Features Of Natural Language Research Articles

Customers' sentiment on food delivery services: An Arabic text mining approach

The Covid-19 pandemic has accelerated the shift in organizations' strategies toward innovative online services. Customer reviews on platforms for online ordering and delivery are a vital source of information about how well a business is performing. Businesses that provide food delivery services (FDS) seek to leverage consumer input to locate areas where customer satisfaction could be raised. Sentiment analysis (SA) has been the subject of an enormous amount of English-language research. Despite Arabic's increasing popularity as a writing language on the Internet, not much study has been conducted on sentiment analysis of Arabic up to this point, with a limited number of publicly available resources for Arabic SA such as datasets and lexicons. The present study collects FDS-related reviews in Arabic to conduct extensive emotion mining, taking advantage of Natural Language Processing, feature selection, and Machine Learning techniques to elicit personal judgments, identify polarity, and recognize customers’ feelings in the FDS domain. To demonstrate that the proposed approach is suitable for analyzing human perceptions of FDS, we designed and carried out excessive experiments that assess the utility of each phase. Our highest categorization accuracy was 90 % using Mutual Information with the SVM classifier. The study's findings provide various managerial insights for improving their plans and service delivery, as well as revealing the main reasons for consumer complaints. It also demonstrates how future academics might harness the power of online business reviews in Arabic using a variety of text-mining approaches.

A framework for the emergence and analysis of language in social learning agents

Neural systems have evolved not only to solve environmental challenges through internal representations but also, under social constraints, to communicate these to conspecifics. In this work, we aim to understand the structure of these internal representations and how they may be optimized to transmit pertinent information from one individual to another. Thus, we build on previous teacher-student communication protocols to analyze the formation of individual and shared abstractions and their impact on task performance. We use reinforcement learning in grid-world mazes where a teacher network passes a message to a student to improve task performance. This framework allows us to relate environmental variables with individual and shared representations. We compress high-dimensional task information within a low-dimensional representational space to mimic natural language features. In coherence with previous results, we find that providing teacher information to the student leads to a higher task completion rate and an ability to generalize tasks it has not seen before. Further, optimizing message content to maximize student reward improves information encoding, suggesting that an accurate representation in the space of messages requires bi-directional input. These results highlight the role of language as a common representation among agents and its implications on generalization capabilities.

Automated Grading Model with Adjusted Level of Lenience for Short Answer Questions using Natural Language Processing

Automated grading of short answer questions is a challenging task that involves understanding and evaluating free-text responses. This research presents an innovative model that combines the capabilities of the language model all-mpnet-base-v2 with a machine learning-based lenience adjustment mechanism to enhance the accuracy and fairness of automated grading systems. The proposed model utilizes all-mpnet-base-v2 for natural language understanding and feature extraction from student responses. To address the variability in acceptable answers and provide a fair grading system, a machine learning-based model is integrated to adjust the level of lenience dynamically. This dual approach ensures that the grading system can handle a wide range of responses while maintaining consistency and reliability. The experimental results demonstrate that the combination of all-mpnet-base-v2 with the lenience adjustment model significantly improves grading accuracy compared to traditional methods. This model represents a significant advancement in the field of educational technology, offering a robust solution for automated grading systems that can adapt to diverse educational contexts and requirements.

Mining software insights: uncovering the frequently occurring issues in low-rating software applications.

