Piece Of Text Research Articles

Conoce Tus Opciones: The Challenges of Communicating Rip Current Information in Spanish

Abstract Rip currents are poorly understood by the public and thus a leading cause of weather-related fatalities in the United States. Individuals who speak languages other than English, such as Spanish-speaking communities, have been historically underserved by communication efforts for natural hazards and thus are at additional risk from rip currents. In response, the National Weather Service (NWS) has developed rip current outreach tools translated into Spanish in recent years, though their efficacy has not been systematically tested. We therefore surveyed members of the Spanish-speaking community in the New York City area to gain insights into how the NWS two-page rip current informational brochure is working and possible improvements to be made. The English version of the brochure was also tested, and the results from the two languages were compared. Quantitative survey results indicate that while both brochures are generally effective, several changes could be made to improve clarity, especially for Spanish speakers, who found more issues with the Spanish version. For example, the translation of the term rip current itself used by the NWS, “corrientes de resaca,” was determined to be confusing by some Spanish speakers, as were other pieces of text with complex scientific terms. Moreover, certain graphics contained in the brochure were difficult to understand for users in both languages. The results of this work can be used by the NWS and other agencies to improve their Spanish risk communication tools and can also be used as a guide when translating scientific information into other languages. Significance Statement Rip currents, narrow channels of fast-moving water that can pull swimmers away from the shore, are among the leading weather-related killers in the United States yet are poorly understood by the public. Recent efforts to reduce society’s vulnerability to rip currents include the translation of outreach materials from English to Spanish. This study evaluated a new National Weather Service Spanish rip current informational brochure and compared it to the English version. The results show that while both the English and Spanish brochures were generally effective at communicating risk, the Spanish version had more issues, primarily surrounding the translation of scientific terms and phrases. Findings from this research can help guide scientists in designing improved non-English severe weather outreach products.

Triangulating evidence in health sciences with Annotated Semantic Queries.

Integrating information from data sources representing different study designs has the potential to strengthen evidence in population health research. However, this concept of evidence "triangulation" presents a number of challenges for systematically identifying and integrating relevant information. These include the harmonization of heterogenous evidence with common semantic concepts and properties, as well as the priortization of the retrieved evidence for triangulation with the question of interest. We present Annotated Semantic Queries (ASQ), a natural language query interface to the integrated biomedical entities and epidemiological evidence in EpiGraphDB, which enables users to extract "claims" from a piece of unstructured text, and then investigate the evidence that could either support, contradict the claims, or offer additional information to the query. This approach has the potential to support the rapid review of preprints, grant applications, conference abstracts, and articles submitted for peer review. ASQ implements strategies to harmonize biomedical entities in different taxonomies and evidence from different sources, to facilitate evidence triangulation and interpretation. ASQ is openly available at https://asq.epigraphdb.org and its source code is available at https://github.com/mrcieu/epigraphdb-asq under GPL-3.0 license.

Ghosts and the machine: testing the use of Artificial Intelligence to deliver historical life course biographies from big data

This article presents the findings of an experiment in the use of Artificial Intelligence text generation processes to convert historical ‘big data’ into narrative text. Using an extensive collection of records pertaining to the Australian colonial settlement of Norfolk Island in the South Pacific (1788–1814), we investigate Generative Large Language Model technology for converting tabulated data from the site into short pieces of novel text, describing the lives of transported convicts and free individuals recorded in our databases. These personalized stories are assessed for fluency and factual correctness. Using this process, we uncover some instructive problems and caveats. We detect AI’s inherent tendency toward bias and uncritical perspectives, including potentially offensive stereotypes. We also discover an unwelcome tendency to summarize data. So, whilst the outputs are for the most part effective and functional, we find that the best results still require artful human intervention to fully capture the most human aspects of history and heritage research.

Social Media Topic Classification on Greek Reddit

Text classification (TC) is a subtask of natural language processing (NLP) that categorizes text pieces into predefined classes based on their textual content and thematic aspects. This process typically includes the training of a Machine Learning (ML) model on a labeled dataset, where each text example is associated with a specific class. Recent progress in Deep Learning (DL) enabled the development of deep neural transformer models, surpassing traditional ML ones. In any case, works of the topic classification literature prioritize high-resource languages, particularly English, while research efforts for low-resource ones, such as Greek, are limited. Taking the above into consideration, this paper presents: (i) the first Greek social media topic classification dataset; (ii) a comparative assessment of a series of traditional ML models trained on this dataset, utilizing an array of text vectorization methods including TF-IDF, classical word and transformer-based Greek embeddings; (iii) a fine-tuned GREEK-BERT-based TC model on the same dataset; (iv) key empirical findings demonstrating that transformer-based embeddings significantly increase the performance of traditional ML models, while our fine-tuned DL model outperforms previous ones. The dataset, the best-performing model, and the experimental code are made public, aiming to augment the reproducibility of this work and advance future research in the field.

