Bayesian validation of grammar productions for the language of thought.
Ontology highlight
ABSTRACT:
SUBMITTER: Romano S
PROVIDER: S-EPMC6039029 | biostudies-literature | 2018
REPOSITORIES: biostudies-literature
Publications
Bayesian validation of grammar productions for the language of thought.
Romano Sergio S Salles Alejo A Amalric Marie M Dehaene Stanislas S Sigman Mariano M Figueira Santiago S
PloS one 20180710 7
Probabilistic proposals of Language of Thoughts (LoTs) can explain learning across different domains as statistical inference over a compositionally structured hypothesis space. While frameworks may differ on how a LoT may be implemented computationally, they all share the property that they are built from a set of atomic symbols and rules by which these symbols can be combined. In this work we propose an extra validation step for the set of atomic productions defined by the experimenter. It sta ...[more]
Similar Datasets
Project description:Formal thought disorder (FTD) is clinically manifested as disorganized speech, but there have been only few investigations of its linguistic properties. We examined how disturbance of thought may relate to the referential function of language as expressed in the use of noun phrases (NPs) and the complexity of sentence structures. We used a comic strip description task to elicit language samples from 30 participants with schizophrenia (SZ), 15 with moderate or severe FTD (SZ + FTD), and 15 minimal or no FTD (SZ-FTD), as well as 15 first-degree relatives of people with SZ (FDRs) and 15 neurotypical controls (NC). We predicted that anomalies in the normal referential use of NPs, sub-divided into definite and indefinite NPs, would identify FTD; and also that FTD would also be linked to reduced linguistic complexity as specifically measured by the number of embedded clauses and of grammatical dependents. Participants with SZ + FTD produced more referential anomalies than NC and produced the fewest definite NPs, while FDRs produced the most and thus also differed from NC. When referential anomalies were classed according to the NP type in which they occurred, the SZ + FTD group produced more anomalies in definite NPs than NC. Syntactic errors did not distinguish groups, but the SZ + FTD group exhibited significantly less syntactic complexity than non-SZ groups. Exploratory regression analyses suggested that production of definite NPs distinguished the two SZ groups. These results demonstrate that FTD can be identified in specific grammatical patterns which provide new targets for detection, intervention, and neurobiological studies.
| S-EPMC6145886 | biostudies-literature
Project description:Languages with many speakers tend to be structurally simple while small communities sometimes develop languages with great structural complexity. Paradoxically, the opposite pattern appears to be observed for non-structural properties of language such as vocabulary size. These apparently opposite patterns pose a challenge for theories of language change and evolution. We use computational simulations to show that this inverse pattern can depend on a single factor: ease of diffusion through the population. A population of interacting agents was arranged on a network, passing linguistic conventions to one another along network links. Agents can invent new conventions, or replicate conventions that they have previously generated themselves or learned from other agents. Linguistic conventions are either Easy or Hard to diffuse, depending on how many times an agent needs to encounter a convention to learn it. In large groups, only linguistic conventions that are easy to learn, such as words, tend to proliferate, whereas small groups where everyone talks to everyone else allow for more complex conventions, like grammatical regularities, to be maintained. Our simulations thus suggest that language, and possibly other aspects of culture, may become simpler at the structural level as our world becomes increasingly interconnected.
| S-EPMC5805949 | biostudies-literature
Project description:IntroductionBrain surgery in the language dominant hemisphere remains challenging due to unintended post-surgical language deficits, despite using pre-surgical functional magnetic resonance (fMRI) and intraoperative cortical stimulation. Moreover, patients are often recommended not to undergo surgery if the accompanying risk to language appears to be too high. While standard fMRI language mapping protocols may have relatively good predictive value at the group level, they remain sub-optimal on an individual level. The standard tests used typically assess lexico-semantic aspects of language, and they do not accurately reflect the complexity of language either in comprehension or production at the sentence level. Among patients who had left hemisphere language dominance we assessed which tests are best at activating language areas in the brain.MethodWe compared grammar tests (items testing word order in actives and passives, wh-subject and object questions, relativized subject and object clauses and past tense marking) with standard tests (object naming, auditory and visual responsive naming), using pre-operative fMRI. Twenty-five surgical candidates (13 females) participated in this study. Sixteen patients presented with a brain tumor, and nine with epilepsy. All participants underwent two pre-operative fMRI protocols: one including CYCLE-N grammar tests (items testing word order in actives and passives, wh-subject and object questions, relativized subject and object clauses and past tense marking); and a second one with standard fMRI tests (object naming, auditory and visual responsive naming). fMRI activations during performance in both protocols were compared at the group level, as well as in individual candidates.ResultsThe grammar tests generated more volume of activation in the left hemisphere (left/right angular gyrus, right anterior/posterior superior temporal gyrus) and identified additional language regions not shown by the standard tests (e.g., left anterior/posterior supramarginal gyrus). The standard tests produced more activation in left BA 47. Ten participants had more robust activations in the left hemisphere in the grammar tests and two in the standard tests. The grammar tests also elicited substantial activations in the right hemisphere and thus turned out to be superior at identifying both right and left hemisphere contribution to language processing.ConclusionThe grammar tests may be an important addition to the standard pre-operative fMRI testing.