In today's digital world, app stores have become an essential part of software distribution, providing customers with a wide range of applications and opportunities for software developers to showcase their work. This study elaborates on the importance of end-user feedback for software evolution. However, in the literature, more emphasis has been given to high-rating & popular software apps while ignoring comparatively low-rating apps. Therefore, the proposed approach focuses on end-user reviews collected from 64 low-rated apps representing 14 categories in the Amazon App Store. We critically analyze feedback from low-rating apps and developed a grounded theory to identify various concepts important for software evolution and improving its quality including user interface (UI) and user experience (UX), functionality and features, compatibility and device-specific, performance and stability, customer support and responsiveness and security and privacy issues. Then, using a grounded theory and content analysis approach, a novel research dataset is curated to evaluate the performance of baseline machine learning (ML), and state-of-the-art deep learning (DL) algorithms in automatically classifying end-user feedback into frequently occurring issues. Various natural language processing and feature engineering techniques are utilized for improving and optimizing the performance of ML and DL classifiers. Also, an experimental study comparing various ML and DL algorithms, including multinomial naive Bayes (MNB), logistic regression (LR), random forest (RF), multi-layer perception (MLP), k-nearest neighbors (KNN), AdaBoost, Voting, convolutional neural network (CNN), long short-term memory (LSTM), bidirectional long short term memory (BiLSTM), gated recurrent unit (GRU), bidirectional gated recurrent unit (BiGRU), and recurrent neural network (RNN) classifiers, achieved satisfactory results in classifying end-user feedback to commonly occurring issues. Whereas, MLP, RF, BiGRU, GRU, CNN, LSTM, and Classifiers achieved average accuracies of 94%, 94%, 92%, 91%, 90%, 89%, and 89%, respectively. We employed the SHAP approach to identify the critical features associated with each issue type to enhance the explainability of the classifiers. This research sheds light on areas needing improvement in low-rated apps and opens up new avenues for developers to improve software quality based on user feedback.

Automatic detection of expressed emotion from Five-Minute Speech Samples: Challenges and opportunities.

Research into clinical applications of speech-based emotion recognition (SER) technologies has been steadily increasing over the past few years. One such potential application is the automatic recognition of expressed emotion (EE) components within family environments. The identification of EE is highly important as they have been linked with a range of adverse life events. Manual coding of these events requires time-consuming specialist training, amplifying the need for automated approaches. Herein we describe an automated machine learning approach for determining the degree of warmth, a key component of EE, from acoustic and text natural language features. Our dataset of 52 recorded interviews is taken from recordings, collected over 20 years ago, from a nationally representative birth cohort of British twin children, and was manually coded for EE by two researchers (inter-rater reliability 0.84-0.90). We demonstrate that the degree of warmth can be predicted with an F1-score of 64.7% despite working with audio recordings of highly variable quality. Our highly promising results suggest that machine learning may be able to assist in the coding of EE in the near future.

NCOG-15. COMPARATIVE ANALYSIS OF GLIOMA INDUCED IMPAIRMENT MODELS

Abstract Speech is produced by the coordination of neural networks involved in language processing and motor control of vocal outputs, both of which can be disrupted by diffuse gliomas. Natural language is the instinctual written or spoken language profile that an individual has developed throughout his or her lifetime. Recent psycholinguistic analyses suggest that patterns of natural language may be an accurate measure for assessing clinical and behavioral conditions. One important feature of natural language is speech onset. Speech delays are common complaints among patients with glioma. Here, we assess the relationship between speech onset and histologic-molecular features of glioma. Speech onset was defined as acoustic measure of the interval of time between the release of the oral constriction in a stop consonant and the onset of subsequent voicing. Audio samples and video analysis of oral kinetics were analyzed from patients with newly diagnosed low- and high-grade glioma. Features were extracted by manual segmentation and aligned for each trial with high-density electrode arrays to record task specific responses from gliomas projecting to the cortical surface. A total of 1,120 stimuli (672 = picture naming (PN) and 448 auditory naming (AN) trials were presented. As expected, patients with glioblastoma had higher picture and auditory naming error rates (p < 0.001, Mann-Whitney U test) compared with diffuse low-grade gliomas. Assuming speech errors produce slower response times, we next measured only correct behavioral responses which demonstrated 7% slower speech initiation responses (p < 0.0001, Mann-Whitney U test) when comparing patients with glioblastoma and diffuse low-grade gliomas. Slowed speech responses did not require glioma infiltration into canonical cortical speech areas of the lateral prefrontal cortex and superior temporal gyrus. Finally, we identified electrophysiological biomarkers of speech errors. These results deepen our understanding of speech processing impairments, which are not based purely on coding errors.

