Text Analysis Tasks Research Articles

ChatGPT for complex text evaluation tasks

AbstractChatGPT and other large language models (LLMs) have been successful at natural and computer language processing tasks with varying degrees of complexity. This brief communication summarizes the lessons learned from a series of investigations into its use for the complex text analysis task of research quality evaluation. In summary, ChatGPT is very good at understanding and carrying out complex text processing tasks in the sense of producing plausible responses with minimum input from the researcher. Nevertheless, its outputs require systematic testing to assess their value because they can be misleading. In contrast to simple tasks, the outputs from complex tasks are highly varied and better results can be obtained by repeating the prompts multiple times in different sessions and averaging the ChatGPT outputs. Varying ChatGPT's configuration parameters from their defaults does not seem to be useful, except for the length of the output requested.

Importance and Didactic Conditions of Teaching the Pedagogical Heritage of Beyimbet Maylin

The article analyzes the pedagogical heritage of Beyimbet Maylin, which has not yet entered scientific circulation, and has not been noticed by methodologists-scientists, and their importance in the development of domestic science is evaluated. For the first time, four educational books written by Beyimbet Maylin, a prominent writer, who have their own place in Kazakh literature, are being studied. In the study, the ways of studying the history of the writing of textbooks left behind by the writer, introducing their content-structural system to future philology teachers were differentiated. Providing information about the history of the formation of domestic textbook science, the possibilities of forming professional-cognitive and research competences of students are analyzed. The main idea in the research work is that students need to effectively use new educational technologies and interactive teaching methods during the analysis of historical heritage, because this, in turn, expands the opportunities for improving the quality of professional education of students. The scientific importance of the article is demonstrated by the unified methodological system of applied teaching and the analysis of educational materials and the effectiveness of using their selection criteria in the organization of students' project work, in the performance of complex text analysis tasks. This proves the innovativeness of the proposed research work, dedicated to a new topic. The article analyzes the effective ways of presenting textbooks written by Beyimbet Maylin to philology students, systematizes the technological conditions. During the pedagogical experiment, it is planned to develop students' professional qualifications related to theoretical and historical-pedagogical heritage, critical thinking skills related to educational materials. They are followed by analyzing the progress and results of work included in one module of a special elective subject. Such complexity ensures the integrity of the research results presented in the article and the reliability of the proposed principles. Historical-pedagogical, system-typological, genre-style analyzes were used in the research work in conjunction with teaching technologies in accordance with modern requirements. The results of the study fully demonstrated the effectiveness of this unit.

XGBoost-Enhanced Graph Neural Networks: A New Architecture for Heterogeneous Tabular Data

Graph neural networks (GNNs) perform well in text analysis tasks. Their unique structure allows them to capture complex patterns and dependencies in text, making them ideal for processing natural language tasks. At the same time, XGBoost (version 1.6.2.) outperforms other machine learning methods on heterogeneous tabular data. However, traditional graph neural networks mainly study isomorphic and sparse data features. Therefore, when dealing with tabular data, traditional graph neural networks encounter challenges such as data structure mismatch, feature selection, and processing difficulties. To solve these problems, we propose a novel architecture, XGNN, which combines the advantages of XGBoost and GNNs to deal with heterogeneous features and graph structures. In this paper, we use GAT for our graph neural network model. We can train XGBoost and GNN end-to-end to fit and adjust the new tree in XGBoost based on the gradient information from the GNN. Extensive experiments on node prediction and node classification tasks demonstrate that the performance of our proposed new model is significantly improved for both prediction and classification tasks and performs particularly well on heterogeneous tabular data.