| S-EPMC5458087 | biostudies-literature
Project description:BackgroundScientific and public fascination with human language have included intensive scrutiny of language disorders as a new window onto the biological foundations of language and its evolutionary origins. Specific language impairment (SLI), which affects over 7% of children, is one such disorder. SLI has received robust scientific attention, in part because of its recent linkage to a specific gene and loci on chromosomes and in part because of the prevailing question regarding the scope of its language impairment: Does the disorder impact the general ability to segment and process language or a specific ability to compute grammar? Here we provide novel electrophysiological data showing a domain-specific deficit within the grammar of language that has been hitherto undetectable through behavioural data alone.Methods and findingsWe presented participants with Grammatical(G)-SLI, age-matched controls, and younger child and adult controls, with questions containing syntactic violations and sentences containing semantic violations. Electrophysiological brain responses revealed a selective impairment to only neural circuitry that is specific to grammatical processing in G-SLI. Furthermore, the participants with G-SLI appeared to be partially compensating for their syntactic deficit by using neural circuitry associated with semantic processing and all non-grammar-specific and low-level auditory neural responses were normal.ConclusionsThe findings indicate that grammatical neural circuitry underlying language is a developmentally unique system in the functional architecture of the brain, and this complex higher cognitive system can be selectively impaired. The findings advance fundamental understanding about how cognitive systems develop and all human language is represented and processed in the brain.
| S-EPMC2268250 | biostudies-literature
Project description:We test whether large language models (LLMs) can be used to simulate human participants in social-science studies. To do this, we ran replications of 14 studies from the Many Labs 2 replication project with OpenAI's text-davinci-003 model, colloquially known as GPT-3.5. Based on our pre-registered analyses, we find that among the eight studies we could analyse, our GPT sample replicated 37.5% of the original results and 37.5% of the Many Labs 2 results. However, we were unable to analyse the remaining six studies due to an unexpected phenomenon we call the "correct answer" effect. Different runs of GPT-3.5 answered nuanced questions probing political orientation, economic preference, judgement, and moral philosophy with zero or near-zero variation in responses: with the supposedly "correct answer." In one exploratory follow-up study, we found that a "correct answer" was robust to changing the demographic details that precede the prompt. In another, we found that most but not all "correct answers" were robust to changing the order of answer choices. One of our most striking findings occurred in our replication of the Moral Foundations Theory survey results, where we found GPT-3.5 identifying as a political conservative in 99.6% of the cases, and as a liberal in 99.3% of the cases in the reverse-order condition. However, both self-reported 'GPT conservatives' and 'GPT liberals' showed right-leaning moral foundations. Our results cast doubts on the validity of using LLMs as a general replacement for human participants in the social sciences. Our results also raise concerns that a hypothetical AI-led future may be subject to a diminished diversity of thought.
| S-EPMC11335848 | biostudies-literature
Project description:All natural languages have formal devices for communicating about number, be they lexical (e.g., two, many) or grammatical (e.g., plural markings on nouns and/or verbs). Here we ask whether linguistic devices for number arise in communication systems that have not been handed down from generation to generation. We examined deaf individuals who had not been exposed to a usable model of conventional language (signed or spoken), but had nevertheless developed their own gestures, called homesigns, to communicate. Study 1 examined four adult homesigners and a hearing communication partner for each homesigner. The adult homesigners produced two main types of number gestures: gestures that enumerated sets (cardinal number marking), and gestures that signaled one vs. more than one (non-cardinal number marking). Both types of gestures resembled, in form and function, number signs in established sign languages and, as such, were fully integrated into each homesigner's gesture system and, in this sense, linguistic. The number gestures produced by the homesigners' hearing communication partners displayed some, but not all, of the homesigners' linguistic patterns. To better understand the origins of the patterns displayed by the adult homesigners, Study 2 examined a child homesigner and his hearing mother, and found that the child's number gestures displayed all of the properties found in the adult homesigners' gestures, but his mother's gestures did not. The findings suggest that number gestures and their linguistic use can appear relatively early in homesign development, and that hearing communication partners are not likely to be the source of homesigners' linguistic expressions of non-cardinal number. Linguistic devices for number thus appear to be so fundamental to language that they can arise in the absence of conventional linguistic input.
