Downloads & Free Reading Options - Results

This cross-sectional study investigates the application of BERT-based natural language processing models for automating the classification of suicidal ideation severity from free-text responses to the Columbia Suicide Severity Rating Scale (C-SSRS). Using data from 1,443 adults who had recently attempted suicide across ten Spanish university hospitals (2019-2023), we developed and evaluated machine learning models to classify responses across five key C-SSRS categories: desire to die, active suicidal thoughts, suicidal ideation with methods, interrupted/aborted attempts, and preparatory behaviors. Our BERT-based models achieved high accuracy across all categories (77.3% to 94.44%), with particularly strong performance in identifying active suicidal ideation (F1-score: 0.98) and classifying preparatory acts (F1-score: 0.97). The models outperformed traditional approaches (SVM, Naive Bayes) in global classification tasks (78.3% vs. 74.6% and 71.2% accuracy respectively). We applied SMOTE to address class imbalance and conducted comprehensive error analysis to identify linguistic factors affecting classification performance. The findings demonstrate that transformer-based NLP can effectively analyze structured clinical interview data for suicide risk assessment, potentially enhancing clinical decision-making and enabling more timely interventions. These automated methods may complement clinical expertise in suicide prevention efforts, though human oversight remains essential. The study contributes to the growing application of advanced computational methods in mental health assessment and provides practical insights for clinical practice in suicide risk evaluation.

This cross-sectional study investigates the application of BERT-based natural language processing models for automating the classification of suicidal ideation severity from free-text responses to the Columbia Suicide Severity Rating Scale (C-SSRS). Using data from 1,443 adults who had recently attempted suicide across ten Spanish university hospitals (2019-2023), we developed and evaluated machine learning models to classify responses across five key C-SSRS categories: desire to die, active suicidal thoughts, suicidal ideation with methods, interrupted/aborted attempts, and preparatory behaviors. Our BERT-based models achieved high accuracy across all categories (77.3% to 94.44%), with particularly strong performance in identifying active suicidal ideation (F1-score: 0.98) and classifying preparatory acts (F1-score: 0.97). The models outperformed traditional approaches (SVM, Naive Bayes) in global classification tasks (78.3% vs. 74.6% and 71.2% accuracy respectively). We applied SMOTE to address class imbalance and conducted comprehensive error analysis to identify linguistic factors affecting classification performance. The findings demonstrate that transformer-based NLP can effectively analyze structured clinical interview data for suicide risk assessment, potentially enhancing clinical decision-making and enabling more timely interventions. These automated methods may complement clinical expertise in suicide prevention efforts, though human oversight remains essential. The study contributes to the growing application of advanced computational methods in mental health assessment and provides practical insights for clinical practice in suicide risk evaluation.

Chat-based counseling became a popular low-threshold intervention for the provision of mental health support to the youth. Simultaneously, advances in artificial intelligence as well as the availability of large datasets at those services make them particularly suitable for the development and implementation of data-driven algorithms to support care. Adding to this line of research, this study explores a natural language processing (NLP) based approach for the automated real-time evaluation of chat conversations. Specifically, text data combined with feedback responses is used to train a classifier to predict how helpful a consultation was, as perceived by the young chatters. Such a system could improve quality monitoring in this highly naturalistic setting but also enable the real-time adjustment of treatment for unhelpful interactions.

The purpose of this project is to build a ChatBot that utilises NLP (Natural Language Processing) and assists customers. A ChatBot is an automated conversation system that replies to users' queries by analysing them using NLP and assists them in every way it can. In this project, we are trying to implement a customer service chatbot that tries to converse and assist the user in some simple scenarios. This chat bot can take simple user queries as input, process them, classify them into one of the existing tags, and respond to them with an appropriate response. If the user's queries are too complex for the bot, it will re-direct the conversation to an actual person. The ChatBot is going to be based on a machine learning model that is built using PyTorch (Python Deep Learning library) and NLTK (Natural Language Tool Kit). The model used here is a feed-forward neural network. There are 3 layers in this neural network, i.e., the input layer, the hidden layer, and the output layer. The number of nodes in the input and hidden layers depends on the total number of distinct words present in the data set. whereas the output contains the same number of nodes as the number of distinct tags the data set is divided into. This kind of neural network is perfect for building simple chatbots as it does not require high computational power either for training or for deploying. The chatbot we built is for a coffee shop, and it performs actions like ordering coffee, telling a joke, suggesting a drink, etc. Although this chatbot is relatively simple, it is highly customizable, thus making it easy to implement it in any scenario.One of the main features of this ChatBot is that the dataset it is trained on is easy to customise and we can add new tags easily, but the neural network need not be altered in most cases, making this a very reliable model. Many chatbots similar to this are being used in fields like medicine, government agencies, automated food ordering systems, etc. This feature also makes training and testing the chatbot very easy to customize.

