences between the languages exist for /r/, these
results, for the first time, seem to suggest that listeners are not
only able to retune their native phoneme categories but also
their non-native phoneme categories to include ambiguous
pronunciations.

Abstract

This paper presents Advanced Glossing, a proposal for a general glossing format designed for language documentation, and a specific setup for the Shoebox-program that implements Advanced Glossing to a large extent. Advanced Glossing (AG) goes beyond the traditional Interlinear Morphemic Translation, keeping syntactic and morphological information apart from each other in separate glossing tables. AG provides specific lines for different kinds of annotation – phonetic, phonological, orthographical, prosodic, categorial, structural, relational, and semantic, and it allows for gradual and successive, incomplete, and partial filling in case that some information may be irrelevant, unknown or uncertain. The implementation of AG in Shoebox sets up several databases. Each documented text is represented as a file of syntactic glossings. The morphological glossings are kept in a separate database. As an additional feature interaction with lexical databases is possible. The implementation makes use of the interlinearizing automatism provided by Shoebox, thus obtaining the table format for the alignment of lines in cells, and for semi-automatic filling-in of information in glossing tables which has been extracted from databases

Abstract

Digitizing and annotating texts and field recordings Given that several initiatives worldwide currently explore the new field of documentation of endangered languages, the E-MELD project proposes to survey and unite procedures, techniques and results in order to achieve its main goal, ''the formulation and promulgation of best practice in linguistic markup of texts and lexicons''. In this context, this year's workshop deals with the processing of recorded texts. I assume the most valuable contribution I could make to the workshop is to show the procedures and methods used in the Awetí Language Documentation Project. The procedures applied in the Awetí Project are not necessarily representative of all the projects in the DOBES program, and they may very well fall short in several respects of being best practice, but I hope they might provide a good and concrete starting point for comparison, criticism and further discussion. The procedures to be exposed include: * taping with digital devices, * digitizing (preliminarily in the field, later definitely by the TIDEL-team at the Max Planck Institute in Nijmegen), * segmenting and transcribing, using the transcriber computer program, * translating (on paper, or while transcribing), * adding more specific annotation, using the Shoebox program, * converting the annotation to the ELAN-format developed by the TIDEL-team, and doing annotation with ELAN. Focus will be on the different types of annotation. Especially, I will present, justify and discuss Advanced Glossing, a text annotation format developed by H.-H. Lieb and myself designed for language documentation. It will be shown how Advanced Glossing can be applied using the Shoebox program. The Shoebox setup used in the Awetí Project will be shown in greater detail, including lexical databases and semi-automatic interaction between different database types (jumping, interlinearization). ( Freie Universität Berlin and Museu Paraense Emílio Goeldi, with funding from the Volkswagen Foundation.)

Abstract

Conclusion In this study we have examined the evidence for the exact genetic position of the Awetí language in the large Tupí family, especially evidence for an internal classification of the larger branch of Tupí called “Mawetí-Guaraní” which comprises the Tupí-Guaraní family, Awetí and Sateré-Mawé. As it turns out, we did not find any clear example of an uncommon sound change which would have happened after the separation of the antecessor of one branch but before the split between the other two. There is some just probability that Awetí belongs somewhat closer to Tupí-Guaraní within Mawetí-Guaraní (configuration A in Table 1), but we did not find any conclusive evidence. All we have are some weak indications the majority of which, however, point in this direction: • a higher number of cognates found between Awetí and proto-Tupí-Guarani; • lexicostatistic results (number of cognates in a 100-item-word-list proposed by Swadesh); • loss of long vowels in Awetí and Tupí-Guaraní, but not in Sateré-Mawé; • some sound changes suggest that in the development to Awetí and to proto-Tupí-Guaraní velar segments changes to dental segments (cf. the discussion of the correspondence set j : t : w); • possibly some of the correspondence sets given in Table 20. We consider it to be too soon to conclude that there is a branch Awetí + Tupí-Guaraní of Mawetí-Guaraní, opposed to Sateré-Mawé, but if there is any grouping, this hypothesis is most promising. 29