A comparative study of lexical bundles in linguistics and biology Ph.D. dissertations

Lexical bundles are considered as one of the main rhetorical features in academic genres. The appropriate use of these features adds to the coherence and naturalness of the texts. Recently, genre analysis studies tend to investigate the similarities and differences across different academic disciplines and the effects of the first language of writers on the employment of these linguistic features. There are limited studies, however, in the field with a mere focus on the use of lexical bundles in PhD dissertations and the similarities and differences on the employment of these features among native and non-native English writers. In this regard, the current study, following a comparative corpus-based approach, investigated the use of lexical bundles in English PhD dissertations written by native English-speaking and nonnative Kurdish-speaking writers across the two disciplines of biology and linguistics. All the compiled dissertations were selected from the ones published between 2010 to 2020 in British universities to keep the compiled corpora comparable. In the next phase, the distribution, linguistic structures, and functions of bundles used in introduction and literature review sections of the compiled dissertations were analyzed using WordSmith 6th edition concordancing software. The analysis provides a list of the most frequently used lexical bundles in each scientific field and among the two groups of writers. The structural analysis showed that noun phrases and prepositional phrases were the most frequently used bundles between both groups of writers and of disciplines. It was also found that passive structures were commonly used in biology dissertations. The functional analysis revealed that non-native Kurdish writers tended to use more research-oriented and text-oriented bundles in comparison with their native counterparts. It was suggested that the rhetorical differences and similarities between the two groups of writers and the disciplines could be attributed to the experience or proficiency levels of writers and the conventions, knowledge construction principles and research approaches of each academic discourse community. Pedagogically speaking, it is discussed that teaching discursive features and including language teaching tasks drawn from authentic texts such as text analysis tasks, focused tasks, extended writing tasks enable EAP learners to learn how to employ these linguistic features effectively to improve the flow of information, coherence and fluency in their academic texts.

Graph Neural Network and BERT Model for Antimalarial Drug Predictions Using Plasmodium Potential Targets

Malaria continues to pose a significant global health burden despite concerted efforts to combat it. In 2020, nearly half of the world’s population faced the risk of malaria, underscoring the urgency of innovative strategies to tackle this pervasive threat. One of the major challenges lies in the emergence of the resistance of parasites to existing antimalarial drugs. This challenge necessitates the discovery of new, effective treatments capable of combating the Plasmodium parasite at various stages of its life cycle. Advanced computational approaches have been utilized to accelerate drug development, playing a crucial role in every stage of the drug discovery and development process. We have witnessed impressive and groundbreaking achievements, with GNNs applied to graph data and BERT from transformers across diverse NLP text analysis tasks. In this study, to facilitate a more efficient and effective approach, we proposed the integration of an NLP based model for SMILES (i.e., BERT) and a GNN model (i.e., RGCN) to predict the effect of antimalarial drugs against Plasmodium. The GNN model was trained using designed antimalarial drug and potential target (i.e., PfAcAS, F/GGPPS, and PfMAGL) graph-structured data with nodes representing antimalarial drugs and potential targets, and edges representing relationships between them. The performance of BERT-RGCN was further compared with that of Mordred-RGCN to evaluate its effectiveness. The BERT-RGCN and Mordred-RGCN models performed consistently well across different feature combinations, showcasing high accuracy, sensitivity, specificity, MCC, AUROC, and AUPRC values. These results suggest the effectiveness of the models in predicting antimalarial drugs against Plasmodium falciparum in various scenarios based on different sets of features of drugs and potential antimalarial targets.

Revolutionizing Healthcare: NLP, Deep Learning, and WSN Solutions for Managing the COVID-19 Crisis

The COVID-19 outbreak in 2020 catalyzed a global socio-economic upheaval, compelling nations to embrace digital technologies as a means of countering economic downturns and ensuring efficient communication systems. This paper delves into the role of Natural Language Processing (NLP) in harnessing wireless connectivity during the pandemic. The examination assesses how wireless networks have affected various facets of crisis management, including virus tracking, optimizing healthcare, facilitating remote education, and enabling unified communications. Additionally, the article underscores the importance of digital inclusion in mitigating disease outbreaks and reconnecting marginalized communities. To address these challenges, a Dual CNN-based BERT model is proposed. BERT model is used to extract the text features, the internal layers of BERT excel at capturing intricate contextual details concerning words and phrases, rendering them highly valuable as features for a wide array of text analysis tasks. The significance of dual CNN is capturing the unique capability to seamlessly integrate both character-level and word-level information. This fusion of insights from different levels of textual analysis proves especially valuable in handling text data that is noisy, complex, or presents challenges related to misspellings and domain-specific terminology. The proposed model is evaluated using the simulated WSN-based text data for crisis management.