GBERT: A hybrid deep learning model based on GPT-BERT for fake news detection

The digital era has expanded social exposure with easy internet access for mobile users, allowing for global communication. Now, people can get to know what is going on around the globe with just a click; however, this has also resulted in the issue of fake news. Fake news is content that pretends to be true but is actually false and is disseminated to defraud. Fake news poses a threat to harmony, politics, the economy, and public opinion. As a result, bogus news detection has become an emerging research domain to identify a given piece of text as genuine or fraudulent. In this paper, a new framework called Generative Bidirectional Encoder Representations from Transformers (GBERT) is proposed that leverages a combination of Generative pre-trained transformer (GPT) and Bidirectional Encoder Representations from Transformers (BERT) and addresses the fake news classification problem. This framework combines the best features of both cutting-edge techniques—BERT's deep contextual understanding and the generative capabilities of GPT—to create a comprehensive representation of a given text. Both GPT and BERT are fine-tuned on two real-world benchmark corpora and have attained 95.30 % accuracy, 95.13 % precision, 97.35 % sensitivity, and a 96.23 % F1 score. The statistical test results indicate the effectiveness of the fine-tuned framework for fake news detection and suggest that it can be a promising approach for eradicating this global issue of fake news in the digital landscape.

1179 Navigating the AI Landscape: Surveying the Use of AI Language Tools for Medical Portfolios

Abstract Aim Recent advances in artificial intelligence (AI) driven natural language processing have made it possible to use tools such as ChatGPT to generate pieces of text which are realistic, eloquent, and challenging to distinguish from human-generated content. This work aims to answer the question: could the use of these tools for medical portfolios represent a positive opportunity to enhance learning, or would it detract from the purpose of reflective practice? Method In total, 98 post-graduate doctors from across the UK were surveyed to gather data on their awareness of, previous use of, disclosure of use and experience with these tools for their professional portfolios. Their opinions on the need for future guidance in this area were also explored. Results From 90 respondents who currently maintain a professional portfolio, 18 (20%) had used AI-driven tools in some form to generate content for it and none disclosed this use. Regarding disclosing the use of AI-driven tools, 67.3% advocate for this whereas 32.7% believe disclosure is unnecessary. There was slightly stronger consensus regarding how useful guidance from regulatory bodies on this practice would be, with 73.5% in support of guidance and 26.5% in opposition. Conclusions Clearly, there is an urgent need for larger studies, discussion, and consensus from regulatory bodies so that guidance can be put in place. Without this, maintaining the integrity of CPD will become challenging for both trainee doctors and assessors of medical portfolios.

Is ChatGPT an Accurate and Readable Patient Aid for Third Molar Extractions?

Artificial intelligence (AI) platforms such as Chat Generative Pre-Trained Transformer (ChatGPT) (Open AI, San Francisco, California, USA) have the capacity to answer health-related questions. It remains unknown whether AI can be a patient-friendly and accurate resource regarding third molar extraction. The purpose was to determine the accuracy and readability of AI responses to common patient questions regarding third molar extraction. This is a cross sectional in-silico assessment of readability and soundness of a computer-generated report. Not applicable. Accuracy, or the ability to provide clinically correct and relevant information, was determined subjectively by 2 reviewers using a 5-point Likert scale, and objectively by comparing responses to American Association of Oral and Maxillofacial Surgeons (AAOMS) clinical consensus papers. Readability, or how easy a piece of text is to read, was assessed using the Flesch Kincaid Reading Ease (FKRE) and Flesch Kincaid Grade Level (FKGL). Both assess readability based on mean number of syllables per word, and words per sentence. To be deemed "readable", FKRE should be >60 and FKGL should be <8. Not applicable. Descriptive statistics were used to analyze the findings of this study. AI-generated responses above the recommended level for the average patient (FKRE: 52; FKGL: 10). The average Likert score was 4.36, suggesting that most responses were accurate with minor inaccuracies or missing information. AI correctly deferred to the provider in instances where no definitive answer exists. Of the responses that addressed content in AAOMS consensus papers, 18/19 responses closely aligned with them. All prompts did not provide citations or references. AI was able to provide mostly accurate responses, and content was closely aligned with AAOMS guidelines. However, responses were too complex for the average third molar extraction patient, and were deficient in citations and references. It is important for providers to educate patients on the utility of AI, and to decide whether to recommend using it for information. Ultimately, the best resource for answers is from the practitioners themselves because the AI platform lacks clinical experience.

