El artículo examina la cuestión de la anámnesis en la filosofía platónica, considerada como la caracterización de todo aprender como “recordar” en el Menón, el Fedro y el Fedón. Para tratar esta problemática, se presentan las presuposiciones necesarias de cualquier lectura de la anámnesis, notablemente la naturaleza de las formas contempladas y cómo se relacionan con nuestra comprensión. Con este fin, se recurre a las interpretaciones de Eric D. Perl y Hans-Georg Gadamer para “situar” la anámnesis dentro de una tradición interpretativa hermenéutico-filosófica de Platón. Posteriormente, se interpreta la anámnesis como una descripción del aprendizaje, y el “medio” que vincula nuestra comprensión y el ser revelado inteligiblemente —punto no desarrollado
por Perl o Gadamer—. La anámnesis se revela como “recordar lo eterno y siempre presente”, pasando de lo conceptual y lingüísticamente indeterminado a lo determinado. Finalmente, se sugieren vínculos entre la interpretación propuesta, la filosofía de Platón, Aristóteles y Gadamer.
The present study portrays some of the key aspects of connected speech in English, as adopted by 42 native Bosnian/Croatian/Serbian-speaking undergraduate students of English in the English Department, University of Tuzla, in the academic year 2013/2014. More specifically, the study shows how successfully these students developed their transcription skills in English, particularly when it comes to the use of diacritics for dental, velarised, and syllabic consonants of English, as well as for aspirated and unreleased (unexploded) English plosives. In addition, the study focuses on the coalescent type of assimilation. Connected speech (also known as rapid, relaxed, casual, or fluent speech) is characterised by a number of phonetic phenomena. The paper also analyses the level to which students enrolled in the English Department in Tuzla have developed a sense of elementary terms in this field, an understanding of the English sound system, and generally speaking, to what extent they developed their broad and narrow transcription skills.
Vasile Păiş, Sara Niţă, Alexandru-Iulius Jerpelea
et al.
Memes are becoming increasingly more popular in online media, especially in social networks. They usually combine graphical representations (images, drawings, animations or video) with text to convey powerful messages. In order to extract, process and understand the messages, AI applications need to employ multimodal algorithms. In this paper, we introduce a curated dataset of real memes in the Romanian language, with multiple annotation levels. Baseline algorithms were employed to demonstrate the usability of the dataset. Results indicate that further research is needed to improve the processing capabilities of AI tools when faced with Internet memes.
Jannik Peters, Constantin Waubert de Puiseau, Hasan Tercan
et al.
The field of emergent language represents a novel area of research within the domain of artificial intelligence, particularly within the context of multi-agent reinforcement learning. Although the concept of studying language emergence is not new, early approaches were primarily concerned with explaining human language formation, with little consideration given to its potential utility for artificial agents. In contrast, studies based on reinforcement learning aim to develop communicative capabilities in agents that are comparable to or even superior to human language. Thus, they extend beyond the learned statistical representations that are common in natural language processing research. This gives rise to a number of fundamental questions, from the prerequisites for language emergence to the criteria for measuring its success. This paper addresses these questions by providing a comprehensive review of 181 scientific publications on emergent language in artificial intelligence. Its objective is to serve as a reference for researchers interested in or proficient in the field. Consequently, the main contributions are the definition and overview of the prevailing terminology, the analysis of existing evaluation methods and metrics, and the description of the identified research gaps.
In recent years, the breakthrough of Large Language Models (LLMs) offers new ideas for achieving universal methods on graph data. The common practice of converting graphs into natural language for LLMs, which refers to graph flattening, exhibits good generalizability and interpretability. However, the poor organization of the textual format results in poor performance in long-distance scenario understanding. Inspired by human cognitive reasoning habits, we propose a novel method for graph flattening to fit LLMs, termed as End-to-End DAG-Path prompting (EEDP). Experiments on real-world datasets show that EEDP enhances the reasoning performance of LLMs in long-distance scenarios while maintaining excellent performance in short-distance scenarios, demonstrating good robustness in the face of distance variations.