A New Generative Model for Textual Descriptions of Medical Images Using Transformers Enhanced with Convolutional Neural Networks.

The automatic generation of descriptions for medical images has sparked increasing interest in the healthcare field due to its potential to assist professionals in the interpretation and analysis of clinical exams. This study explores the development and evaluation of a generalist generative model for medical images. Gaps were identified in the literature, such as the lack of studies that explore the performance of specific models for medical description generation and the need for objective evaluation of the quality of generated descriptions. Additionally, there is a lack of model generalization to different image modalities and medical conditions. To address these issues, a methodological strategy was adopted, combining natural language processing and features extraction from medical images and feeding them into a generative model based on neural networks. The goal was to achieve model generalization across various image modalities and medical conditions. The results showed promising outcomes in the generation of descriptions, with an accuracy of 0.7628 and a BLEU-1 score of 0.5387. However, the quality of the generated descriptions may still be limited, exhibiting semantic errors or lacking relevant details. These limitations could be attributed to the availability and representativeness of the data, as well as the techniques used.

Detecting bots in social-networks using node and structural embeddings

Users on social networks such as Twitter interact with each other without much knowledge of the real-identity behind the accounts they interact with. This anonymity has created a perfect environment for bot accounts to influence the network by mimicking real-user behaviour. Although not all bot accounts have malicious intent, identifying bot accounts in general is an important and difficult task. In the literature there are three distinct types of feature sets one could use for building machine learning models for classifying bot accounts. These feature-sets are: user profile metadata, natural language features (NLP) extracted from user tweets and finally features extracted from the the underlying social network. Profile metadata and NLP features are typically explored in detail in the bot-detection literature. At the same time less attention has been given to the predictive power of features that can be extracted from the underlying network structure. To fill this gap we explore and compare two classes of embedding algorithms that can be used to take advantage of information that network structure provides. The first class are classical embedding techniques, which focus on learning proximity information. The second class are structural embedding algorithms, which capture the local structure of node neighbourhood. We show that features created using structural embeddings have higher predictive power when it comes to bot detection. This supports the hypothesis that the local social network formed around bot accounts on Twitter contains valuable information that can be used to identify bot accounts.

Distribution of sentence length and dependency distance in children's compositions: Characteristics of natural language and variations in language development

This study analyzed the distribution of the sentence length and mean of dependency distances (MDD) in Japanese sentences, comparing data from random sources with that obtained from children's compositions, and identifying changes in distribution according to grade level. Findings reveal that the sentence length in random data is well suited to a geometric distribution, whereas MDD is well suited to a lognormal distribution. In contrast, data from children's compositions show a shift in the distribution of the number of clauses from a lognormal to a gamma distribution, depending on the school year, with MDD suiting a gamma distribution. Mean MDD increases exponentially with the logarithm of the number of clauses in random data, while it increases linearly in composition data, thus generally supporting previous findings that dependency distances are optimized in natural language. However, MDDs exhibit non-monotonic changes with grades, suggesting the complexity of children's language development.

Discovering Critical Factors in the Content of Crowdfunding Projects

Crowdfunding can simplify the financing process to raise large amounts of money to complete projects for startups. However, improving the success rate has become one of critical issues. To achieve this goal, fundraisers need to create a short video, attractive promotional content, and present themselves on social media to attract investors. Previous studies merely discussed project factors that affect crowdfunding success rates. However, from the available literature, relatively few studies have studied what elements should be involved in the project content for the success of crowdfunding projects. Consequently, this study aims to extract the crucial factors that can enhance the crowdfunding project success rate based on the project content description. To identify the crucial project content factors of movie projects, this study employed two real cases from famous platforms by using natural language processing (NLP) and feature selection algorithms including rough set theory (RST), decision trees (DT), and ReliefF, from 12 pre-defined candidate factors. Then, support vector machines (SVM) were used to evaluate the performance. Finally, “Role”, “Cast”, “Merchandise”, “Sound effects”, and “Sentiment” were identified as important content factors for movie projects. The findings also could provide fundraisers with suggestions on how to make their movie crowdfunding projects more successful.

