We propose a method for training language models in an interactive setting inspired by child language acquisition. In our setting, a speaker attempts to communicate some information to a listener in a single-turn dialogue and receives a reward if communicative success is achieved. Unlike earlier related work using image--caption data for interactive reference games, we operationalize communicative success in a more abstract language-only question--answering setting. First, we present a feasibility study demonstrating that our reward provides an indirect signal about grammaticality. Second, we conduct experiments using reinforcement learning to fine-tune language models. We observe that cognitively plausible constraints on the communication channel lead to interpretable changes in speaker behavior. However, we do not yet see improvements on linguistic evaluations from our training regime. We outline potential modifications to the task design and training configuration that could better position future work to use our methodology to observe the benefits of interaction on language learning in computational cognitive models.
Recent advances in self-supervised learning have attracted significant attention from both machine learning and neuroscience. This is primarily because self-supervised methods do not require annotated supervisory information, making them applicable to training artificial networks without relying on large amounts of curated data, and potentially offering insights into how the brain adapts to its environment in an unsupervised manner. Although several previous studies have elucidated the correspondence between neural representations in deep convolutional neural networks (DCNNs) and biological systems, the extent to which unsupervised or self-supervised learning can explain the human-like acquisition of categorically structured information remains less explored. In this study, we investigate the correspondence between the internal representations of DCNNs trained using a self-supervised contrastive learning algorithm and human semantics and recognition. To this end, we employ a few-shot learning evaluation procedure, which measures the ability of DCNNs to recognize novel concepts from limited exposure, to examine the inter-categorical structure of the learned representations. Two comparative approaches are used to relate the few-shot learning outcomes to human semantics and recognition, with results suggesting that the representations acquired through contrastive learning are well aligned with human cognition. These findings underscore the potential of self-supervised contrastive learning frameworks to model learning mechanisms similar to those of the human brain, particularly in scenarios where explicit supervision is unavailable, such as in human infants prior to language acquisition.
Given the massive integration of AI technologies into our daily lives, AI-related concepts are being used to metaphorically compare AI systems with human behaviour and/or cognitive abilities like language acquisition. Rightfully, the epistemic success of these metaphorical comparisons should be debated. Against the backdrop of the conflicting positions of the 'computational' and 'meat' chauvinisms, we ask: can the conceptual constellation of the computational and AI be applied to the human domain and what does it mean to do so? What is one doing when the conceptual constellations of AI in particular are used in this fashion? Rooted in a Wittgensteinian view of concepts and language-use, we consider two possible answers and pit them against each other: either these examples are conceptual metaphors, or they are attempts at conceptual engineering. We argue that they are conceptual metaphors, but that (1) this position is unaware of its own epistemological contingency, and (2) it risks committing the ''map-territory fallacy''. Down at the conceptual foundations of computation, (3) it most importantly is a misleading 'double metaphor' because of the metaphorical connection between human psychology and computation. In response to the shortcomings of this projected conceptual organisation of AI onto the human domain, we argue that there is a semantic catch. The perspective of the conceptual metaphors shows avenues for forms of conceptual engineering. If this methodology's criteria are met, the fallacies and epistemic shortcomings related to the conceptual metaphor view can be bypassed. At its best, the cross-pollution of the human and AI conceptual domains is one that prompts us to reflect anew on how the boundaries of our current concepts serve us and how they could be approved.
This article presents experiments performed using a computational laboratory environment for language acquisition experiments. It implements a multi-agent system consisting of two agents: an adult language model and a daughter language model that aims to learn the mother language. Crucially, the daughter agent does not have access to the internal knowledge of the mother language model but only to the language exemplars the mother agent generates. These experiments illustrate how this system can be used to acquire abstract grammatical knowledge. We demonstrate how statistical analyses of patterns in the input data corresponding to grammatical categories yield discrete grammatical rules. These rules are subsequently added to the grammatical knowledge of the daughter language model. To this end, hierarchical agglomerative cluster analysis was applied to the utterances consecutively generated by the mother language model. It is argued that this procedure can be used to acquire structures resembling grammatical categories proposed by linguists for natural languages. Thus, it is established that non-trivial grammatical knowledge has been acquired. Moreover, the parameter configuration of this computational laboratory environment determined using training data generated by the mother language model is validated in a second experiment with a test set similarly resulting in the acquisition of non-trivial categories.