Perception and Evaluation of Text from an Interview with a Real Person and an Interview with an Artificial Intelligence

The article presents a study of perception, understanding and evaluation of two interviews, the first one is performed with a real person (from an expert source), and the second is generated by the ChatGPT artificial intelligence (AI). The study covers 84 participants between the ages of 20 and 23, students in “Psychology”, “Bulgarian Philology”, “Slavic Philology”, and “Russian Philology” from the Sofia University “St. Kliment Ohridski”. The research design is of the “between-group” type (in this case, two-group). The first experimental group consisted of 40 subjects evaluates the interview with a real celebrity. The second experimental group consisted of 44 subjects evaluates the interview taken by an artificial intelligence. Both interviews consisted of ten identical questions and ten corresponding answers. Each of the ten answers is evaluated by the participants using a Likert Scale composed of 5-point score scales, each of which is “continuous” with 11 possible answer positions (from “0” to “10”), as follows: [1] inadequate / fully adequate; [2] non-specific (general) / completely specific; [3] non-exhaustive / fully exhaustive; [4] questionable / completely realistic; and [5] unsatisfactory / 10-completely satisfactory. A statistical analysis was made by a two-group test (Two Sample T-Test Unequal Variance). It was established that: consolidated answers (series of maximum ratings) when working with the “artificial” interview; in two of the ten questions asked, a statistically significant difference was observed in favor of the artificial text; in one of the questions, a statistically significant difference is observed in favor of the real interview; no difference in emotional impact and interest in summary assessment of the both pieces of text.

TASP: Topic-based abstractive summarization of Facebook text posts

Summarizing trending topics in large collections of Facebook posts is particularly relevant to profile social user activities and interests. However, automatically generating these summaries poses significant challenges due to the high heterogeneity of the input data, the limited fluency of extractive summaries, and the absence of abstractive summarization methods capable of handling multiple posts simultaneously. Existing abstractive models are either not suited to handle large post collections or disregard topic-level text relations. In this work, we present TASP, a novel tool for trending topic detection and summarization from English-written Facebook posts. It trains abstractive summarization models on multi-post collections by leveraging a shortlist of authoritative posts published by renowned newspapers. At inference time, TASP first creates clusters of semantically similar social posts, each one representing a distinct topic, using pre-trained transformer-based language models. Then, it generates abstractive summaries of the clusters for which authoritative information is missing. To the best of our knowledge, TASP is the first available tool suited to abstractive multi-post summarization. We test our approach on a large-scale dataset of real Facebook posts. The results show (1) The higher effectiveness of transformer-based approaches in generating topic-specific post clusters compared to traditional methods. (2) The importance of attending long pieces of text in multi-post abstractive summary generation.

A BERT-GRU Model for Measuring the Similarity of Arabic Text

Semantic Textual Similarity (STS) aims to assess the semantic similarity between two pieces of text. As a challenging task in natural language processing, various approaches for STS in high-resource languages, such as English, have been proposed. In this paper, we are concerned with STS in low resource languages such as Arabic. A baseline approach for STS is based on vector embedding of the input text and application of similarity metric on the embedding space. In this contribution, we propose a cross-encoder neural network (Cross-BERT-GRU) to handle semantic similarity of Arabic sentences that benefits from both the strong contextual understanding of BERT and the sequential modeling capabilities of GRU. The architecture begins by inputting the BERT word embeddings for each word into a GRU cell to model long-term dependencies. Then, max pooling and average pooling are applied to the hidden outputs of the GRU cell, serving as the sentence -pair encoder. Finally, a softmax layer is utilized to predict the degree of similarity. The experiment results show a Spearman correlation coefficient of around 0.9 and that Cross-BERT-GRU outperforms the other BERT models in predicting the semantic textual similarity of Arabic sentences. The experimentation results also indicate that the performance improves by integrating data augmentation techniques.