RANDOM FOREST-BASED CLASSIFIER FOR AUTOMATIC SARCASM CLASSIFICATION ON TWITTER DATA USING MULTIPLE FEATURES

Sarcasm is one of the nonliteral languages usually employed in social networks and microblogging websites to convey implicit information in an individual communication message. This could lead to the misclassification of tweets. This paper focuses on sarcasm detection on tweets, which has been experimented with the use of textual features. The textual features comprise the Neural language fusion and Natural language features, which include sentiment-related features, semantic and synthetic features, punctuation-related features, and GloVe embedding features. The features mentioned above were extracted separately from the target tweet and fused to form fused features for the target tweet. The proposed predictive model attained an accuracy of 86.9% with a random forest classifier, which outperformed other models employed in the experiment, such as DT (83.9), SVM (80.5), KNN (83.1), and LR (82.9).

Long-term memory predictors of adult language learning at the interface between syntactic form and meaning.

Recent neurocognitive models of second language learning have posited specific roles for declarative and procedural memory in the processing of novel linguistic stimuli. Pursuing this line of investigation, the present exploratory study examined the role of declarative and procedural memory abilities in the early stages of adult comprehension of sentences in a miniature language with natural language characteristics (BrocantoJ). Thirty-six native Italian young adults were aurally exposed to BrocantoJ in the context of a computer game over three sessions on consecutive days. Following vocabulary training and passive exposure, participants were asked to perform game moves described by aural sentences in the language. Game trials differed with respect to the information the visual context offered. In part of the trials processing of relationships between grammatical properties of the language (word order and morphological case marking) and noun semantics (thematic role) was necessary in order reach an accurate outcome, whereas in others nongrammatical contextual cues were sufficient. Declarative and procedural learning abilities were respectively indexed by visual and verbal declarative memory measures and by a measure of visual implicit sequence learning. Overall, the results indicated a substantial role of declarative learning ability in the early stages of sentence comprehension, thus confirming theoretical predictions and the findings of previous similar studies in miniature artificial language paradigms. However, for trials that specifically probed the learning of relationships between morphosyntax and semantics, a positive interaction between declarative and procedural learning ability also emerged, indicating the cooperative engagement of both types of learning abilities in the processing of relationships between ruled-based grammar and interpretation in the early stages of exposure to a new language in adults.

Joint visual template and natural language for robust visual tracking

At present, the target of interest in visual tracking is given in the form of a bounding box. Due to the randomness of the target shape, the bounding box may contain a lot of non-target information. When encountering complex tracking scenarios, the performance of the tracker reduces severely. To address this problem, in this letter, the authors propose a novel tracking framework based on the joint of the visual template and natural language (VNTrack) to alleviate the impact of bounding box ambiguity. Specifically, the authors first use a pre-trained language model to extract the features of the language description of the target. Then, a feature alignment module is designed to align and enhance the visual template feature and natural language feature. In addition, the authors design a multimodal query module to fuse the visual template, natural language, and search region information. Experimental results over tracking benchmarks with language annotations show that the proposed VNTrack is competitive among the state-of-the-art trackers.

