Piece Of Text Research Articles

Utility-Preserving Privacy Protection of Textual Documents via Word Embeddings

A great variety of mechanisms have been proposed to protect structured databases with numerical and categorical attributes; however, little attention has been devoted to unstructured textual data. Textual data protection requires first detecting sensitive pieces of text and then masking those pieces via suppression or generalization. Current solutions rely on classifiers that can recognize a fixed set of (allegedly sensitive) named entities. Yet, such approaches fall short of providing adequate protection because in reality references to sensitive information are not limited to a predefined set of entity types, and not all the appearances of certain entity type result in disclosure. In this work we propose a more general and flexible based on the notion of word embedding. By means of word embeddings we build vectors that numerically capture the semantic relationships of the textual terms. Then we evaluate the disclosure caused by the terms on the entity to be protected according to the similarity between their vector representations. Our method also preserves the semantics (and, therefore, the utility) of the document by replacing risky terms with privacy-preserving generalizations. Empirical results show that our approach offers much more robust protection and greater utility preservation than methods based on named entity recognition.

Пространство «вечного дома» в истории текста романа М.А. Булгакова «Мастер и Маргарита»

One of the important questions in bulgakovian studies is connected with the space of “the eternal house”, peace, which the master had deserved. The author of this paper try to understand the writer’s intention based on a textual analysis of the novel. In our work, we rely on the system of editions of the novel that we have established and its main text, reflecting the last creative will of the author to the fullest extent [1]. The line “The Master – Margarita” is outlined in drafts of the 1931 novel and is developed in its second edition (1932–1936). The third (1936) and fourth (1937) editions are incomplete — they don’t have the episodes considered in this paper. Therefore, our study uses the texts of drafts of 1931, the second, fifth (last handwritten, 1937–1938) and the sixth (final, 1938–1940) editions of the novel. The draft texts are conveyed by dynamic transcription, which will make visible the process of writer’s work on the creation and allow us to see the formation of the author’s intention. Graphic conventions are used for this: a piece of text crossed out by the writer — [text]; an insert during the writing process — text; an insert crossed out — [text]; a later insert — {text}; a later insert crossed out — {text}; a conjecture — <text>; reliability of the transmitted author’s text — <sic>; the end of a page and the transition to the next one are indicated by two straight vertical lines ||.

Элементы фикциональности в медиатекстах: столкновение реального и вымышленного

The current article is dedicated to the analysis of fictionality in modern English media texts. Fictionality is a term applied in narrative theory, and traditionally associated with the belles-lettres style (i.e. fictional narrative). In brief, fictionality is an intentional use of invented stories, which is opposed to factuality. We found that, being a fiction-specific narratological category, fictionality may appear in some kinds of media texts in the shape of separate pieces of text of different length - fictional inclusions. In our study we relied on the works by W. Schmid, J. Jenette, R. Walsh, Paul Dawson, H. S. Nielsen, J. Phelan etc. The main aim of the work is to prove that narratological analysis may be applied not only to the belles-lettres style, but also to media texts, along with other types of analysis. Also, we focused on the types, features and functions of the phenomenon under study. Investigation showed that opinion articles are more prone to contain fictional inclusions than others, which accounts for the fact that this type of articles possesses many features of the belles-lettres style. The application of stylistic and narratological analysis to fictional inclusions led us to the following conclusions. Functions of fictionality in media texts may be numerous, including attracting attention, setting up a contact with the addressee, avoiding judgment, sharing opinion, discrediting of an opponent etc. Fictional inclusions may help in creating such stylistic devices as metaphor, irony, hyperbole, implication. Altogether, the devices and functions mentioned above contribute to language manipulation of the recipient. Also, textual markers of fictionality in media texts were found, among which there were such language means as the change of narrative modes, along with modal constructions, conditional mood, negation, change of register and hyperbolization. The given means help the reader identify fictional inclusions in media texts. The results of the research support the idea voiced by modern narratologists about blurring boundaries between fiction and other cultural spheres.

User Stories and Natural Language Processing: A Systematic Literature Review

<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Context:</i> User stories have been widely accepted as artifacts to capture the user requirements in agile software development. They are short pieces of texts in a semi-structured format that express requirements. Natural language processing (NLP) techniques offer a potential advantage in user story applications. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Objective:</i> Conduct a systematic literature review to capture the current state-of-the-art of NLP research on user stories. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Method:</i> The search strategy is used to obtain relevant papers from SCOPUS, ScienceDirect, IEEE Xplore, ACM Digital Library, SpringerLink, and Google Scholar. Inclusion and exclusion criteria are applied to filter the search results. We also use the forward and backward snowballing techniques to obtain more comprehensive results. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Results:</i> The search results identified 718 papers published between January 2009 to December 2020. After applying the inclusion/exclusion criteria and the snowballing technique, we identified 38 primary studies that discuss NLP techniques in user stories. Most studies used NLP techniques to extract aspects of who, what, and why from user stories. The purpose of NLP studies in user stories is broad, ranging from discovering defects, generating software artifacts, identifying the key abstraction of user stories, and tracing links between model and user stories. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Conclusion:</i> NLP can help system analysts manage user stories. Implementing NLP in user stories has many opportunities and challenges. Considering the exploration of NLP techniques and rigorous evaluation methods is required to obtain quality research. As with NLP research in general, the ability to understand a sentence’s context continues to be a challenge.