Abstract

Traditionally, there have been three standard assumptions made about the Parallelism Effect on VP-ellipsis, namely that the effect is categorical, that it applies asymmetrically and that it is uniquely due to syntactic factors. Based on the results of a series of experiments involving online and offline tasks, it will be argued that the Parallelism Effect is instead noncategorical and interactional. The factors investigated include construction type, conceptual and morpho-syntactic recoverability, finiteness and anaphor type (to test VP-anaphora). The results show that parallelism is gradient rather than categorical, effects both VP-ellipsis and anaphora, and is influenced by both structural and non-structural factors.

Abstract

The Island Melanesia project was initiated to collect data on the little-known Papuan languages of Island Melanesia, and to explore the origins of and relationships between these languages. The project materials from the 2003 field season focus on language related to cultural domains (e.g., material culture) and on targeted grammatical description. Five tasks are included: Proto-Oceanic lexicon, Grammatical questionnaire and lexicon, Kinship questionnaire, Domains of likely pre-Austronesian terminology, and Botanical collection questionnaire.

Abstract

This task is designed to elicit virtual 3D ‘models’ created in gesture space using iconic and other representational gestures. This task has been piloted with Lao speakers, where two speakers were asked to explain the meaning of terms referring to different kinds of fish trap mechanisms. The task elicited complex performances involving a range of iconic gestures, and with especially interesting use of (a) the ‘model/diagram’ in gesture space as a virtual object, (b) the non-dominant hand as a prosodic/semiotic anchor, (c) a range of different techniques (indexical and iconic) for evoking meaning with the hand, and (d) the use of nearby objects and parts of the body as semiotic ‘props’.

Abstract

Laos features a high level of linguistic diversity, with more than 70 languages from four different major language families (Tai, Mon-Khmer, Hmong-Mien, Tibeto-Burman). Mon-Khmer languages were spoken in Laos earlier than other languages, with incoming migrations by Tai speakers (c. 2000 years ago) and Hmong-Mien speakers (c. 200 years ago). There is widespread language contact and multilingualism in upland minority communities, while lowland-dwelling Lao speakers are largely monolingual. Lao is the official national language. Most minority languages are endangered, with a few exceptions (notably Hmong and Kmhmu). There has been relatively little linguistic research on languages of Laos, due to problems of both infrastructure and administration.

Abstract

This chapter discusses some of the elements of human sociality that serve as the social and cognitive infrastructure or preconditions for the use of requests and other kinds of recruitments in interaction. The notion of an agent with goals is a canonical starting point, though importantly agency tends not to be wholly located in individuals, but rather is socially distributed. This is well illustrated in the case of requests, in which the person or group that has a certain goal is not necessarily the one who carries out the behavior towards that goal. The chapter focuses on the role of semiotic (mostly linguistic) resources in negotiating the distribution of agency with request-like actions, with examples from video-recorded interaction in Lao, a language spoken in Laos and nearby countries. The examples illustrate five hallmarks of requesting in human interaction, which show some ways in which our ‘manipulation’ of other people is quite unlike our manipulation of tools: (1) that even though B is being manipulated, B wants to help, (2) that while A is manipulating B now, A may be manipulated in return later; (3) that the goal of the behavior may be shared between A and B, (4) that B may not comply, or may comply differently than requested, due to actual or potential contingencies, and (5) that A and B are accountable to one another; reasons may be asked for, and/or given, for the request. These hallmarks of requesting are grounded in a prosocial framework of human agency.

Abstract

Research on video- and audio-recordings of spontaneous naturally-occurring conversation in English has shown that conversation is a rule-guided, practice-oriented domain that can be investigated for its underlying mechanics or structure. Systematic study could yield something like a grammar for conversation. The goal of this task is to acquire a corpus of video-data, for investigating the underlying structure(s) of interaction cross-linguistically and cross-culturally

Abstract

We want to know how people think about their field of kin, on the supposition that it is quasi-spatial. To get some insights here, we need to video a discussion about kinship reckoning, the kinship system, marriage rules and so on, with a view to looking at both the linguistic expressions involved, and the gestures people use to indicate kinship groups and relations. Unlike the task in the 2001 manual, this task is a direct interview method.