An improved Bulgarian natural language processing pipeline

In this paper, we present a language pipeline for processing Bulgarian language data. The pipeline consists of the following steps: tokenization, sentence splitting, part-of-speech tagging, dependency parsing, named entity recognition, lemmatization, and word sense disambiguation. The first two components are based on rules and lists of words specific to the Bulgarian language, while the rest of the components use machine learning algorithms trained on universal dependency data and pretrained word vectors. The pipeline is implemented in the Python library spaCy (https://spacy.io/) and achieves significant results on all the included subtasks. The pipeline is open source and is available on Github (https://github.com/melaniab/spacy-pipeline-bg/) for use by researchers and developers for a variety of natural language processing and text analysis tasks.

A Comparison of LSTM and BERT model for sarcasm prediction

Sarcasm prediction is a text analysis task that aims to identify sarcastic and non-sarcastic statements in text. Sarcasm is a figure of speech that uses opposite or contradictory language to express a certain meaning or idea. Sarcasm is usually cryptic, vague, and suggestive, which makes sarcasm prediction a challenging task. In sarcasm prediction projects, techniques of natural language processing are usually leveraged to analyze and classify the text. The main challenge of this task lies in the fact that sarcasm usually has multiple manifestations and needs to consider the contextual and semantic information of the text. The prediction of sarcasm holds significant application value in natural language processing, such as social media analysis, public opinion monitoring, sentiment analysis and so on. In this paper, by controlling variables, the influence of adding the long short-term memory (LSTM) layer and changing the grid structure of the model on the accuracy of prediction results is explored. Moreover, accuracy of the LSTM prediction performance is compared with that of the bidirectional encoder representations from Transformers (BERT) model. At the same time, this paper analyzed and discussed the phenomenon that adding the number of LSTM model layers could not obtain higher prediction accuracy, and the accuracy gap of prediction results between LSTM model and BERT model, and finally obtained relevant conclusions.

The hypergeometric test performs comparably to TF-IDF on standard text analysis tasks

The hypergeometric test performs comparably to TF-IDF on standard text analysis tasks

Влияние установки на чтение на стратегию чтения цифрового текста у учащихся старшего школьного возраста: результаты айтрекингового исследования

<p style="text-align: right;">The paper aims to investigate the correlation between reading setup and digital reading strategies used by Russian schoolchildren. To study the influence of the reading setup factor, we conducted an eye-tracking study of high school students (N=66), which showed the dependence of digital reading strategies on the pre-reading task by experienced readers. The study showed that the pattern of skimming digital texts does not always appear explicitly, and revealed a significant influence on reading strategies of both reader's individual differences and the rhetorical organization of the text, as well as the reading task factor. The task of searching for specific information led to a more thorough primary reading of the entire text with fewer returns and transitions between the other semantic elements of the digital educational text. In the case of the text analysis task, the first reading was more cursory, but there was more jumping back and forth between paragraphs to establish the logical linkage of the text information. The influence of the reading setup on the digital reading strategy illustrates the importance of formulating the reading task correctly in order to ensure an effective digital reading experience.</p>

Knowledge-Fusion-Based Iterative Graph Structure Learning Framework for Implicit Sentiment Identification.