Computational Linguistics with Deep-Learning-Based Intent Detection for Natural Language Understanding

Computational linguistics explores how human language is interpreted automatically and then processed. Research in this area takes the logical and mathematical features of natural language and advances methods and statistical procedures for automated language processing. Slot filling and intent detection are significant modules in task-based dialogue systems. Intent detection is a critical task in any natural language understanding (NLU) system and constitutes the base of a task-based dialogue system. In order to build high-quality, real-time conversational solutions for edge gadgets, there is a demand for deploying intent-detection methods on devices. This mandates an accurate, lightweight, and fast method that effectively operates in a resource-limited environment. Earlier works have explored the usage of several machine-learning (ML) techniques for detecting intent in user queries. In this article, we propose Computational Linguistics with Deep-Learning-Based Intent Detection and Classification (CL-DLBIDC) for natural language understanding. The presented CL-DLBIDC technique receives word embedding as input and learned meaningful features to determine the probable intention of the user query. In addition, the presented CL-DLBIDC technique uses the GloVe approach. In addition, the CL-DLBIDC technique makes use of the deep learning modified neural network (DLMNN) model for intent detection and classification. For the hyperparameter tuning process, the mayfly optimization (MFO) algorithm was used in this study. The experimental analysis of the CL-DLBIDC method took place under a set of simulations, and the results were scrutinized for distinct aspects. The simulation outcomes demonstrate the significant performance of the CL-DLBIDC algorithm over other DL models.

Language and Intelligence, Artificial vs. Natural or What Can and What Cannot AI Do with NL?

In this talk, I argue that there are certain pragmatic features of natural language (that I will call 'productivity' and 'malleability', on top of syntactical generativity and semantical compositionality), which are not only hard, but even impossible to capture in an artificial language used by an AI system, and the reason for this is to be found in certain deep, metaphysical differences between artificial and natural intelligence, accounting for the differences in their respective processes of concept-formation.

Classification of specialities in textual medical reports based on natural language processing and feature selection

Nowadays, a great deal of detailed information about patients, including disease status, medication history, and side effects, is collected in an electronic format; called an electronic medical record (EMR), and the data serves as a valuable resource for further analysis, diagnosis, and treatment. The huge q uantity of detailed patient information in these medical texts produces a huge challenge in terms of processing this data efficiently, however. Machine learning (ML) algorithms, artificial intelligence techniques, and natural language processing tools can have the potential effect of simplifying unstructured data, which could positively affect medical report analysis. Natural language processing (NLP) has recently made huge advances on a variety of tasks. In this paper, an automatic system was thus produced to classify specialist consultant interactions based on patients’ medical reports. NLP was used as a pre-processing step on a dataset formed of unstructured medical reports. Feature extraction and selection methods were used to convert the textual reports into sets of features and to extract the most effective features to increase classification accuracy and reduce execution time. Various classification methods were then applied (ML perceptron, logistic regression random forest (RF), and linear support vec tor classifier (LSVC)). The highest accuracy (99.39%) was achieved in ML-perceptron classification techniques .

Modelling dynamic behaviour of agents in a multiagent world: Logical analysis of Wh-questions and answers

Abstract In a multiagent and multi-cultural world, the fine-grained analysis of agents’ dynamic behaviour, i.e. of their activities, is essential. Dynamic activities are actions that are characterized by an agent who executes the action and by other participants of the action. Wh-questions on the participants of the actions pose a difficult particular challenge because the variability of the types of possible answers to such questions is huge. To deal with the problem, we propose the analysis and classification of Wh-questions apt for agents’ communication in a multiagent system (MAS). Our proposal of such a system consists of agents who communicate with their fellow agents by messaging so that each autonomous agent, though resource-bounded, can make less or more rational decisions to meet its own and collective goals. In addition, by communicating with other fellow agents and their environment, agents can learn new concepts and enrich their ontology so that their behaviour is dynamic. We aim to make a general proposal of the system so that the ‘envelope’ of agents’ messages can be formalized in any MAS standard, be it The Foundation for Intelligent Physical Agents - Agent Communication Language (FIPA-ACL) or Knowledge Query and Manipulation Language (KQML). Yet, the content of messages is encoded in a formalized natural language. To this end, we apply Transparent Intensional Logic (TIL) with its procedural semantics which is particularly apt for a fine-grained analysis in which all the semantically salient features of natural language can be plausibly formalized. In this paper, we concentrate on analysing the content of query messages, particularly the content of those that encode Wh-questions and the answers to them. We also summarize TIL deduction system that makes it possible to answer such questions in an intelligent way. Linguists distinguish several subtypes of Wh-questions. Though the linguistic classification is helpful, it is not always suitable for agents’ communication and reasoning. We need the classification based on a logical analysis of Wh-questions so that the agents can infer possible answers to such questions rather than only looking for them by keywords. This paper aims to apply an appropriate classification of the logical types of Wh-questions and the analysis of such questions; we concentrate in particular on questions concerning the participants of activities. The application of these results to the analysis of processes and events based on verb valency frames is another novelty of the paper.