Abstract

What causes a language to be the way it is? Some features are universal, some are inherited, others are borrowed, and yet others are internally innovated. But no matter where a bit of language is from, it will only exist if it has been diffused and kept in circulation through social interaction in the history of a community. This book makes the case that a proper understanding of the ontology of language systems has to be grounded in the causal mechanisms by which linguistic items are socially transmitted, in communicative contexts. A biased transmission model provides a basis for understanding why certain things and not others are likely to develop, spread, and stick in languages. Because bits of language are always parts of systems, we also need to show how it is that items of knowledge and behavior become structured wholes. The book argues that to achieve this, we need to see how causal processes apply in multiple frames or 'time scales' simultaneously, and we need to understand and address each and all of these frames in our work on language. This forces us to confront implications that are not always comfortable: for example, that "a language" is not a real thing but a convenient fiction, that language-internal and language-external processes have a lot in common, and that tree diagrams are poor conceptual tools for understanding the history of languages. By exploring avenues for clear solutions to these problems, this book suggests a conceptual framework for ultimately explaining, in causal terms, what languages are like and why they are like that.

Abstract

This task is a complement to the questionnaire ‘Multimodal interaction in your field site: a preliminary investigation’. The objective of the task is to obtain high quality video data on structured and symmetrical dyadic multimodal interaction. The features of interaction we are interested in include turn organization in speech and nonverbal behavior, eye-gaze behavior, use of composite signals (i.e. communicative units of speech-combined-with-gesture), and linguistic and other resources for ‘navigating’ interaction (e.g. words like okay, now, well, and um).

Abstract

This article introduces a new speech corpus, the Nijmegen Corpus of Casual Czech (NCCCz), which contains more than 30 hours of high-quality recordings of casual conversations in Common Czech, among ten groups of three male and ten groups of three female friends. All speakers were native speakers of Czech, raised in Prague or in the region of Central Bohemia, and were between 19 and 26 years old. Every group of speakers consisted of one confederate, who was instructed to keep the conversations lively, and two speakers naive to the purposes of the recordings. The naive speakers were engaged in conversations for approximately 90 minutes, while the confederate joined them for approximately the last 72 minutes. The corpus was orthographically annotated by experienced transcribers and this orthographic transcription was aligned with the speech signal. In addition, the conversations were videotaped. This corpus can form the basis for all types of research on casual conversations in Czech, including phonetic research and research on how to improve automatic speech recognition. The corpus will be freely available

Abstract

With the aim to clarify the definition of humans as “linguistic animals”, in the present paper I functionally distinguish three types of language competences: i) language as a general biological tool for communication, ii) “perceptual syntax”, iii) propositional language. Following this terminological distinction, I review pivotal findings on animals' communication systems, which constitute useful evidence for the investigation of the nature of three core components of humans' faculty of language: semantics, syntax, and theory of mind. In fact, despite the capacity to process and share utterances with an open-ended structure is uniquely human, some isolated components of our linguistic competence are in common with nonhuman animals. Therefore, as I argue in the present paper, the investigation of animals' communicative competence provide crucial insights into the range of cognitive constraints underlying humans' ability of language, enabling at the same time the analysis of its phylogenetic path as well as of the selective pressures that have led to its emergence.

