Màster Oficial - Ciència Cognitiva i Llenguatge (CCiL)

URI permanent per a aquesta col·leccióhttps://hdl.handle.net/2445/61025

Treballs Finals del Màster de Ciència Cognitiva i Llenguatge de la Facultat de Filosofia de la Universitat de Barcelona.

Estadístiques

Examinar

Mostrant 1 - 20 de 25

Applying semantic prosody for machine translation improvement on English-Chinese passive sentences
(2024-09) Ma, Xinyue; Farrús, Mireia
Passive sentences are widely used in English and have a neutral semantic prosody, while in Chinese they are much less frequent and with negative semantic prosody. Thus, many times English passives should be translated into active Chinese sentences and passive voice should be mainly reserved for unfavorable content. This work uses focused datasets to fine-tune a neural network machine translation model and improves its performance on translating English BE passives to Chinese. Evaluation after fine-tuning suggests training on English-Chinese sentence pairs that contain BE passives that are translated into actives by human translator, but to BEI passives by the model leads to considerable improvement of model performance. Also, data shows that the semantic orientation score of BE passives translated into Chinese passives by human translators are significantly different from those translated into active sentences, providing quantifiable way to address this issue.
Acoustic study of non-interrogative usages of the interrogative pronoun in Mandarin
(2024-09) Wang, Jiaqi; Farrús, Mireia
In Mandarin Chinese, the interrogative pronoun “shen2me0” not only can be used to express interrogation, but also has multiple non-interrogative usages. In sentences with same syntactic structure, by applying different intonations, this wh-word can convey various meanings. However, due to the lack of grammatical markers, it could be a complex problem for Automatic Speech Recognition (ASR). Therefore, this study chose to investigate the acoustic features of these usages of “shen2me0”. Through two experiments, this present compared the acoustic features including pitch contours, sentence stress, duration, pitch range, boundary tone of sentence and the wh-word of the interrogative, empty reference, rhetorical, and referential substitution usages in same sentences under different contexts. Also, the use of modal particles at the end of the sentences was considered. The results showed that interrogative usage had moderate pitch fluctuation and the use of modal particle would influence how people pronounced the wh-word. In empty reference usage, “shen2me0” had a neutral nature as a placeholder rather than a focus, and thus the sentence showed a flat pitch curve. Rhetorical usage had dynamic pitch changes, especially at the word “hai2”, to express strong emotion. Referential substitution usage had a flatter pitch curve at the beginning, which rose higher at the end, with prolonged pronunciation of “me0”. Moreover, this study also discussed about meaning for ASR and the improvements in further study.
The Influence of Individuals’ Musical Training and Pragmatic Skills on the Perception of f0 Manipulation in a Foreign Language
(2024-09) Gerenčir, Larisa; Elvira-García, Wendy
In speech perception studies, particularly of intonation, several factors have previously been noticed to influence one’s perceptive skills and, accordingly, their ability to perceive smaller changes in intonation. While sociodemographic factors like gender and age have often been studied in this context, other factors like musical training and individual differences, particularly individual pragmatic skills, are gradually becoming more interesting for those studying speech perception. Playing an instrument is argued to involve many highly complex cognitive functions and affect neuroplasticity and importantly, it has been studied in connection with auditory processing. Furthermore, the study of the influence of individual skills on auditory processing shows a positive relation between higher pragmatic skills and perception. While both of these factors are proven significant, the research tends to focus on one of them, rarely comparing the two skills or examining which one has a higher influence on the perception of intonation. The present study addresses this gap by examining Croatian native speakers’ perception of f0 stimulation in Catalan, a language they are unfamiliar with. 91 participants took a discrimination test in which they listened to 20 pairs of stimuli and had to answer whether the stimuli in each pair were the same. The analysis was done in two steps, first, through a confusion matrix and analysis of answers for each pair of stimuli, and second, by building three linear regression models in order to compare the relationship between musical training and pragmatic skills and the correct answers. The results demonstrate a good overall precision, but lower recall score and establish the fourth pair of stimuli as a threshold in which the majority of participants observe the difference between the stimuli. The second model, which included musical training, but no pragmatic skills as a variable, has proven to be the best-performing one, showing that in this study musical training played a more significant role than pragmatic skills in individuals’ ability to perceive changes of f0 in an unfamiliar language.