I argue that third person is not underspecified: there must be a distinct third person feature. I add to the existing body of morphological arguments for this conclusion (Nevins 2007; Trommer 2008, a.o.) a syntactic argument: I show that there is omnivorous third person agreement in Algonquian languages. I focus here on two, Blackfoot (Plains Algonquian) and Plains Cree (Central Algonquian), demonstrating that they have an agreement suffix (the peripheral suffix, analyzed as a probe in C) that indexes the number, animacy, and obviation of the structurally-highest third person argument, skipping over first and second person if it has to. I argue that alternative analyses of this agreement pattern in terms of animacy, obviation, and the categorial feature [D] do not work; thus, third person must be specified even in the syntax (contra Preminger 2019).
Ilker Kesen, Andrea Pedrotti, Mustafa Dogan
et al.
With the ever-increasing popularity of pretrained Video-Language Models (VidLMs), there is a pressing need to develop robust evaluation methodologies that delve deeper into their visio-linguistic capabilities. To address this challenge, we present ViLMA (Video Language Model Assessment), a task-agnostic benchmark that places the assessment of fine-grained capabilities of these models on a firm footing. Task-based evaluations, while valuable, fail to capture the complexities and specific temporal aspects of moving images that VidLMs need to process. Through carefully curated counterfactuals, ViLMA offers a controlled evaluation suite that sheds light on the true potential of these models, as well as their performance gaps compared to human-level understanding. ViLMA also includes proficiency tests, which assess basic capabilities deemed essential to solving the main counterfactual tests. We show that current VidLMs' grounding abilities are no better than those of vision-language models which use static images. This is especially striking once the performance on proficiency tests is factored in. Our benchmark serves as a catalyst for future research on VidLMs, helping to highlight areas that still need to be explored.
Tree-controlled grammars are context-free grammars where the derivation process is controlled in such a way that every word on a level of the derivation tree must belong to a certain control language. We investigate the generative capacity of such tree-controlled grammars where the control languages are special regular sets, especially strictly locally testable languages or languages restricted by resources of the generation (number of non-terminal symbols or production rules) or acceptance (number of states). Furthermore, the set theoretic inclusion relations of these subregular language families themselves are studied.
Bram M. A. van Dijk, Tom Kouwenhoven, Marco R. Spruit
et al.
Current Large Language Models (LLMs) are unparalleled in their ability to generate grammatically correct, fluent text. LLMs are appearing rapidly, and debates on LLM capacities have taken off, but reflection is lagging behind. Thus, in this position paper, we first zoom in on the debate and critically assess three points recurring in critiques of LLM capacities: i) that LLMs only parrot statistical patterns in the training data; ii) that LLMs master formal but not functional language competence; and iii) that language learning in LLMs cannot inform human language learning. Drawing on empirical and theoretical arguments, we show that these points need more nuance. Second, we outline a pragmatic perspective on the issue of `real' understanding and intentionality in LLMs. Understanding and intentionality pertain to unobservable mental states we attribute to other humans because they have pragmatic value: they allow us to abstract away from complex underlying mechanics and predict behaviour effectively. We reflect on the circumstances under which it would make sense for humans to similarly attribute mental states to LLMs, thereby outlining a pragmatic philosophical context for LLMs as an increasingly prominent technology in society.
Mai Ha Vu, Rahmad Akbar, Philippe A. Robert
et al.
Deep neural-network-based language models (LMs) are increasingly applied to large-scale protein sequence data to predict protein function. However, being largely black-box models and thus challenging to interpret, current protein LM approaches do not contribute to a fundamental understanding of sequence-function mappings, hindering rule-based biotherapeutic drug development. We argue that guidance drawn from linguistics, a field specialized in analytical rule extraction from natural language data, can aid with building more interpretable protein LMs that are more likely to learn relevant domain-specific rules. Differences between protein sequence data and linguistic sequence data require the integration of more domain-specific knowledge in protein LMs compared to natural language LMs. Here, we provide a linguistics-based roadmap for protein LM pipeline choices with regard to training data, tokenization, token embedding, sequence embedding, and model interpretation. Incorporating linguistic ideas into protein LMs enables the development of next-generation interpretable machine-learning models with the potential of uncovering the biological mechanisms underlying sequence-function relationships.