Implicit sentiment identification is a significant classical task in text analysis. Graph neural networks (GNNs) have recently been successful in implicit sentiment identification, but the current approaches still suffer from two problems. On the one hand, there is a lack of structural information carried by the single-view graph structure of implicit sentiment texts to accurately capture obscure sentiment expressions. On the other hand, the predefined fixed graph structure may contain some noisy edges that cannot represent semantic information using an accurate topology, which can seriously impair the performance of implicit sentiment analysis. To address these problems, we introduce a knowledge-fusion-based iterative graph structure learning framework (KIG). Specifically, for the first problem, KIG constructs graph structures based on three views, namely, co-occurrence statistics, cosine similarity, and syntactic dependency trees through prior knowledge, which provides rich multi-source information for implicit sentiment analysis and facilitates the capture of implicit obscure sentiment expressions. To address the second problem, KIG innovatively iterates the three original graph structures and searches for their implicit graph structures to better fit the data themselves to optimize the downstream implicit sentiment analysis task. We compared our method with the mainstream implicit sentiment identification methods on two publicly available datasets, and ours outperformed both benchmark models. The accuracy, recall, and F1 values of KIG on the Pun of the Day dataset reached 89.2%, 93.7%, and 91.1%, respectively. Extensive experimental results demonstrate the superiority of our proposed method for the implicit sentiment identification task.

Social Media Opinion Analysis Model Based on Fusion of Text and Structural Features

The opinion recognition for comments in Internet media is a new task in text analysis. It takes comment statements as the research object, by learning the opinion tendency in the original text with annotation, and then performing opinion tendency recognition on the unannotated statements. However, due to the uncertainty of NLP (natural language processing) in short scenes and the complexity of Chinese text, existing methods have some limitations in accuracy and application scenarios. In this paper, we propose an opinion tendency recognition model HGAT (heterogeneous graph attention network) that integrates text vector and context structure methods to address the above problems. This method first trains a text vectorization model based on annotation text content, then constructs an isomorphic graph with annotation, news, and theme as its apex, and then optimizes the feature vectors of all nodes using an isomorphic graph neural network model with attention mechanism. In addition, this article collected 1,684,318 news items and 57,845,091 comments based on Toutiao, sifted through 511 of those stories and their corresponding 103,787 comments, and tested the impact of HGAT on this dataset. Experiments show that this method has stable improvement effect on different NLP methods, increasing accuracy by 2–10%, and provides a new perspective for opinion tendency recognition.

Building a text retrieval system for the Sanskrit language: Exploring indexing, stemming, and searching issues

Stemming is an important pre-processing step in the text analysis domains such as text mining, text summarization and information retrieval (IR). In this study, we build a Sanskrit text collection and explore different indexing, stemming and searching strategies in Sanskrit. We also propose two stemmers: a ‘light’ and an ‘aggressive’ and evaluate their effectiveness in the text analysis task. The performance of the stemmers is evaluated in two ways: a direct and an indirect IR-based evaluation. In direct evaluation, we found that the stemmers are effective. In indirect evaluation, we apply different retrieval models such as BM25, TF–IDF, Divergence from Randomness (DFR) based and language models. The proposed stemmers are compared with GRAS stemmer, language-independent indexing (trunc-n) and no stemming approach. Among different stemming methods, aggressive stemmers provide the best performance. Hiemstra language model outperforms other retrieval models we experimented with. In statistical analysis, we found that the proposed stemming approaches produce significantly better results than the no-stemming approach.

Application of text mining technology for comparative analysis of trends in the labor market

Object: The main purpose of the article is to analyze the employment of the population in the post-Soviet space in the context of world events to identify positive and negative trends in employment, as well as characteristic trajectories of development directions.Methods: Modern methods of word processing, in particular text mining, word cloud were used.Findings: The rapid development of modern technologies, Internet applications is accompanied by the generation of large amounts of data, the timely processing of which is today one of the main problems in various spheres of life - social, economic, political, and others. In solving this global problem, modern methods of processing text information, the so-called text mining technologies, come to the rescue. These tools allow to increase the efficiency of solving problems of different levels. The algorithms embedded in the text mining technology reveal the basic concepts of the text, the content and the relationship between them.The integration of modern text mining systems and the R-Studio programming language makes it possible to conduct research in the field of text analysis and processing. These systems, using statistical methods, process the rating of news documents, materials of scientific documents, blogs, tweets, emails, advertisements and other information. The main task of text analysis is to get a clear idea about the topics of interest, to extract important information. The analysis of text documents by text mining methods is carried out in several stages: information search, text preprocessing, extraction of the required information, application of text methods, analysis and interpretation of the obtained results. For the analysis of texts, articles in Russian in PDF format were selected, including information on trends in the labor market and employment for 1995–2020.Conclusions: over the past 20 years there have been significant changes in the labor market and employment. Text analysis technologies made it possible to reveal that during the study period, the labor market issues of unemployment, employment, employment transformation, the emergence of new forms of employment, social and gender problems, and others raised.

