Testing a computational model of causative overgeneralizations: Child judgment and production data from English, Hebrew, Hindi, Japanese and K’iche’

How do language learners avoid the production of verb argument structure overgeneralization errors (*The clown laughed the man c.f. The clown made the man laugh), while retaining the ability to apply such generalizations productively when appropriate? This question has long been seen as one that is both particularly central to acquisition research and particularly challenging. Focussing on causative overgeneralization errors of this type, a previous study reported a computational model that learns, on the basis of corpus data and human-derived verb-semantic-feature ratings, to predict adults’ by-verb preferences for less- versus more-transparent causative forms (e.g., * The clown laughed the man vs The clown made the man laugh) across English, Hebrew, Hindi, Japanese and K’iche Mayan. Here, we tested the ability of this model (and an expanded version with multiple hidden layers) to explain binary grammaticality judgment data from children aged 4;0-5;0, and elicited-production data from children aged 4;0-5;0 and 5;6-6;6 (N=48 per language). In general, the model successfully simulated both children’s judgment and production data, with correlations of r=0.5-0.6 and r=0.75-0.85, respectively, and also generalized to unseen verbs. Importantly, learners of all five languages showed some evidence of making the types of overgeneralization errors – in both judgments and production – previously observed in naturalistic studies of English (e.g., *I’m dancing it). Together with previous findings, the present study demonstrates that a simple learning model can explain (a) adults’ continuous judgment data, (b) children’s binary judgment data and (c) children’s production data (with no training of these datasets), and therefore constitutes a plausible mechanistic account of the acquisition of verbs’ argument structure restrictions.

Lexical Input in the Grammatical Expression of Stance: A Collexeme Analysis of the INTRODUCTORY IT PATTERN

Previous research on the INTRODUCTORY IT PATTERN unveiled various lexical and grammatical aspects of its use as a grammatical stance device, including the range of the most frequently used adjectival and verbal stance lexemes, associated stance meanings, the most frequent sub-patterns, and the distinct uses in various contextual settings of the pattern. However, the stance meanings of the pattern, which are deeply rooted in the associated lexical resources, are still understudied. This study explores the meanings of the INTRODUCTORY IT PATTERN by referring to the stance meanings of the pattern associated with the adjectival and verbal lexemes that are statistically attracted to the pattern. The research samples were extracted from the British component of the International Corpus of English (ICE-GB). The samples were manually annotated for different stance types and a collexeme analysis was performed to identify the full range of stance lexemes statistically associated with the INTRODUCTORY IT PATTERN (collexemes). The results show that both adjectival and verbal collexemes are statistically and functionally significant for the delivery of discrete stance types/subtypes. Adjectival collexemes are frequently deployed for all four stance types: Epistemic stance, Evaluation stance, Dynamic stance, and Deontic stance, while verbal collexemes are valuable lexical resources for the Epistemic stance, as their use entails modalized evidentiality, pointing to epistemic judgment of the writer-speaker toward events/propositions. Close examination of the use of adjectival and verbal collexemes identified three fundamental meanings of the INTRODUCTORY IT PATTERN. First, the pattern is inherently evaluative as it tends to attract more lexemes with evaluative meanings and associates evaluative meanings with superficially non-evaluative lexemes. Second, it features a scalarized expression of diversified stance types/subtypes, thus, especially reflective of the scalarized semantic feature of stance expression. Third, it connotates an overwhelmingly positive likelihood judgment. The article concludes by discussing the limitations of this study.

Results of Cluster Analysis of Semantic Features “What is the Person’s Gender? and “What is the Person’s Appearance?” for the Stimulus “Flirting Person””