We introduce GPT-NeoX-20B, a 20 billion parameter autoregressive language model trained on the Pile, whose weights will be made freely and openly available to the public through a permissive license. It is, to the best of our knowledge, the largest dense autoregressive model that has publicly available weights at the time of submission. In this work, we describe \model{}'s architecture and training and evaluate its performance on a range of language-understanding, mathematics, and knowledge-based tasks. We find that GPT-NeoX-20B is a particularly powerful few-shot reasoner and gains far more in performance when evaluated five-shot than similarly sized GPT-3 and FairSeq models. We open-source the training and evaluation code, as well as the model weights, at https://github.com/EleutherAI/gpt-neox.
The short story could be the media to teach something to the readers. “Two old cherry trees still in bloom” for example, is written by Saikaku. Through his work, Saikaku would like to show the readers the implementation of 4 identity-statuses. He brought the readers into critical reading to help them find the identity. To analyze the 4 identity status content in his short story, the coding analysis was used. By analyzing, it could be seen that 4 identity-status categories comprise identity diffusion, identity foreclosure, identity moratorium, and identity achievement. Each status has two codes like uncertainty and unknown in identity diffusion, admiration and sympathy in identity foreclosure, struggle and chasing in identity moratorium, and settle and completion in identity achievement. Through the analysis, Saikaku shows the readers that to harmonize oneself as an individual and community member, someone needs to develop his curiosity through implementing identity moratorium to satisfy his lacking of background knowledge. Moreover, developing a sense of love is very essential to keep someone digging something out sufficiently through identity foreclosure to reach the ultimate goal of life.
Language. Linguistic theory. Comparative grammar, Computational linguistics. Natural language processing
Decades of research has studied how language learning infants learn to discriminate speech sounds, segment words, and associate words with their meanings. While gradual development of such capabilities is unquestionable, the exact nature of these skills and the underlying mental representations yet remains unclear. In parallel, computational studies have shown that basic comprehension of speech can be achieved by statistical learning between speech and concurrent referentially ambiguous visual input. These models can operate without prior linguistic knowledge such as representations of linguistic units, and without learning mechanisms specifically targeted at such units. This has raised the question of to what extent knowledge of linguistic units, such as phone(me)s, syllables, and words, could actually emerge as latent representations supporting the translation between speech and representations in other modalities, and without the units being proximal learning targets for the learner. In this study, we formulate this idea as the so-called latent language hypothesis (LLH), connecting linguistic representation learning to general predictive processing within and across sensory modalities. We review the extent that the audiovisual aspect of LLH is supported by the existing computational studies. We then explore LLH further in extensive learning simulations with different neural network models for audiovisual cross-situational learning, and comparing learning from both synthetic and real speech data. We investigate whether the latent representations learned by the networks reflect phonetic, syllabic, or lexical structure of input speech by utilizing an array of complementary evaluation metrics related to linguistic selectivity and temporal characteristics of the representations. As a result, we find that representations associated...