CGKPN: Cross-Graph Knowledge Propagation Network with Adaptive Connection for Reasoning-Based Machine Reading Comprehension

The task of machine reading comprehension (MRC) is to enable machine to read and understand a piece of text and then answer the corresponding question correctly. This task requires machine to not only be able to perform semantic understanding but also possess logical reasoning capabilities. Just like human reading, it involves thinking about the text from two interacting perspectives of semantics and logic. However, previous methods based on reading comprehension either consider only the logical structure of the text or only the semantic structure of the text and cannot simultaneously balance semantic understanding and logical reasoning. This single form of reasoning cannot make the machine fully understand the meaning of the text. Additionally, the issue of sparsity in composition presents a significant challenge for models that rely on graph-based reasoning. To this end, a cross-graph knowledge propagation network (CGKPN) with adaptive connection is presented to address the above issues. The model first performs self-view node embedding on the constructed logical graph and semantic graph to update the representations of the graphs. Specifically, a relevance matrix between nodes is introduced to adaptively adjust node connections in response to the challenge posed by sparse graph. Subsequently, CGKPN conducts cross-graph knowledge propagation on nodes that are identical in both graphs, effectively resolving conflicts arising from identical nodes in different views, and enabling the model to better integrate the logical and semantic relationships of the text through efficient interaction. Experiments on the two MRC datasets ReClor and LogiQA indicate the superior performance of our proposed model CGKPN compared to other existing baselines.

SUMEX: A hybrid framework for Semantic textUal siMilarity and EXplanation generation

Measuring semantic similarity between two pieces of text is a widely known problem in Natural language processing(NLP). It has many applications, such as finding similar medical notes of patients to accelerate the diagnosis process, plagiarism detection, and document clustering. Most state-of-the-art models are based on machine/deep learning and lack sufficient explanations for their results, limiting their adoption in critical domains like healthcare. This paper presents a hybrid framework SUMEX (Semantic textUal siMilarity and EXplanation generation) that uniquely combines ontology with a state-of-the-art embedding-based model for semantic textual similarity. The primary strength of the framework is that it explains its results in human-understandable natural language, which is vital in critical domains such as healthcare. Experiments have been conducted on two datasets of clinical notes using four embeddings: ScispaCy, BioWord2Vec, ClinicalBERT, and a customized Word2Vec trained on clinical notes. The SUMEX framework outperforms the embedding-based model on the benchmark datasets of ClinicalSTS by improving average precision scores by 7 % and reducing the false-positives-rate by 23 %. On the Patients Similarity Dataset, the average top-five and top-three precision scores were improved by 14% and 10%, respectively, using SUMEX. The SUMEX also generates explanations for its results in natural language. The domain experts evaluated the quality of the explanations. The results show that the generated explanations are of significantly good quality, with a score of 90 % and 93 % for measures of Completeness and Correctness, respectively. In addition, ChatGPT was also used for similarity score and generating explanations. The experiments show that the SUMEX framework performed better than the ChatGPT.

SENTIMENT TAHLIL UCHUN MATNDAN VALENTLIK,HISSIYOTVABOSHQATA’SIRCHAN HOLATLARNI ANIQLASH

Thetermsentimentanalysiscanbeusedtorefertomanydifferent, butrelated,problems.Mostcommonly,itisusedtorefertothetaskofau-tomatically determining the valence or polarity of a piece of text, wheth-er it is positive, negative, or neutral. However, more generally, it refers to determiningone’sattitudetowardsaparticulartargetortopic.Here,attitude can mean an evaluative judgment, such as positive or negative, or an emotional or affectual attitude such as frustration, joy, anger, sadness, excitement, and so on. In this article, we use ‘sentiment analysis’ to refer tothetaskofautomaticallydeterminingfeelingsfromtext,inotherwords, automatically determining valence, emotions, and other affectual states from text.

Enhancing toponym identification: Leveraging Topo-BERT and open-source data to differentiate between toponyms and extract spatial relationships

Abstract. Geoparsing, the process of linking locations within text to sets of geographic coordinates, plays an important role in the extraction and analysis of information from unstructured textual data. With the rapid growth in availability of user-generated data from online sources, there is increasing demand for reliable geoparsing methods. Central to many of these methods is the accurate identification of toponyms within text. For some applications, however, simple identification of toponyms is insufficient. Problems which require the association of a piece of text containing multiple toponyms to a singular location require a more nuanced approach. In this paper, we show that a transformer based deep learning model, is able to identify the subject toponym within a given text, and classify other toponyms in terms of their spatial relationship with the subject. We curate a dataset of text taken from Wikipedia pages representing 5252 locations, and use OpenStreetMap data to classify toponyms within the text in terms of their spatial relationship with the subject of each article. This dataset is then used to train a transformer based deep-learning model. On a human labelled test set, our model achieves an F1 score of 0.916 when identifying the subject toponym, and 0.884 and 0.793 when identifying toponyms representing parent and child locations of the subject, respectively. We also consider the more complex adjacent and crossing relationships - with the model achieving F1 scores of 0.548 and 0.704 in these categories, respectively.