Abstract

The aim of this study is to investigate the word-learning phenomenon utilizing a new model that integrates three processes: a) extracting a word out of a continuous sounds sequence, b) inducing referential meanings, c) mapping a word onto its intended referent, with the possibility to extend the acquired word over a potentially infinite sets of objects of the same semantic category, and over not-previously-heard utterances. Previous work has examined the role of statistical learning and/or of prosody in each of these processes separately. In order to examine the multilayered word-learning task, we integrate these two strands of investigation into a single approach. We have conducted the study on adults and included six different experimental conditions, each including specific perceptual manipulations of the signal. In condition 1, the only cue to word-meaning mapping was the co-occurrence between words and referents (“statistical cue”). This cue was present in all the conditions. In condition 2, we added infant-directed-speech (IDS) typical pitch enhancement as a marker of the target word and of the statistical cue. In condition 3 we placed IDS typical pitch enhancement on random words of the utterances, i.e. inconsistently matching the statistical cue. In conditions 4, 5 and 6 we manipulated respectively duration, a non-prosodic acoustic cue and a visual cue as markers of the target word and of the statistical cue. Systematic comparisons between learning performance in condition 1 with the other conditions revealed that the word-learning process is facilitated only when pitch prominence consistently marks the target word and the statistical cue…

Abstract

The mysterious human propensity for acquiring speech and language has fascinated scientists for decades. A substantial body of evidence suggests that this capacity is rooted in aspects of neurodevelopment that are specified at the genomic level. Researchers have begun to identify genetic factors that increase susceptibility to developmental disorders of speech and language, thereby offering the first molecular entry points into neuronal mechanisms underlying human vocal communication. The identification of genetic variants influencing language acquisition facilitates the analysis of animal models in which the corresponding orthologs are disrupted. At face value, the situation raises aperplexing question: if speech and language are uniquely human, can any relevant insights be gained from investigations of gene function in other species? This chapter addresses the question using the example of FOXP2, a gene implicated in a severe monogenic speech and language disorder. FOXP2 encodes a transcription factor that is highly conserved in vertebrate species, both in terms of protein sequence and expression patterns. Current data suggest that an earlier version of this gene, present in the common ancestor of humans, rodents, and birds, was already involved in establishing neuronal circuits underlying sensory-motor integration and learning of complex motor sequences. This may have represented one of the factors providing a permissive neural environment for subsequent evolution of vocal learning. Thus, dissection of neuromolecular pathways regulated by Foxp2 in nonlinguistic species is a necessary prerequisite for understanding the role of the human version of the gene in speech and language.

Abstract

Relative clauses are a syntactic device to create complex sentences and they make language structurally productive. Despite a considerable number of experimental studies, it is still largely unclear how children learn relative clauses and how these are processed in the language system. Researchers at the MPI for Psycholinguistics used a computational learning model to gain novel insights into these issues. The model explains the differential development of relative clauses in English as well as cross-linguistic differences

Abstract

This chapter connects the grammar of the first person collective pronoun in the Cha’palaa language of Ecuador with its use in interaction for collective reference and social category membership attribution, addressing the problem posed by the fact that non-singular pronouns do not have distributional semantics (“speakers”) but are rather associational (“speaker and relevant associates”). It advocates a cross-disciplinary approach that jointly considers elements of linguistic form, situated usages of those forms in instances of interaction, and the broader ethnographic context of those instances. Focusing on large-scale and relatively stable categories such as racial and ethnic groups, it argues that looking at how speakers categorize themselves and others in the speech situation by using pronouns provides empirical data on the status of macro-social categories for members of a society

Abstract

This paper examines code and style shifting during sales transactions based on two market case studies from highland Ecuador. Bringing together ideas of linguistic economy with work on stylistic variation and ethnohistorical research on Andean markets, I study bartering, market calls and sales pitches to show how sellers create stylistic performances distinguished by contrasts of code, register and poetic features. The interaction of the symbolic value of language with the economic values of the market presents a place to examine the relationship between discourse and the material world.

Abstract

Reading is an audiovisual process that requires the learning of systematic links between graphemes and phonemes. It is thus possible that reading impairments reflect an audiovisual processing deficit. In this study, we compared audiovisual processing in adults with developmental dyslexia and adults without reading difficulties. We focused on differences in cross-modal temporal sensitivity both for speech and for non-speech events. When compared to adults without reading difficulties, adults with developmental dyslexia presented a wider temporal window in which unsynchronized speech events were perceived as synchronized. No differences were found between groups for the non-speech events. These results suggests a deficit in dyslexia in the perception of cross-modal temporal synchrony for speech events.