Identification of maternal depression risk from natural language collected in a mobile health app

Depression is one of the most common pregnancy complications, affecting approximately 15% of pregnant people. While valid psychometric measures of depression risk exist, they are not consistently administered at routine prenatal care, exacerbating the problem of adequate detection. The language we use in daily life offers a window into our psychological wellbeing. In this longitudinal observational cohort study of prenatal patients using a prenatal care mobile health app, we examine how features of app-entered natural language and other app-entered patient-reported data may be used as indicators for validated depression risk measures. Patient participants (n=1091) were prescribed a prenatal care app as part of a quality improvement initiative in the UPMC healthcare system from September 2019 – May 2022. Natural language from open-ended writing prompts in the app and self-reported daily mood, were entered by patients using the tool. Participants also completed a validated measure of depression risk - the Edinburgh Postnatal Depression Scale (EPDS) - at least once in their pregnancy. A variety of natural language processing tools were used to score sentiment, categorize topics, and capture other semantic and syntactic information from text entries. LASSO was used to model the relationship between the natural language features and depression risk. Open-ended text within a 30-day and 60-day timeframe of completing an EPDS was found to be moderately predictive of moderate to severe depression risk (AUROC=0.66 and 0.67, for each respective timeframe). When combined with average daily reported mood, open-ended text showed good predictive power (AUROC=0.87). Consistently predictive language features across all models included themes of “money” and “sadness.” The combination of natural language and other user-reported data collected through a mobile health app offers an opportunity for identifying depression risk among a pregnant population.

Human-Generated Web Data Disentanglement for Complex Event Processing

In social media, human-generated web data from real-world events have become exponentially complex due to the chaotic and spontaneous features of natural language. This may create an information overload for the information consumers, and in turn not easily digest a large amount of information in a limited time. To tackle this issue, we propose to use Complex Event Processing (CEP) and semantic web reasoners to disentangle the human-generated data and present users with only relevant and important data. However, one of the key obstacles is that the human-generated data can have no structured meaning sometimes even for the speaker, hindering the output of the CEP. Therefore, in order to adapt to the CEP inputs, we present two different techniques that allow for the discrimination and digestion of value of human-generated data. The first one relies on the Variable Sharing Property that was developed for relevance logics, while the second one is based on semantic equivalence and natural language processing. The results can be given to CEP for further semantic reasoning and generate digested information for users.

A spatiotemporal multi-feature extraction framework for opinion mining

With the rapid development of Internet technology and the explosive growth of digital text, opinion mining has become one of the important research hotspots in the field of natural language processing (NLP). In recent years, neural network based deep learning algorithms have been applied in the field of opinion mining. Considering the relation between temporal and spatial dimensions of text data and the characteristics of natural language itself, traditional deep learning algorithms cannot be comprehensive in the processing of fully feature extraction. In this paper, we propose a new deep learning framework for opinion mining, which includes a temporal feature extraction layer that consists of two layers of bidirectional simple recurrent unit (Bi-SRU) networks extracting features at the word and grammar levels; a semantic feature extraction layer that mainly contains a multi-head attention module; a spatial feature extraction layer with dilated convolution that is used to extract opinion preference features. The Internet movie database (IMDb) is used to verify the performance of the proposed framework. The experiment results show that the proposed framework can effectively improve the classification accuracy, whose performance is better than that of the compared algorithms.