Predicting the words that a child is going to learn next can be useful for boosting language acquisition, and such predictions have been shown to be possible with both neural network techniques (looking at changes in the vocabulary state over time) and graph model (looking at data pertaining to the relationships between words). However, these models do not fully capture the complexity of the language learning process of an infant when used in isolation. In this paper, we examine how a model of language acquisition for infants and young children can be constructed and adapted for use in a Spatio-Temporal Graph Convolutional Network (STGCN), taking into account the different types of linguistic relationships that occur during child language learning. We introduce a novel approach for predicting child vocabulary acquisition, and evaluate the efficacy of such a model with respect to the different types of linguistic relationships that occur during language acquisition, resulting in insightful observations on model calibration and norm selection. An evaluation of this model found that the mean accuracy of models for predicting new words when using sensorimotor relationships (0.733) and semantic relationships (0.729) were found to be superior to that observed with a 2-layer Feed-forward neural network. Furthermore, the high recall for some relationships suggested that some relationships (e.g. visual) were superior in identifying a larger proportion of relevant words that a child should subsequently learn than others (such as auditory).
Auditory working memory is essential for various daily activities, such as language acquisition, conversation. It involves the temporary storage and manipulation of information that is no longer present in the environment. While extensively studied in neuroscience and cognitive science, research on its modeling within neural networks remains limited. To address this gap, we propose a general framework based on a close-loop predictive coding paradigm to perform short auditory signal memory tasks. The framework is evaluated on two widely used benchmark datasets for environmental sound and speech, demonstrating high semantic similarity across both datasets.
This paper presents BAMBI (BAby language Models Boostrapped for Italian), a series of Baby Language Models (BabyLMs) trained on data that mimic the linguistic input received by a five-year-old Italian-speaking child. The BAMBI models are tested using a benchmark specifically designed to evaluate language models, which takes into account the amount of training input the models received. The BAMBI models are compared against a large language model (LLM) and a multimodal language model (VLM) to study the contribution of extralinguistic information for language acquisition. The results of our evaluation align with the existing literature on English language models, confirming that while reduced training data support the development of relatively robust syntactic competence, they are insufficient for fostering semantic understanding. However, the gap between the training resources (data and computation) of the BAMBI models and the LLMs is not fully reflected in their performance: despite LLMs' massive training, their performance is not much better than that of BAMBI models. This suggests that strategies beyond scaling training resources, such as data curation, inclusion of multimodal input, and other training strategies such as curriculum learning, could play a crucial role in shaping model performance.
In this work, we explain our approach employed in the BabyLM Challenge, which uses various methods of training language models (LMs) with significantly less data compared to traditional large language models (LLMs) and are inspired by how human children learn. While a human child is exposed to far less linguistic input than an LLM, they still achieve remarkable language understanding and generation abilities. To this end, we develop a model trained on a curated dataset consisting of 10 million words, primarily sourced from child-directed transcripts. The 2024 BabyLM Challenge initial dataset of 10M words is filtered to 8.5M. Next, it is supplemented with a randomly selected subset of TVR dataset consisting of 1.5M words of television dialogues. The latter dataset ensures that similar to children, the model is also exposed to language through media. Furthermore, we reduce the vocabulary size to 32,000 tokens, aligning it with the limited vocabulary of children in the early stages of language acquisition. We use curriculum learning and is able to match the baseline on certain benchmarks while surpassing the baseline on others. Additionally, incorporating common LLM training datasets, such as MADLAD-400, degrades performance. These findings underscore the importance of dataset selection, vocabulary scaling, and curriculum learning in creating more data-efficient language models that better mimic human learning processes.
Text simplification is crucial for improving accessibility and comprehension for English as a Second Language (ESL) learners. This study goes a step further and aims to facilitate ESL learners' language acquisition by simplification. Specifically, we propose simplifying complex sentences to appropriate levels for learners while also increasing vocabulary coverage of the target level in the simplifications. We achieve this without a parallel corpus by conducting reinforcement learning on a large language model. Our method employs token-level and sentence-level rewards, and iteratively trains the model on its self-generated outputs to guide the model to search for simplification hypotheses that satisfy the target attributes. Experiment results on CEFR-SP and TurkCorpus datasets show that the proposed method can effectively increase the frequency and diversity of vocabulary of the target level by more than $20\%$ compared to baseline models, while maintaining high simplification quality.
Chatbots based on large language models offer cheap conversation practice opportunities for language learners. However, they are hard to control for linguistic forms that correspond to learners' current needs, such as grammar. We control grammar in chatbot conversation practice by grounding a dialogue response generation model in a pedagogical repository of grammar skills. We also explore how this control helps learners to produce specific grammar. We comprehensively evaluate prompting, fine-tuning, and decoding strategies for grammar-controlled dialogue response generation. Strategically decoding Llama3 outperforms GPT-3.5 when tolerating minor response quality losses. Our simulation predicts grammar-controlled responses to support grammar acquisition adapted to learner proficiency. Existing language learning chatbots and research on second language acquisition benefit from these affordances. Code available on GitHub.