The power of graphs in medicine: Introducing BioGraphSum for effective text summarization

In biomedicine, the expansive scientific literature combined with the frequent use of abbreviations, acronyms, and symbols presents considerable challenges for text processing and summarization. The Unified Medical Language System (UMLS) has been a go-to for extracting concepts and determining correlations in these studies; hence, the BioGraphSum model introduced in this study aims to reduce this UMLS dependence. Through adoption of an innovative perspective, sentences within a piece of text are graphically conceptualized as nodes, enabling the concept of “Malatya centrality” to be leveraged. This approach focuses on pinpointing influential nodes on a graph and, by analogy, the most pertinent sentences within the text for summarization. In order to evaluate the performance of the BioGraphSum approach, a corpus was curated that consisted of 450 contemporary scientific research articles available on the PubMed database, aligned with proven research methodology. The BioGraphSum model was subjected to rigorous testing against this corpus in order to demonstrate its capabilities. Preliminary results, especially in the precision-based and f-score-based ROUGE-(1–2), ROUGE-L, and ROUGE-SU metrics reported significant improvements when compared to other existing models considered state-of-the-art in text summarization.

Evaluating Arabic Emotion Recognition Task Using ChatGPT Models: A Comparative Analysis between Emotional Stimuli Prompt, Fine-Tuning, and In-Context Learning

Textual emotion recognition (TER) has significant commercial potential since it can be used as an excellent tool to monitor a brand/business reputation, understand customer satisfaction, and personalize recommendations. It is considered a natural language processing task that can be used to understand and classify emotions such as anger, happiness, and surprise being conveyed in a piece of text (product reviews, tweets, and comments). Despite the advanced development of deep learning and particularly transformer architectures, Arabic-focused models for emotion classification have not achieved satisfactory accuracy. This is mainly due to the morphological richness, agglutination, dialectal variation, and low-resource datasets of the Arabic language, as well as the unique features of user-generated text such as noisiness, shortness, and informal language. This study aims to illustrate the effectiveness of large language models on Arabic multi-label emotion classification. We evaluated GPT-3.5 Turbo and GPT-4 using three different settings: in-context learning, emotional stimuli prompt, and fine-tuning. The ultimate objective of this research paper is to determine if these LLMs, which have multilingual capabilities, could contribute to enhancing the aforementioned task and encourage its use within the context of an e-commerce environment for example. The experimental results indicated that the fine-tuned GPT-3.5 Turbo model achieved an accuracy of 62.03%, a micro-averaged F1-score of 73%, and a macro-averaged F1-score of 62%, establishing a new state-of-the-art benchmark for the task of Arabic multi-label emotion recognition.

Semi-Automated Nonresponse Detection for Open-Text Survey Data

Open-ended survey questions can enable researchers to gain insights beyond more commonly used closed-ended question formats by allowing respondents an opportunity to provide information with few constraints and in their own words. Open-ended web probes are also increasingly used to inform the design and evaluation of survey questions. However, open-ended questions are more susceptible to insufficient or irrelevant responses that can be burdensome and time-consuming to identify and remove manually, often resulting in underuse of open-ended questions and, when used, potential inclusion of poor-quality data. To address these challenges, we developed and publicly released the Semi-Automated Nonresponse Detection for Survey text (SANDS), an item nonresponse detection approach based on a Bidirectional Transformer for Language Understanding model, fine-tuned using Simple Contrastive Sentence Embedding and targeted human coding, to categorize open-ended text data as valid or likely nonresponse. This approach is powerful in that it uses natural language processing as opposed to existing nonresponse detection approaches that have relied exclusively on rules or regular expressions or used bag-of-words approaches that tend to perform less well on short pieces of text, typos, or uncommon words, often prevalent in open-text survey data. This paper presents the development of SANDS and a quantitative evaluation of its performance and potential bias using open-text responses from a series of web probes as case studies. Overall, the SANDS model performed well in identifying a dataset of likely valid results to be used for quantitative or qualitative analysis, particularly on health-related data. Developed for generalizable use and accessible to others, the SANDS model can greatly improve the efficiency of identifying inadequate and irrelevant open-text responses, offering expanded opportunities for the use of open-text data to inform question design and improve survey data quality.