Exploring the role of prosodic information as a modality in hate speech detection
(2024-09) Duque Maldonado, Alejandra; Farrús, Mireia
The multimodal detection of hate speech has been a trending topic across different disciplines in the recent years. New approaches make use of multimodal techniques to target and mitigate toxic behaviour. Current proposals, despite relying on various modalities, tend to prioritize the use of text in monolingual settings, typically with English. In this thesis, we want to avoid using textual data, and rather focus on the audio modality to see if its properties can help us target toxic speech. Given that within speech prosody we could target possible cues related to indicators of hate speech (e.g. emotional speech), our aim is to test the effectiveness of predicting toxic speech based on prosodic features obtained from the audio. We used two different classification methods to test whether the use of these audio properties was satisfactory. These algorithms were trained on a database of YouTube videos in Spanish, within which there were examples of hate speech towards gender and equality speeches. These examples were processed to work only with the speech signal, which naturally reflected prosodic properties. In the scope of our application, we saw that our traditional machine learning approach suggests it is possible to detect hate speech based on prosodic information. By using frame-wise information of the audio, we have seen that it is possible to classify hate speech automatically. Our proposal opens the door for future works to continue testing the effectiveness of including this information, and encourages future proposals to merge prosodic information with other modalities.
On the relation between language and abstract thought: a cross-linguistic analysis on Infinity and its origins
(2024-09) Calonge Cases, Anna; Dediu, Dan
Delving into the study of abstract concept development from an interdisciplinary perspective, mainly philosophical and linguistic, the main purpose of the thesis is to examine the interrelation between linguistic phenomena and the emergence and development of one very specific (although highly complex and multifaceted) abstract concept: infinity. Following the hypothesis that the abstract notion of infinity is a linguistically modulated concept, this research aims to investigate the potentially substantial role that language may have played in the emergence and evolution of the infinity concept by examining whether morphological and syntactic processes may have facilitated or promoted its development. Through the analysis of all the words used in modern English to designate infinity, whether temporal, limitless, numerical and so on, and through the cross-linguistic examination of the infinity words and phrases in the New Testament across English, Spanish, Latin and Koine Greek (the original language), several findings were obtained. Firstly, the detection of a pattern of lexicalisation across the languages: beyond the general notion of infinity as a lack of limits or measurability, the specific notions of power-related infinity and, most noticeably, temporal infinity have been largely lexicalised, manifesting a cultural need to discuss, reflect and talk about omnipotence and, especially, eternity. The prominence of temporal infinity is also evidenced by the fact that the only simple lexemes refering to infinity directly, and not resulting from lexicogenic operations, are words designating temporal infinity. All other expressions bring about the meaning of infinitude from compositional processes, in particular, morphosyntactic processes of negation, universal quantification and the depiction of forward motion across a temporal medium. While it has not been possible to confirm the hypothesis whether these linguistic processes are necessary or influential for the inception of the abstract concept, given that the four analysed languages share the same mechanisms, the findings open new research paths. Particularly, this methodology can be replicated for other languages, thus building a large cross-linguistic database of infinity words and expressions in the New Testament, aiming at a better understanding of the interrelation between language and cognition, particularly in abstract thought. Ultimately, the prominence of temporal infinity, from which other notions of specific infinitude and the general concept of infinity itself probably arose, constitutes a topic worth exploring.
Two Types of Exclamatives in Mandarin Chinese
(2023) Wang, Yining; Mendia, Jon Ander
This paper explores two types of exclamatives in Mandarin Chinese, as Badan and Cheng (2015) identified. Whilst type I exclamatives employ degree demonstratives zheme/name (this.ME/that.ME) and exclusively have the surprise reading, type II exclamatives recruit the degree quantifier duome (much.ME) and exclusively have the non-surprise reading. In this paper, a lexical approach is taken to account for the contrast in their expressions of surprise and non-surprise. The central claim is that this distinction is firmly rooted in the semantics of the degree adverbs involved. To be precise, the degree demonstrative zheme/name operates within equative comparatives, where it selects the expected degree according to the speaker's perspective as the standard of comparison, and the sense of surprise emerges when the actual degree surpasses this initial expectation. In contrast, the degree quantifier duome predominantly contributes to positive constructions, simply demanding the degree in question to be noteworthy in comparison to a contextually given standard, without invoking surprise. Moreover, their syntactic properties and the resulting compositional consequences are also discussed.