| S-EPMC3870334 | biostudies-literature
Project description:The cognitive approach to psychotherapy aims to change patients' maladaptive schemas, that is, overly negative views on themselves, the world, or the future. To obtain awareness of these views, they record their thought processes in situations that caused pathogenic emotional responses. The schemas underlying such thought records have, thus far, been largely manually identified. Using recent advances in natural language processing, we take this one step further by automatically extracting schemas from thought records. To this end, we asked 320 healthy participants on Amazon Mechanical Turk to each complete five thought records consisting of several utterances reflecting cognitive processes. Agreement between two raters on manually scoring the utterances with respect to how much they reflect each schema was substantial (Cohen's κ = 0.79). Natural language processing software pretrained on all English Wikipedia articles from 2014 (GLoVE embeddings) was used to represent words and utterances, which were then mapped to schemas using k-nearest neighbors algorithms, support vector machines, and recurrent neural networks. For the more frequently occurring schemas, all algorithms were able to leverage linguistic patterns. For example, the scores assigned to the Competence schema by the algorithms correlated with the manually assigned scores with Spearman correlations ranging between 0.64 and 0.76. For six of the nine schemas, a set of recurrent neural networks trained separately for each of the schemas outperformed the other algorithms. We present our results here as a benchmark solution, since we conducted this research to explore the possibility of automatically processing qualitative mental health data and did not aim to achieve optimal performance with any of the explored models. The dataset of 1600 thought records comprising 5747 utterances is published together with this article for researchers and machine learning enthusiasts to improve upon our outcomes. Based on our promising results, we see further opportunities for using free-text input and subsequent natural language processing in other common therapeutic tools, such as ecological momentary assessments, automated case conceptualizations, and, more generally, as an alternative to mental health scales.
| S-EPMC8523074 | biostudies-literature
Project description:Chain-of-thought prompting enhances the abilities of large language models (LLMs) significantly. It not only makes these models more specific and context-aware but also impacts the wider field of artificial intelligence (AI). This approach broadens the usability of AI, increases its efficiency, and aligns it more closely with human thinking and decision-making processes. As we improve this method, it is set to become a key element in the future of AI, adding more purpose, precision, and ethical consideration to these technologies. In medicine, the chain-of-thought prompting is especially beneficial. Its capacity to handle complex information, its logical and sequential reasoning, and its suitability for ethically and context-sensitive situations make it an invaluable tool for healthcare professionals. Its role in enhancing medical care and research is expected to grow as we further develop and use this technique. Chain-of-thought prompting bridges the gap between AI's traditionally obscure decision-making process and the clear, accountable standards required in healthcare. It does this by emulating a reasoning style familiar to medical professionals, fitting well into their existing practices and ethical codes. While solving AI transparency is a complex challenge, the chain-of-thought approach is a significant step toward making AI more comprehensible and trustworthy in medicine. This review focuses on understanding the workings of LLMs, particularly how chain-of-thought prompting can be adapted for nephrology's unique requirements. It also aims to thoroughly examine the ethical aspects, clarity, and future possibilities, offering an in-depth view of the exciting convergence of these areas.
| S-EPMC10819595 | biostudies-literature
Project description:Do principles of language processing in the brain affect the way grammar evolves over time or is language change just a matter of socio-historical contingency? While the balance of evidence has been ambiguous and controversial, we identify here a neurophysiological constraint on the processing of language that has a systematic effect on the evolution of how noun phrases are marked by case (i.e. by such contrasts as between the English base form she and the object form her). In neurophysiological experiments across diverse languages we found that during processing, participants initially interpret the first base-form noun phrase they hear (e.g. she…) as an agent (which would fit a continuation like … greeted him), even when the sentence later requires the interpretation of a patient role (as in … was greeted). We show that this processing principle is also operative in Hindi, a language where initial base-form noun phrases most commonly denote patients because many agents receive a special case marker ("ergative") and are often left out in discourse. This finding suggests that the principle is species-wide and independent of the structural affordances of specific languages. As such, the principle favors the development and maintenance of case-marking systems that equate base-form cases with agents rather than with patients. We confirm this evolutionary bias by statistical analyses of phylogenetic signals in over 600 languages worldwide, controlling for confounding effects from language contact. Our findings suggest that at least one core property of grammar systematically adapts in its evolution to the neurophysiological conditions of the brain, independently of socio-historical factors. This opens up new avenues for understanding how specific properties of grammar have developed in tight interaction with the biological evolution of our species.
| S-EPMC4534460 | biostudies-literature
Project description:When sending text messages on their mobile phone to friends, children often use a special type of register, which is called textese. This register allows the omission of words and the use of textisms: instances of non-standard written language such as 4ever (forever). Previous studies have shown that textese has a positive effect on children's literacy abilities. In addition, it is possible that children's grammar system is affected by textese as well, as grammar rules are often transgressed in this register. Therefore, the main aim of this study was to investigate whether the use of textese influences children's grammar performance, and whether this effect is specific to grammar or language in general. Additionally, studies have not yet investigated the influence of textese on children's cognitive abilities. Consequently, the secondary aim of this study was to find out whether textese affects children's executive functions. To investigate this, 55 children between 10 and 13 years old were tested on a receptive vocabulary and grammar performance (sentence repetition) task and various tasks measuring executive functioning. In addition, text messages were elicited and the number of omissions and textisms in children's messages were calculated. Regression analyses showed that omissions were a significant predictor of children's grammar performance after various other variables were controlled for: the more words children omitted in their text messages, the better their performance on the grammar task. Although textisms correlated (marginally) significantly with vocabulary, grammar and selective attention scores and omissions marginally significantly with vocabulary scores, no other significant effects were obtained for measures of textese in the regression analyses: neither for the language outcomes, nor for the executive function tasks. Hence, our results show that textese is positively related to children's grammar performance. On the other hand, use of textese does not affect--positively nor negatively--children's executive functions.
| S-EPMC4816572 | biostudies-literature