Large language models exhibit general linguistic abilities but significantly differ from humans in their efficiency of language acquisition. This study proposes a method for integrating the developmental characteristics of working memory during the critical period, a stage when human language acquisition is particularly efficient, into language models. The proposed method introduces a mechanism that initially constrains working memory during the early stages of training and gradually relaxes this constraint in an exponential manner as learning progresses. Targeted syntactic evaluation shows that the proposed method outperforms conventional models without memory constraints or with static memory constraints. These findings not only provide new directions for designing data-efficient language models but also offer indirect evidence supporting the underlying mechanisms of the critical period hypothesis in human language acquisition.
This study underscores the pivotal role of syntax feedback in augmenting the syntactic proficiency of students. Recognizing the challenges faced by learners in mastering syntactic nuances, we introduce a specialized dataset named Essay-Syntax-Instruct designed to enhance the understanding and application of English syntax among these students. Leveraging the capabilities of Large Language Models (LLMs) such as GPT3.5-Turbo, Llama-2-7b-chat-hf, Llama-2-13b-chat-hf, and Mistral-7B-Instruct-v0.2, this work embarks on a comprehensive fine-tuning process tailored to the syntax improvement task. Through meticulous evaluation, we demonstrate that the fine-tuned LLMs exhibit a marked improvement in addressing syntax-related challenges, thereby serving as a potent tool for students to identify and rectify their syntactic errors. The findings not only highlight the effectiveness of the proposed dataset in elevating the performance of LLMs for syntax enhancement but also illuminate a promising path for utilizing advanced language models to support language acquisition efforts. This research contributes to the broader field of language learning technology by showcasing the potential of LLMs in facilitating the linguistic development of Students.
Recursion is the nature of human natural language. Since Chomsky proposed generative grammar, many scholars have studied recursion either theoretically or empirically. However, by observing children's acquisition of tail recursion sequences, we can verify the nativism of language supported by universal grammar and reveal the cognitive mechanism of human brain. To date, our understanding of children's acquisition path of recursion and influencing factors still remain controversial. This systematic review summarizes the research of tail recursive sequence by taking possessive recursion and locative recursion as examples, focusing on the experimental methods, acquisition paths, and influencing factors of tail recursive sequence. The current behavioural experiments reveal that, the debate about children's performance revolves around: 1) Gradual acquisition or synchronous acquisition. 2) symmetry or asymmetry between the acquisition of locative recursion sequences and possessive recursion sequences. We presume that children can acquire recursion quickly in a short period of time thanks to the language acquisition device, though there are also scholars who believe that a third factor also plays a role.
In second-language acquisition, predictive modeling aids educators in implementing diverse teaching strategies, attracting significant research attention. However, while model accuracy is widely explored, model fairness remains under-examined. Model fairness ensures equitable treatment of groups, preventing unintentional biases based on attributes such as gender, ethnicity, or economic background. A fair model should produce impartial outcomes that do not systematically disadvantage any group. This study evaluates the fairness of two predictive models using the Duolingo dataset's en\_es (English learners speaking Spanish), es\_en (Spanish learners speaking English), and fr\_en (French learners speaking English) tracks. We analyze: 1. Algorithmic fairness across platforms (iOS, Android, Web). 2. Algorithmic fairness between developed and developing countries. Key findings include: 1. Deep learning outperforms machine learning in second-language knowledge tracing due to improved accuracy and fairness. 2. Both models favor mobile users over non-mobile users. 3. Machine learning exhibits stronger bias against developing countries compared to deep learning. 4. Deep learning strikes a better balance of fairness and accuracy in the en\_es and es\_en tracks, while machine learning is more suitable for fr\_en. This study highlights the importance of addressing fairness in predictive models to ensure equitable educational strategies across platforms and regions.
As recursion has been underlying any linguistic work for the last 60 years, the acquisition of recursive structures by children during language learning has become a focal point of inquiry. This study delves into the developmental trajectory of Mandarin-speaking children's acquisition of recursive possessives and locatives, assessing the impact of structural diversity on language acquisition. The research contrasts the comprehension of two-level recursive structures among children aged 3 to 7 years, employing answering question while seeing a picture task to elicit responses. The findings indicate that children do not attain adult-like proficiency in two-level recursion until the age of 6, and there exists a notable asymmetry in the acquisition of recursive possessives versus locatives. These results underscore the primacy of structural complexity and cognitive factors in the acquisition process, enhancing our comprehension of the cognitive foundations of language development and the pivotal role of recursion in child language acquisition.