Moral Perception: seeing the Good, the Bad and the Ugly
(2023) González Arias, Lucía; Toribio Mateas, Josefa
Attentional moral perception (AMP) is the most recent and less-controversial version of moral perceptualism, the claim that we can represent moral properties such as being right or wrong in perception. AMP features an attentional, albeit perceptual sensitivity to moral difference-makers that need not involve the perceptual representation of the moral properties themselves. In this essay, I examine the empirical and theoretical basis of this proposal, delving into the methodological imprecisions and problematic implications that comes with endorsing it. On the empirical side, I cast doubt on the validity of the data and the fitness of the experimental designs typically invoked to defend the view. On the theoretical side, I articulate a more controlled example and, borrowing a classic argument from Speaks (2010), I argue that AMP ultimately needs to commit to phenomenologically rich ‘attentional states’ and renounce the transparency of perception. Clinical and evolutive aspects of AMP are also discussed.
Modalized Robust Virtue Epistemology Defended
(2023) Lara Crosas, Marc; Broncano, Fernando, 1954-
This essay addresses one of the most central and longstanding problems in epistemology: What is the nature of knowledge? According to robust virtue epistemology, which is a popular answer to this problem, knowledge is explained solely by appealing to cognitive abilities, epistemic competences, or intellectual virtues. In this essay, I defend a reliabilist and modalized version of robust virtue epistemology. More specifically, the main thesis of this essay is that Mortini's (2022) reformulation of the safety principle as "environment-relative safety", which is the best response to Kelp's (2009; 2016; 2018) safety dilemma, makes robust virtue epistemology even more plausible. In section 2, after the introduction, I will present Pritchard's defense of a modest (non-robust) virtue epistemology based on the independence thesis. In section 3, Kelp's (2009; 2016; 2018) safety dilemma, which is an objection to the necessity of safety for knowledge, is presented. However, I will argue that Mortini's (2022) reformulation of safety satisfactorily answers this objection and that it is the best answer to this dilemma thus far in the literature. Finally, in section 4, I argue that the manifestation of cognitive abilities ought to be relativized to actual features of the environment. And I will argue that the satisfaction of the ability condition relativized to those features entails the satisfaction of environment-relative safety.
Semántica de los refranes: la importancia del contexto cultural y del contexto lingüístico
(2022) Oteiza Irigoyen, Maialen; Castellón Masalles, Irene
Debido a su largo recorrido, existen varios puntos de vista desde los que afrontar la paremiología o estudio de los refranes. El presente trabajo tratará la semántica de dichos elementos que tanto usamos los hablantes de una lengua para expresar o reforzar ciertas ideas. Todos nosotros utilizamos proverbios en nuestro día a día, pero ¿quiere decir esto que todos los refranes son iguales o que se forman y entienden de la misma manera? Esta es la pregunta base de la que parte este estudio. Más concretamente, se ha querido ver cómo es la semántica de los refranes en diferentes lenguas y si el contexto cultural y el contexto lingüístico afectan a la hora de crear, usar o comprender las diferentes paremias. Para conseguir dicho objetivo, este trabajo consta de tres secciones principales. La primera, el marco teórico, nos explica los conceptos teóricos que se analizarán en las próximas secciones. De esta manera, se hablará de la difícil definición de refrán, de las visiones que existen para enfocar el estudio del proverbio, del significado literal y figurado de los refranes, del contexto cultural y el contexto lingüístico, así como de la importancia que estos pueden tener en las paremias, de las figuras que aparecen en los refranes – haciendo especial referencia a las metáforas conceptuales y a su posible universalidad – y de las dos ramas de la lingüística que se han tenido en cuenta en el enfoque de este trabajo.