PharmKE: Knowledge Extraction Platform for Pharmaceutical Texts Using Transfer Learning

Even though named entity recognition (NER) has seen tremendous development in recent years, some domain-specific use-cases still require tagging of unique entities, which is not well handled by pre-trained models. Solutions based on enhancing pre-trained models or creating new ones are efficient, but creating reliable labeled training for them to learn on is still challenging. In this paper, we introduce PharmKE, a text analysis platform tailored to the pharmaceutical industry that uses deep learning at several stages to perform an in-depth semantic analysis of relevant publications. The proposed methodology is used to produce reliably labeled datasets leveraging cutting-edge transfer learning, which are later used to train models for specific entity labeling tasks. By building models for the well-known text-processing libraries spaCy and AllenNLP, this technique is used to find Pharmaceutical Organizations and Drugs in texts from the pharmaceutical domain. The PharmKE platform also incorporates the NER findings to resolve co-references of entities and examine the semantic linkages in each phrase, creating a foundation for further text analysis tasks, such as fact extraction and question answering. Additionally, the knowledge graph created by DBpedia Spotlight for a specific pharmaceutical text is expanded using the identified entities. The obtained results with the proposed methodology result in about a 96% F1-score on the NER tasks, which is up to 2% better than those of the fine-tuned BERT and BioBERT models developed using the same dataset. The ultimate benefits of the platform are that pharmaceutical domain specialists may more easily identify the knowledge extracted from the input texts thanks to the platform’s visualization of the model findings. Likewise, the proposed techniques can be integrated into mobile and pervasive systems to give patients more relevant and comprehensive information from scanned medication guides. Similarly, it can provide preliminary insights to patients and even medical personnel on whether a drug from a different vendor is compatible with the patient’s prescription medication.

A Scene-Text Synthesis Engine Achieved Through Learning from Decomposed Real-World Data.

Scene-text image synthesis techniques that aim to naturally compose text instances on background scene images are very appealing for training deep neural networks due to their ability to provide accurate and comprehensive annotation information. Prior studies have explored generating synthetic text images on two-dimensional and three-dimensional surfaces using rules derived from real-world observations. Some of these studies have proposed generating scene-text images through learning; however, owing to the absence of a suitable training dataset, unsupervised frameworks have been explored to learn from existing real-world data, which might not yield reliable performance. To ease this dilemma and facilitate research on learning-based scene text synthesis, we introduce DecompST, a real-world dataset prepared from some public benchmarks, containing three types of annotations: quadrilateral-level BBoxes, stroke-level text masks, and text-erased images. Leveraging the DecompST dataset, we propose a Learning-Based Text Synthesis engine (LBTS) that includes a text location proposal network (TLPNet) and a text appearance adaptation network (TAANet). TLPNet first predicts the suitable regions for text embedding, after which TAANet adaptively adjusts the geometry and color of the text instance to match the background context. After training, those networks can be integrated and utilized to generate the synthetic dataset for scene text analysis tasks. Comprehensive experiments were conducted to validate the effectiveness of the proposed LBTS along with existing methods, and the experimental results indicate the proposed LBTS can generate better pretraining data for scene text detectors. Our dataset and code are made available at: https://github.com/iiclab/DecompST.