The present work continues a series of studies concerning ludic competence/ playfulness by means of psycholinguistic instruments and is devoted to description of the behaviour pattern of the ludic position “Diplomat”, which corresponds to flirting as one of the components of playfulness. The key research method is psycholinguistic experiment whose main stage is the controlled association experiment (CAE) with the stimulus “flirting person”. The sample consisted of 215 young respondents (age 21-35). The instructions for the controlled association experiment was developed in the frame of the parametric concept of I.A. Sternin. 23 questions (semantic features) were formulated. They were recognized as relevant to communication for the stimulus “flirting person” and allowed to obtain the material for describing the behaviour pattern of ludic position Diplomat (“flirting person”) reflecting the reality of linguistic consciousness of native speakers. This study presents the results of cluster analysis of two association fields built for the following semantic features: “What is the person’s gender?” and “What is the person’s appearance?”. The results of cluster analysis of the association field built for the semantic feature “What is the person’s gender?” indicate that in the linguistic consciousness of the inhabitants of Ukraine, the overwhelming majority of respondents (93%) accept the binary concept of gender, for 3% of respondents gender does not matter, and only one respondent considers the transgender identity. The analysis of the association field built for the semantic feature “What is the person’s appearance?” allows us to assert that the absolute majority of respondents (78%) demonstrates an emotionally positive attitude towards “flirting person”.

Java Bytecode Control Flow Classification: Framework for Guiding Java Decompilation

Decompilation is the main process of software development, which is very important when a program tries to retrieve lost source codes. Although decompiling Java bytecode is easier than bytecode, many Java decompilers cannot recover originally lost sources, especially the selection statement, i.e., if statement. This deficiency affects directly decompilation performance. In this paper, we propose the methodology for guiding Java decompiler to deal with the aforementioned problem. In the framework, Java bytecode is transformed into two kinds of features called frame feature and latent semantic feature. The former is extracted directly from the bytecode. The latter is achieved by two-step transforming the Java bytecode to bigram and then term frequency-inverse document frequency (TFIDF). After that, both of them are fed to the genetic algorithm to reduce their dimensions. The proposed feature is achieved by converting the selected TFIDF to a latent semantic feature and concatenating it with the selected frame feature. Finally, KNN is used to classify the proposed feature. The experimental results show that the decompilation accuracy is 93.68 percent, which is obviously better than Java Decompiler.

Modelling concrete and abstract concepts using brain-constrained deep neural networks

A neurobiologically constrained deep neural network mimicking cortical areas relevant for sensorimotor, linguistic and conceptual processing was used to investigate the putative biological mechanisms underlying conceptual category formation and semantic feature extraction. Networks were trained to learn neural patterns representing specific objects and actions relevant to semantically ‘ground’ concrete and abstract concepts. Grounding sets consisted of three grounding patterns with neurons representing specific perceptual or action-related features; neurons were either unique to one pattern or shared between patterns of the same set. Concrete categories were modelled as pattern triplets overlapping in their ‘shared neurons’, thus implementing semantic feature sharing of all instances of a category. In contrast, abstract concepts had partially shared feature neurons common to only pairs of category instances, thus, exhibiting family resemblance, but lacking full feature overlap. Stimulation with concrete and abstract conceptual patterns and biologically realistic unsupervised learning caused formation of strongly connected cell assemblies (CAs) specific to individual grounding patterns, whose neurons were spread out across all areas of the deep network. After learning, the shared neurons of the instances of concrete concepts were more prominent in central areas when compared with peripheral sensorimotor ones, whereas for abstract concepts the converse pattern of results was observed, with central areas exhibiting relatively fewer neurons shared between pairs of category members. We interpret these results in light of the current knowledge about the relative difficulty children show when learning abstract words. Implications for future neurocomputational modelling experiments as well as neurobiological theories of semantic representation are discussed.

Sentiment Weighted Word Embedding for Big Text Data

Sentiment analysis is the practice of eliciting a sentiment orientation of people's opinions (i.e. positive, negative and neutral) toward the specific entity. Word embedding technique like Word2vec is an effective approach to encode text data into real-valued semantic feature vectors. However, it fails to preserve sentiment information that results in performance deterioration for sentiment analysis. Additionally, big sized textual data consisting of large vocabulary and its associated feature vectors demands huge memory and computing power. To overcome these challenges, this research proposed a MapReduce based Sentiment weighted Word2Vec (MSW2V), which learns the sentiment and semantic feature vectors using sentiment dictionary and big textual data in a distributed MapReduce environment, where memory and computing power of multiple computing nodes are integrated to accomplish the huge resource demand. Experimental results demonstrate the outperforming performance of the MSW2V compared to the existing distributed and non-distributed approaches.