Aproximación al deísmo en español: propuesta de clasificación y análisis
(2022) Bañeras Carrió, Maria; Hernanz, Ma. Lluïsa
El objetivo de este estudio es profundizar en el deísmo en español, que consiste en la aparición inesperada de de encabezando ciertas construcciones de infinitivo. Quizás por ser considerado un fenómeno vulgar y fuera de la norma, ha recibido escasa atención en las gramáticas de la lengua española de referencia. De hecho, en la actual Nueva gramática de la lengua española (RAE-ASALE 2009) solamente se le dedica un breve subapartado, el 43.6u. Ahí se describe el deísmo como una característica propia de la lengua medieval que “permanece todavía en el español no estándar de algunas regiones, aunque es uso muy desprestigiado”, añadiendo que “estas construcciones se consideran anómalas en el español actual, por lo que se recomienda evitarlas” (RAE-ASALE 2009). Este trabajo aspira a ser una aproximación no exhaustiva al fenómeno, entendiendo que es fundamental estudiarlo en relación con construcciones similares en otras lenguas romances. No se pretende, por otra parte, llevar a cabo un estudio minucioso a nivel descriptivo y dialectal del fenómeno, sino tener una visión de conjunto y sugerir posibles explicaciones.
The Bilingual Moses Illusion: Evidence for Semantic Illusions in Spanish-Catalan Bilingual Speakers
(2022-09) Bautista Martín, Ana María; Ferré Romeu, Pilar; Comesaña, Montserrat
The tendency to overlook semantic inaccuracies in questions like ‘How many animals of each kind did Moses take on the ark?’, also known as the Moses illusion, was introduced into research almost forty years ago. However, the cross-linguistic validation of this effect has been limited, and studies regarding bilingual speakers are still sparse and controversial, even though some theoretical accounts predict that processing in a second language may modulate the rates of semantic illusions by either boosting or hindering the detection of inaccuracies. The present investigation specifically addressed these two concerns, firstly by examining whether the Moses illusion takes place in Spanish native speakers, and subsequently by exploring how it is elicited in the languages of early and highly proficient Spanish-Catalan bilingual speakers. Implemented in the form of a sentential judgement task for semantically correct and incorrect sentences adding the Self-Paced reading technique, the experiments’ results revealed that the Moses illusion happened in Spanish (Experiment 1), and with no statistical differences across Spanish and Catalan (Experiment 2). These findings indicate that semantic illusions take place similarly in both languages of Spanish-Catalan bilingual speakers. Further research is required to consolidate these findings and to build a more solid conceptual framework for the bilingual Moses illusion and second language processing.
The imperative future in Nort-Western Catalan
(2022) Escolà Sallán, Laia; Espinal, M. Teresa (Maria Teresa)
This study deals with a North-Western Catalan phenomenon called imperative future. The paper aims to give this phenomenon an accurate representation. The imperative future consists of using the future form of the verb to give an order. The future is usually used to refer to an event posterior to the time of the utterance. Hence, using this verbal form to give an order is somewhat unexpected. As its name indicates, it has a similar meaning to an imperative at first sight, as it is used to command the addressee.
Lexical cohesion and LSA coherence in schizophrenic patients with and without thought disorder
(2022) Olmo Encabo, Pablo del; Rosselló Ximenes, Joana
Language abnormalities called Formal Thought Disorder may be present in patients with schizophrenia. Disorganized speech is one of these abnormalities, and it may be studied by examining the semantic relationships between parts of the text. Two such approaches are lexical cohesion and LSA coherence. Previous studies have found apparently contradictory evidence, with lexical cohesion being increased in the discourse of patients with Formal Thought Disorder and LSA coherence being decreased. In this study, we tested the idea that these findings are not contradictory, and that one may find both these disturbances in the same sample of patients, with both explaining relevant clinical variables independently of each other. Results didn’t provide strong support for our hypotheses. We attempt to provide explanations for this failure to find the expected results. Methodological differences with previous literature, low subject number and lack of control of certain symptomatology variables may explain our null findings.