Abstract

Several sources of information are used in choosing the intended referent of an ambiguous pronoun. The two sources considered in this paper are foregrounding and context. The first refers to the accessibility of discourse entities. An entity that is foregrounded is more likely to become the pronoun’s referent than an entity that is not. Context information affects pronoun resolution when world knowledge is needed to find the referent. The model presented here simulates how world knowledge invoked by context, together with foregrounding, influences pronoun resolution. It was developed as an extension to the Distributed Situation Space (DSS) model of knowledge-based inferencing in story comprehension (Frank, Koppen, Noordman, & Vonk, 2003), which shall be introduced first.

Abstract

This project is part of a collaborative project with the research group “Reciprocals across languages” led by Nick Evans. One goal of this project is to develop a typology of reciprocals. This questionnaire is designed to help field workers get an overview over the type of markers used in the expression of reciprocity in the language studied.

Abstract

The aim of this paper is to explore whether indexing terms for an audiovisual program can be derived from contextual texts automatically. For this we apply natural-language processing techniques to contextual texts of two Dutch TV-programs. We use a Dutch domain thesaurus to derive possible metadata. This possible metadata is ranked by an algorithm which uses the relations of the thesaurus. We evaluate the results by comparing them to human made descriptions.

Abstract

We present a solution to the problem of online speaker/signer diarization - the task of determining "who spoke/signed when?". Our solution is based on the idea that gestural activity (hands and body movement) is highly correlated with uttering activity. This correlation is necessarily true for sign languages and mostly true for spoken languages. The novel part of our solution is the use of motion history images (MHI) as a likelihood measure for probabilistically detecting uttering activities. MHI is an efficient representation of where and how motion occurred for a fixed period of time. We conducted experiments on 4.9 hours of a publicly available dataset (the AMI meeting data) and 1.4 hours of sign language dataset (Kata Kolok data). The best performance obtained is 15.70% for sign language and 31.90% for spoken language (measurements are in DER). These results show that our solution is applicable in real-world applications like video conferences.

Abstract

We demonstrate how the problem of speaker diarization can be solved using both gesture and speaker parametric models. The novelty of our solution is that we approach the speaker diarization problem as a speaker recognition problem after learning speaker models from speech samples corresponding to gestures (the occurrence of gestures indicates the presence of speech and the location of gestures indicates the identity of the speaker). This new approach offers many advantages: comparable state-of-the-art performance, faster computation and more adaptability. In our implementation, parametric models are used to model speakers' voice and their gestures: more specifically, Gaussian mixture models are used to model the voice characteristics of each person and all persons, and gamma distributions are used to model gestural activity based on features extracted from Motion History Images. Tests on 4.24 hours of the AMI meeting data show that our solution makes DER score improvements of 19% on speech-only segments and 4% on all segments including silence (the comparison is with the AMI system).

Abstract

Prior research on language identification focused primarily on text and speech. In this paper, we focus on the visual modality and present a method for identifying sign languages solely from short video samples. The method is trained on unlabelled video data (unsupervised feature learning) and using these features, it is trained to discriminate between six sign languages (supervised learning). We ran experiments on video samples involving 30 signers (running for a total of 6 hours). Using leave-one-signer-out cross-validation, our evaluation on short video samples shows an average best accuracy of 84%. Given that sign languages are under-resourced, unsupervised feature learning techniques are the right tools and our results indicate that this is realistic for sign language identification.

Abstract

Acquiring the sounds of a language involves learning to recognize distributional patterns present in the input. We show that among adult learners, this distributional learning of auditory categories (which are conceived of here as probability density functions in a multidimensional space) is constrained by the salience of the dimensions that form the axes of this perceptual space. Only with a particular ratio of variation in the perceptual dimensions was category learning driven by the distributional properties of the input.