VictimFinder: Harvesting rescue requests in disaster response from social media with BERT

Social media platforms are playing increasingly critical roles in disaster response and rescue operations. During emergencies, users can post rescue requests along with their addresses on social media, while volunteers can search for those messages and send help. However, efficiently leveraging social media in rescue operations remains challenging because of the lack of tools to identify rescue request messages on social media automatically and rapidly. Analyzing social media data, such as Twitter data, relies heavily on Natural Language Processing (NLP) algorithms to extract information from texts. The introduction of bidirectional transformers models, such as the Bidirectional Encoder Representations from Transformers (BERT) model, has significantly outperformed previous NLP models in numerous text analysis tasks, providing new opportunities to precisely understand and classify social media data for diverse applications. This study developed and compared ten VictimFinder models for identifying rescue request tweets, three based on milestone NLP algorithms and seven BERT-based. A total of 3191 manually labeled disaster-related tweets posted during 2017 Hurricane Harvey were used as the training and testing datasets. We evaluated the performance of each model by classification accuracy, computation cost, and model stability. Experiment results show that all BERT-based models have significantly increased the accuracy of categorizing rescue-related tweets. The best model for identifying rescue request tweets is a customized BERT-based model with a Convolutional Neural Network (CNN) classifier. Its F1-score is 0.919, which outperforms the baseline model by 10.6%. The developed models can promote social media use for rescue operations in future disaster events.

Formation of a system of intelligent analysis of geological data based on ontological engineering

This article discusses the relevance and options for using the tools of ontological modeling, in relation to the objects of the mineral resource complex. Currently available geological information resources are consistently converted into electronic form, which creates a significant array of electronic scanned copies of documents that must first be translated into electronic text (recognition operation), and then analyze the electronic text for the presence of geological, economic, geological and technical parameters relevant for making managerial decisions in the field of subsoil use. The solution of the text recognition problem is relevant for quite “old” geological information resources, while the task of text analysis is relevant for each of them. One of the options for forming the semantics of interaction is ontology. It has a significant potential for application in various fields of knowledge, including in the field of geology and subsoil use. Ontologies allow integrating disparate data in order to form their thesaurus, which is especially necessary in geology, where the ambiguity of terminology interpretations creates significant barriers in the formation of ontologies of geological information resources. This problem is aggravated by a decrease in the number of specialized specialists in geological exploration organizations. The developed ontology is designed to form a knowledge base in the structure of geological information resources as the basis of a management decision support system in this industry.

SDTM: A Supervised Bayesian Deep Topic Model for Text Analytics

This study proposes a novel supervised deep topic modeling approach for effective text analysis. This approach leverages the auxiliary data associated with text, such as ratings in consumer reviews or categories of posts in online forums, to enhance the discovery of latent topics in text. The proposed approach can effectively improve topic modeling performance in several ways. First, the learned latent topics are more meaningful and distinguishable, which helps text data exploration. Second, the latent topics discovered by the novel supervised deep topic model are more accurate, which improves the performance of downstream econometrics and predictive analytics that utilize latent topics as inputs. Given the prevalence of auxiliary data in real-world text analysis tasks and the wide adoption of topic modeling in business research and practice, the study offers an effective solution for extracting insights from text data.

Deep Learning and Machine Learning-Based Model for Conversational Sentiment Classification

In the current era of the internet, people use online media for conversation, discussion, chatting, and other similar purposes. Analysis of such material where more than one person is involved has a spate challenge as compared to other text analysis tasks. There are several approaches to identify users’ emotions from the conversational text for the English language, however regional or low resource languages have been neglected. The Urdu language is one of them and despite being used by millions of users across the globe, with the best of our knowledge there exists no work on dialogue analysis in the Urdu language. Therefore, in this paper, we have proposed a model which utilizes deep learning and machine learning approaches for the classification of users’ emotions from the text. To accomplish this task, we have first created a dataset for the Urdu language with the help of existing English language datasets for dialogue analysis. After that, we have preprocessed the data and selected dialogues with common emotions. Once the dataset is prepared, we have used different deep learning and machine learning techniques for the classification of emotion. We have tuned the algorithms according to the Urdu language datasets. The experimental evaluation has shown encouraging results with 67% accuracy for the Urdu dialogue datasets, more than 10, 000 dialogues are classified into five emotions i.e., joy, fear, anger, sadness, and neutral. We believe that this is the first effort for emotion detection from the conversational text in the Urdu language domain.