The Dominant Participle Construction as a Syntax-Semantics Interface Phenomenon:The Case Study of Latin
(2022) Vendrell Morist, Maria; Mateu Fontanals, Jaume
The syntactic representation of participles has remained a disputed issue at the centre of linguistic research for some time. Much of the confusion surrounding participles is the result of their categorial indeterminacy, which ranges between the category of verbs and adjectives. The blended categorial properties of participles are mirrored in their flexible syntactic distribution. For example, participles are flexible enough to fulfil prototypical adjectival functions in attributive and copular structures, e.g. ‘the sewed button’ and ‘the button is sewed’, while they may also exhibit a verbal distribution in periphrastic occurrences, e.g. ‘Duna has sewed the button’. In such a way, the categorial flexibility of participles allows for a wide spectrum of different uses.
Functional brain correlates of auditory verbal hallucinations in schizophrenia: a design of an fMRI study testing perceptual and cognitive models
(2021) Gawel, Olivia; Fuentes Claramonte, Paola; Rosselló Ximenes, Joana
Auditory verbal hallucinations (AVHs, or ‘hearing voices’) are a cardinal symptom of schizophrenia, and yet, their biological basis has not been fully determined. As of now, theories that attempt to disentangle the origins of AVHs can be separated into two main types (or models): perceptual versus cognitive. The former has considered AVHs to be due to malfunction in perceptual processing, namely an abnormal activation in the auditory cortex, as well as having a top-down cognitive influence. The latter considers AVHs to be due to one of two cognitive processes: the misinterpretation of intrusive memories (which posit that AVHs are the result of a breakdown in the processes monitoring the source of memories) or to the malfunction of inner speech (which posits that AVHs are due to dysfunction of speech monitoring). The current study aims to propose an adequate experimental design of a prospective fMRI study that will test both the perceptual and cognitive approaches allowing to fill the gaps in the general framework for AVHs. Firstly, to test the perceptual model, an experimental design borrowed from Fuentes-Claramonte and colleagues (2021) will be adapted, with a modification controlling for motor activity. To test one side of the Cognitive Model, the theory of intrusive memory, the experimental paradigm that has been created by Fuentes-Claramonte and colleagues (2019) and validated by Martin-Subero and colleagues (2021) will be adapted to test schizophrenic patients with AVHs, which has not been done before. It will elicit negatively valenced autobiographical memories, which has been shown to activate parts of the default mode network, a circuit thought to be impaired in schizophrenia. Thirdly, to test the other side of the Cognitive Model, namely the theory of inner speech, an experimental paradigm will be proposed called the Rhyming task, a phonological encoding task that is known to activate brain regions involved in subvocal rehearsal and short-term storage of information. However, because the stimuli for this task is lacking for the Spanish population, a pilot study (an online survey) was conducted, presenting healthy participants with pairs of objects (created partly from a personalized corpus), and asked them to do three tasks: to decide whether the names of both objects rhyme; provide the name of each object; and rate the object on a 1 to 5 Likert scale for the purposes of determining emotional valence. The results of the pilot study guided the selection of the appropriate stimuli for the prospective imaging study. The proposed fMRI study tackles the biological basis of AVHs from different perspectives, helping to improve patients' lives that are touched by this cardinal symptom, and thus enabling future research to sculpt appropriate clinical intervention thanks to pinpointing the exact biological basis of hearing voices.