As a vast and diverse linguistic grouping, Tibeto-Burman languages vary in their usage of time constructs, both morphologically and semantically. Even between genetically related languages within the Tibeto-Burman language family, approaches to elements such as suffixation vary widely, while vocabulary from Indo-Aryan and distantly related Sinitic languages is differently incorporated and borrowed. In this article, we identify trends that only become apparent through the process of data collation and the careful comparison of numerous grammatical sketches and dictionaries. We further expand this rich, if understudied, area through the incorporation of original fieldwork data from the Thangmi/Thami-speaking communities of Nepal undertaken by one of the co-authors, and supplemented by the researcher’s residence in the Himalayan region from 1996 to 2009. The literature review and linguistic scope of this survey includes multiple grammars of languages spoken across the Greater Himalayan region, with specific emphasis on the Rāī-Kiranti sub-branch of languages autochthonous to eastern Nepal. In our comparative analysis, we focus on apparent cognates and shared paradigms with an emphasis on systems of segmental time measurement (e.g. ‘two days hence,’ ‘this year’) rather than on relative ones (e.g. ‘now,’ ‘then’). Through this compilation, the relationship between Tibeto-Burman languages and their often-dominant regional Indo-Aryan counterparts becomes more visible, mediated by a better understanding of the shared yet conflicting epistemological, astrological, and organizational views of time held by the communities who speak Tibeto-Burman languages. Features of note include the assimilation of Chinese and Indian religious and spiritual systems, as well as imported vocabulary that does not always replace—but is in fact sometimes incorporated into—the lexicon of a given language by the speech community. It is our observation that in Tibeto-Burman languages, Indigenous concepts, categories and classifications of time are usually grammatically encoded in adverbial forms, while the influential Indo-Aryan languages of the region mostly make use of nominal morphology in order to express temporal concepts. In addition, reflexes of Proto-Tibeto-Burman (hereafter PTB) nouns are still evident across the language family. To conclude, we position this survey as a comparative and analytical contribution which focuses attention on the region’s rich linguistic variation and the importance of rigorous documentation, conservation and revitalisation programs for Indigenous languages of the Tibeto-Burman family, as the communities who speak these languages continue to grapple with severe socio-political challenges and face the hegemonic pressures of linguistic assimilation.
L’articolo analizza le prime nove opere uscite nella collana di poesia Croma k per l’editore Oèdipus. La finalità è quella di ricollegare le caratteristiche formali di queste opere a una possibile tendenza della poesia contemporanea italiana.
Language. Linguistic theory. Comparative grammar, Style. Composition. Rhetoric
In this paper, we develop a mathematical model of awareness based on the idea of plurality. Instead of positing a singular principle, telos, or essence as noumenon, we model it as plurality accessible through multiple forms of awareness ("n-awareness"). In contrast to many other approaches, our model is committed to pluralist thinking. The noumenon is plural, and reality is neither reducible nor irreducible. Nothing dies out in meaning making. We begin by mathematizing the concept of awareness by appealing to the mathematical formalism of higher category theory. The beauty of higher category theory lies in its universality. Pluralism is categorical. In particular, we model awareness using the theories of derived categories and $(\infty, 1)$-topoi which will give rise to our meta-language. We then posit a "grammar" ("n-declension") which could express n-awareness, accompanied by a new temporal ontology ("n-time"). Our framework allows us to revisit old problems in the philosophy of time: how is change possible and what do we mean by simultaneity and coincidence? Another question which could be re-conceptualized in our model is one of soteriology related to this pluralism: what is a self in this context? A new model of "personal identity over time" is thus introduced.
Natural language generation provides designers with methods for automatically generating text, e.g. for creating summaries, chatbots and game content. In practise, text generators are often either learned and hard to interpret, or created by hand using techniques such as grammars and templates. In this paper, we introduce a novel grammar induction algorithm for learning interpretable grammars for generative purposes, called Gitta. We also introduce the novel notion of template trees to discover latent templates in corpora to derive these generative grammars. By using existing human-created grammars, we found that the algorithm can reasonably approximate these grammars using only a few examples. These results indicate that Gitta could be used to automatically learn interpretable and easily modifiable grammars, and thus provide a stepping stone for human-machine co-creation of generative models.
Hiroaki Naganuma, Diptarama Hendrian, Ryo Yoshinaka
et al.
We propose a new approach for universal lossless text compression, based on grammar compression. In the literature, a target string $T$ has been compressed as a context-free grammar $G$ in Chomsky normal form satisfying $L(G) = \{T\}$. Such a grammar is often called a \emph{straight-line program} (SLP). In this paper, we consider a probabilistic grammar $G$ that generates $T$, but not necessarily as a unique element of $L(G)$. In order to recover the original text $T$ unambiguously, we keep both the grammar $G$ and the derivation tree of $T$ from the start symbol in $G$, in compressed form. We show some simple evidence that our proposal is indeed more efficient than SLPs for certain texts, both from theoretical and practical points of view.