The course in computational linguistics described in this paper was given at The American University during the spring semester of 1969. The purpose of the course was "to convey to students with no previous experience an appreciation of the growing art of computational linguistics which encompasses every use to which computers can be put in manipulation of natural language." Each of the 16 class sessions is briefly outlined and a number of articles and books for each class are listed for recommended reading. The majority of these references are available in published form or from the Clearinghouse for Federal Scientific and Technical Information in Springfield, Virginia. (JD)

Summarization is an effective strategy to promote and enhance learning and deep comprehension of texts. However, summarization is seldom implemented by teachers in classrooms because the manual evaluation of students' summaries requires time and effort. This problem has led to the development of automated models of summarization quality. However, these models often rely on features derived from expert ratings of student summarizations of specific source texts and are therefore not generalizable to summarizations of new texts. Further, many of the models rely of proprietary tools that are not freely or publicly available, rendering replications difficult. In this study, we introduce an automated summarization evaluation (ASE) model that depends strictly on features of the source text or the summary, allowing for a purely textbased model of quality. This model effectively classifies summaries as either low or high quality with an accuracy above 80%. Importantly, the model was developed on a large number of source texts allowing for generalizability across texts. Further, the features used in this study are freely and publicly available affording replication. [This paper was published in: S. Isotani et al. (Eds.), "AIED 2019" (pp. 84-95). Switzerland: Springer.]

A survey is presented of Automated Language Processing done in 1966. It is limited in scope to analytical processing of natural language, excluding work in programming languages, speech recognition, and statistical processing of text. It focuses on work aimed at generating and analyzing sentences of a natural language. This survey has four major sections: The first, on syntactic theory, contains a summary of the principal assumptions underlying work in generative grammar and a report of the most significant developments in theoretical and descriptive work in syntax. The second section, on semantic theory, attempts to provide some dimensions along which we can judge various theories that have been proposed and developed in the literature in 1966. A number of empirical studies related to semantics and psycholinguistics are reported in a third section. Finally, a fourth section discusses various computer systems for manipulation of natural language. These range from systems that support linguistic studies to systems that are attempting to utilize natural language as a communication medium.

Includes bibliographies

Automating the process of assigning security classifications to unstructured text would facilitate a transition to a data-centric architecture-one that promotes information sharing, in which all data in an organization are electronically labelled. In this document, we report the results of a series of experiments conducted to investigate the effectiveness of using statistical natural language processing and machine learning techniques to automatically assign security classifications to documents. We present guidelines for selecting parameters to maximize the accuracy of a machine learning algorithm's classification decisions for several well-defined collections of documents. We examine the significance of a document's topic and the effect of security policy changes on the ability of our system to automate classification; we include design recommendations to address both topic and policy considerations. Our classification techniques prove effective at assessing a document's sensitivity, achieving accuracies upwards of 80%.

This article aims to clarify the aspects of teaching the Arabic language in a modern way. By benefits from cognitive linguistics as a mental approach that works on the mechanisms of perception and the creative role played by the human mind on one hand, and the applications of artificial intelligence on the other hand. This study deals with the new vision of the cognitive linguistic approach to language and the world’s openness to the field of artificial intelligence by investing in effectiveness- cognitive transformation and linking language to the human brain. The study benefited from the appropriate descriptive approach to display and analyze work data. Reaching results demonstrated that artificial intelligence gained an important position in the field of cognitive sciences; as it revealed human recognition of information technology and the exploitation of computer science in the service of humanity. Artificial intelligence science is also based on two pillars, which are computer software and the machine, as the program represents the human mind on one hand, and on the other hand, the machine with its tools represents the human body with its organs. DOI: https://doi.org/10.70091/Atras/vol06no01.47