Machine-translation inspired reordering as preprocessing for cross-lingual sentiment analysis
(2018-08) Ramírez Atrio, Alejandro; Badia, Antoni
In this thesis we study the effect of word reordering as preprocessing for Cross-Lingual Sentiment Analysis. We try different reorderings in two target languages (Spanish and Catalan) so that their word order more closely resembles the one from our source language (English). Our original expectation was that a Long Short Term Memory classifier trained on English data with bilingual word embeddings would internalize English word order, resulting in poor performance when tested on a target language with different word order. We hypothesized that the more the word order of any of our target languages resembles the one of our source language, the better the overall performance of our sentiment classifier would be when analyzing the target language. We tested five sets of transformation rules for our Part of Speech reorderings of Spanish and Catalan, extracted mainly from two sources: two papers by Crego and Mariño (2006a and 2006b) and our own empirical analysis of two corpora: CoStEP and Tatoeba. The results suggest that the bilingual word embeddings that we are training our Long Short Term Memory model with do not improve any English word order learning by part of the model when used cross-lingually. There is no improvement when reordering the Spanish and Catalan texts so that their word order more closely resembles English, and no significant drop in result score even when applying a random reordering to them making them almost unintelligible, neither when classifying between 2 options (positive-negative) nor between 4 (strongly positive, positive, negative, strongly negative). We also replicated this with two different classifiers: a Convolutional Neural Network and a Support Vector Machine. The Convolutional Neural Network should primarily learn only short-range word order, while the Long Short Term Memory network should be expected to learn as well more long-range orderings. The Support Vector Machine does not take into account word order. Subsequently, we analyzed the prediction biases of these models to see how they affect the reordering results. Based on this analysis, we conclude that the lacking results of the Long Short Term Memory classifier when fed a reordered text do not respond to a problem of prediction bias. In the process of training our models, we use two bilingual lexicons (English-Spanish and English-Catalan) (Hu and Liu 2004) that contain words that typically are key for analyzing the sentiment of a sentence that we use to project our bilingual word embeddings between each language pair. Due to the results we got in the reordering experiments, we conjectured that what determines how our models are classifying the sentiment of the target languages is whether these lexicon words appear or not in the input sentence. Finally, because of this, we test different alterations on the target languages corpora to determine whether this conjecture is strengthened or not. The results seem to go in favor of it. Our main conclusion, therefore, is that Part of Speech-based word reordering of a target language to make its word order more similar to a source language does not improve the results on sentiment classification of our Long Short Term Memory classifier trained on source language data, regardless of the granularity of the sentiment, based on our bilingual word embeddings.
Differential Object Marking in Catalan: Contexts of Appearance and Analysis
(2017-09) Benito Galdeano, Rut; Gallego, Ángel J.; Roca, Francesc
The present paper studies the nature of the phenomenon called Differential Object Marking (DOM hereafter), a device that distinguishes direct objects (DO) by a morphological marker (typically, a preposition). The study focuses on Catalan, paying special attention to the Barcelonese geolect, namely the Catalan variety spoken in the metropolitan area of Barcelona.
Long distance agreement in Spanish
(2017-09) Fernández Serrano, Irene; Gallego, Ángel J.; Hernanz, Maria Lluïsa
From a musical protolanguage to rhythm and tonality
(2014-08) Celma Miralles, Alexandre; Rosselló Ximenes, Joana
[eng] Music and language are two faculties that have only evolved in humans, and by mutual interaction. As Darwin (1871) suggested, before speaking, our ancestors were able to sing in a way structurally and functionally similar to what birds do. At that stage, a musical protolanguage with beat yielded a common basis for music and language. Hierarchical recursion along with grammar and lexical meaning joined this musical protolanguage and gave rise to language. Linguistic recursion, in turn, made meter possible. Rhythm therefore would have preceded tonality. Subsequently, in parallel to the emergence of grammar, harmony and tonality were added to the meter. That beat is more primitive than meter is suggested by the fact that some animals perceive but do not externalize it. Crucially, they are all vocal learners. Externalization, either in musical rhythm or language, requires a complex social behaviour, which for rhythm is already present in the drumming behaviour of certain primates. The role of vocalizations, in turn, goes even further: their harmonic spectrum underpinned the tones of our musical scales. Thus, driven to a large extent by language, music has turned out to be as we know it nowadays.
La distintividad en la sintaxis. El caso de la combinación de clíticos en la lenguas iberorrománicas
(2016-09) Colomina Samitier, M. Pilar; Brucart, José Ma.; Gallego, Ángel J.
En el presente trabajo, nos centraremos en un caso de combinación de elementos idénticos: la combinación de clíticos (ingl. clitic cluster). Los análisis que se ofrecen de estas construcciones presentan enfoques diversos: desde una perspectiva morfológica esta restricción se ha relacionado con la combinación de determinados rasgos. Desde una perspectiva sintáctica, esta incompatibilidad se explica por la imposibilidad de que ambos clíticos ocupen una misma posición. En el presente trabajo consideramos oportuno explorar esta restricción dentro del marco general de la distintividad, puesto que permite relacionar las combinaciones anteriores con otro tipo de combinaciones.

Examinar

Enviaments recents