Understanding how infants perceive speech sounds and language structures is still an open problem. Previous research in artificial neural networks has mainly focused on large dataset-dependent generative models, aiming to replicate language-related phenomena such as ''perceptual narrowing''. In this paper, we propose a novel approach using a small-sized generative neural network equipped with a continual learning mechanism based on predictive coding for mono-and bilingual speech sound learning (referred to as language sound acquisition during ''critical period'') and a compositional optimization mechanism for generation where no learning is involved (later infancy sound imitation). Our model prioritizes interpretability and demonstrates the advantages of online learning: Unlike deep networks requiring substantial offline training, our model continuously updates with new data, making it adaptable and responsive to changing inputs. Through experiments, we demonstrate that if second language acquisition occurs during later infancy, the challenges associated with learning a foreign language after the critical period amplify, replicating the perceptual narrowing effect.
Human languages have evolved to be structured through repeated language learning and use. These processes introduce biases that operate during language acquisition and shape linguistic systems toward communicative efficiency. In this paper, we investigate whether the same happens if artificial languages are optimised for implicit biases of Large Language Models (LLMs). To this end, we simulate a classical referential game in which LLMs learn and use artificial languages. Our results show that initially unstructured holistic languages are indeed shaped to have some structural properties that allow two LLM agents to communicate successfully. Similar to observations in human experiments, generational transmission increases the learnability of languages, but can at the same time result in non-humanlike degenerate vocabularies. Taken together, this work extends experimental findings, shows that LLMs can be used as tools in simulations of language evolution, and opens possibilities for future human-machine experiments in this field.
In many fields, such as language acquisition, neuropsychology of language, the study of aging, and historical linguistics, corpora are used for estimating the diversity of grammatical structures that are produced during a period by an individual, community, or type of speakers. In these cases, treebanks are taken as representative samples of the syntactic structures that might be encountered. Generalizing the potential syntactic diversity from the structures documented in a small corpus requires careful extrapolation whose accuracy is constrained by the limited size of representative sub-corpora. In this article, I demonstrate -- theoretically, and empirically -- that a grammar's derivational entropy and the mean length of the utterances (MLU) it generates are fundamentally linked, giving rise to a new measure, the derivational entropy rate. The mean length of utterances becomes the most practical index of syntactic complexity; I demonstrate that MLU is not a mere proxy, but a fundamental measure of syntactic diversity. In combination with the new derivational entropy rate measure, it provides a theory-free assessment of grammatical complexity. The derivational entropy rate indexes the rate at which different grammatical annotation frameworks determine the grammatical complexity of treebanks. I introduce the Smoothed Induced Treebank Entropy (SITE) as a tool for estimating these measures accurately, even from very small treebanks. I conclude by discussing important implications of these results for both NLP and human language processing.
Whether and how language models (LMs) acquire the syntax of natural languages has been widely evaluated under the minimal pair paradigm. However, a lack of wide-coverage benchmarks in languages other than English has constrained systematic investigations into the issue. Addressing it, we first introduce ZhoBLiMP, the most comprehensive benchmark of linguistic minimal pairs for Chinese to date, with 118 paradigms, covering 15 linguistic phenomena. We then train 20 LMs of different sizes (14M to 1.4B) on Chinese corpora of various volumes (100M to 3B tokens) and evaluate them along with 14 off-the-shelf LLMs on ZhoBLiMP. The overall results indicate that Chinese grammar can be mostly learned by models with around 500M parameters, trained on 1B tokens with one epoch, showing limited benefits for further scaling. Most (N=95) linguistic paradigms are of easy or medium difficulty for LMs, while there are still 13 paradigms that remain challenging even for models with up to 32B parameters. In regard to how LMs acquire Chinese grammar, we observe a U-shaped learning pattern in several phenomena, similar to those observed in child language acquisition.
Large language models (LLMs) now exhibit near human-level performance in various tasks, but their performance drops drastically after a handful of high-resource languages due to the imbalance in pre-training data. Inspired by the human process of second language acquisition, particularly code-switching (the practice of language alternation in a conversation), we propose code-switching curriculum learning (CSCL) to enhance cross-lingual transfer for LLMs. CSCL mimics the stages of human language learning by progressively training models with a curriculum consisting of 1) token-level code-switching, 2) sentence-level code-switching, and 3) monolingual corpora. Using Qwen 2 as our underlying model, we demonstrate the efficacy of the CSCL in improving language transfer to Korean, achieving significant performance gains compared to monolingual continual pre-training methods. Ablation studies reveal that both token- and sentence-level code-switching significantly enhance cross-lingual transfer and that curriculum learning amplifies these effects. We also extend our findings into various languages, including Japanese (high-resource) and Indonesian (low-resource), and using two additional models (Gemma 2 and Phi 3.5). We further show that CSCL mitigates spurious correlations between language resources and safety alignment, presenting a robust, efficient framework for more equitable language transfer in LLMs. We observe that CSCL is effective for low-resource settings where high-quality, monolingual corpora for language transfer are hardly available.