A Parallel Model for Jointly Extracting Entities and Relations

Extracting relational triples from a piece of text is an essential task in knowledge graph construction. However, most existing methods either identify entities before predicting their relations, or detect relations before recognizing associated entities. This order may lead to error accumulation because once there is an error in the initial step, it will accumulate to subsequent steps. To solve this problem, we propose a parallel model for jointly extracting entities and relations, called PRE-Span, which consists of two mutually independent submodules. Specifically, candidate entities and relations are first generated by enumerating token sequences in sentences. Then, two independent submodules (Entity Extraction Module and Relation Detection Module) are designed to predict entities and relations. Finally, the predicted results of the two submodules are analyzed to select entities and relations, which are jointly decoded to obtain relational triples. The advantage of this method is that all triples can be extracted in just one step. Extensive experiments on the WebNLG*, NYT*, NYT and WebNLG datasets show that our model outperforms other baselines at 94.4%, 88.3%, 86.5% and 83.0%, respectively.

MuLX-QA: Classifying Multi-Labels and Extracting Rationale Spans in Social Media Posts

While social media platforms play an important role in our daily lives in obtaining the latest news and trends from across the globe, they are known to be prone to widespread proliferation of harmful information in different forms leading to misconceptions among the masses. Accordingly, several prior works have attempted to tag social media posts with labels/classes reflecting their veracity, sentiments, hate content, and so on. However, in order to have a convincing impact, it is important to additionally extract the post snippets on which the labelling decision is based. We call such a post snippet the rationale . These rationales significantly improve human trust and debuggability of the predictions, especially when detecting misinformation or stigmas from social media posts. These rationale spans or snippets are also helpful in post-classification social analysis, such as for finding out the target communities in hate-speech, or for understanding the arguments or concerns against the intake of vaccines. Also it is observed that a post may express multiple notions of misinformation, hate, sentiment, and the like. Thus, the task of determining (one or multiple) labels for a given piece of text, along with the text snippets explaining the rationale behind each of the identified labels is a challenging multi-label, multi-rationale classification task, which is still nascent in the literature. While transformer -based encoder-decoder generative models such as BART and T5 are well suited for the task, in this work we show how a relatively simpler encoder-only discriminative question-answering (QA) model can be effectively trained using simple template-based questions to accomplish the task. We thus propose MuLX-QA and demonstrate its utility in producing (label, rationale span) pairs in two different settings: multi-class (on the HateXplain dataset related to hate speech on social media), and multi-label (on the CAVES dataset related to COVID-19 anti-vaccine concerns). MuLX-QA outperforms heavier generative models in both settings. We also demonstrate the relative advantage of our proposed model MuLX-QA over strong baselines when trained with limited data. We perform several ablation studies, and experiments to better understand the effect of training MuLX-QA with different question prompts, and draw interesting inferences. Additionally, we show that MuLX-QA is effective on social media posts in resource-poor non-English languages as well. Finally, we perform a qualitative analysis of our model predictions and compare them with those of our strongest baseline.

Manifesto of Deep Learning Architecture for Aspect Level Sentiment Analysis to extract customer criticism

Sentiment analysis, a critical task in natural language processing, aims to automatically identify and classify the sentiment expressed in textual data. Aspect-level sentiment analysis focuses on determining sentiment at a more granular level, targeting specific aspects or features within a piece of text. In this paper, we explore various techniques for sentiment analysis, including traditional machine learning approaches and state-of-the-art deep learning models. Additionally, deep learning techniques has been utilized to identifying and extracting specific aspects from text, addressing aspect-level ambiguity, and capturing nuanced sentiments for each aspect. These datasets are valuable for conducting aspect-level sentiment analysis. In this article, we explore a language model based on pre-trained deep neural networks. This model can analyze sequences of text to classify sentiments as positive, negative, or neutral without explicit human labeling. To evaluate these models, data from Twitter's US airlines sentiment database was utilized. Experiments on this dataset reveal that the BERT, RoBERTA and DistilBERT model outperforms than the ML based model in accuracy and is more efficient in terms of training time. Notably, our findings showcase significant advancements over previous state-of-the-art methods that rely on supervised feature learning, bridging existing gaps in sentiment analysis methodologies. Our findings shed light on the advancements and challenges in sentiment analysis, offering insights for future research directions and practical applications in areas such as customer feedback analysis, social media monitoring, and opinion mining.