Twitter Sentiment Analysis Using Convolutional Neural Network

Opinion examination via online media stages, for example, Twitter, Facebook, YouTube, E-commerce sites like Amazon, Flipkart and so on has become a vital and testing task. We can discover the supposition about an item or a film or an assessment utilizing conclusion examination, rather than perusing every one of the audits. The presence of tweet length, spelling mistakes, contractions, and extraordinary characters – The opinion investigation task requires a non-customary methodology. Evaluate this feeling with a good or negative value, called extremity. The general feeling is frequently gathered as good, nonpartisan or negative from the indication of the extremity. In opinion examination the extremity of a sentence decides the feeling of a sentence. Assessment of a sentence can be ordered into three classes - positive, negative and unbiased. Most current online media slant order strategies judge the conclusion extremity essentially as per printed substance and disregard other data on these stages. Profound learning has arisen as a solid AI methods that learns various layers of highlights of the information and produces cutting edge expectations results. In opinion investigation, most normally utilized profound learning strategies are convolutional neural organizations (CNN) and intermittent neural organizations (RNN). CNN highlights a convolutional layer to remove data by a greater piece of text. The outcome shows that it accomplishes better exactness in Twitter feeling grouping than some of the typical strategies like SVM and Naive Bayes techniques. Supposition investigation (or assessment mining) might be a tongue preparing strategy wont to decide if information is positive, negative or neutral.

Deep Bidirectional LSTM Network Learning-Based Sentiment Analysis for Arabic Text

Abstract Sentiment analysis aims to predict sentiment polarities (positive, negative or neutral) of a given piece of text. It lies at the intersection of many fields such as Natural Language Processing (NLP), Computational Linguistics, and Data Mining. Sentiments can be expressed explicitly or implicitly. Arabic Sentiment Analysis presents a challenge undertaking due to its complexity, ambiguity, various dialects, the scarcity of resources, the morphological richness of the language, the absence of contextual information, and the absence of explicit sentiment words in an implicit piece of text. Recently, deep learning has obviously shown a great success in the field of sentiment analysis and is considered as the state-of-the-art model in Arabic Sentiment Analysis. However, the state-of-the-art accuracy for Arabic sentiment analysis still needs improvements regarding contextual information and implicit sentiment expressed in different real cases. In this paper, an efficient Bidirectional LSTM Network (BiLSTM) is investigated to enhance Arabic Sentiment Analysis, by applying Forward-Backward encapsulate contextual information from Arabic feature sequences. The experimental results on six benchmark sentiment analysis datasets demonstrate that our model achieves significant improvements over the state-of-art deep learning models and the baseline traditional machine learning methods.

The Effect of Direct-Focused and Unfocused Written Corrective Feedback Toward Freshmen Foreign Language Writing

This study aims at finding out the effect of direct-focused and direct-unfocused written corrective feedback in improving freshmen essay writing the use of preposition, article and past tense are investigated in students revision text and new pieces of essay writing. The study uses an experimental method and true-experimental design. Data collected from 60 students for 3 month. Data are analyzed by two-way ANOVAs using SPSS 21. The study found the significant effect of Direct-Focused CF. Likewise, using Direct-Unfocused CF has positive effect. Both groups outperform in revision and new pieces of writing than group without corrective feedback. In addition, there is positive interaction effect among the use of such corrective feedback and exposure. For that reason, it seems that is evidence underpinning the importance of such feedbacks and exposure in increasing students’ writing accuracy at using any grammatical items. Further study needs to be held in order to reveal the relationship or effect of students’ motivation, perception and engagement and typology of feedback in improving students writing performance.

A Systematic Literature Review of Different Machine Learning Methods on Hate Speech Detection

Hate speech is one of the most challenging problem internet is facing today. This systematic literature review examine hate speech detection problem and will be used to do an experimental approach on detecting hate speech and abusive language. This work also provide an overview of previous research, including methods, algorithms, and main features used. We use two research questions in this literature review which will be the foundation of the next experimental research. Correctly classifying a piece of text as an actual hate speech requires a lot of correctly labelled data. Most common challenges are different languages, out of vocabulary words, long range dependencies and many more.Â

A comparative study of effective approaches for Arabic sentiment analysis

Sentiment analysis (SA) is a natural language processing (NLP) application that aims to analyse and identify sentiment within a piece of text. Arabic SA started to receive more attention in the last decade with many approaches showing some effectiveness for detecting sentiment on multiple datasets. While there have been some surveys summarising some of the approaches for Arabic SA in literature, most of these approaches are reported on different datasets, which makes it difficult to identify the most effective approaches among those. In addition, those approaches do not cover the recent advances in NLP that use transformers. This paper presents a comprehensive comparative study on the most effective approaches used for Arabic sentiment analysis. We re-implement most of the existing approaches for Arabic SA and test their effectiveness on three of the most popular benchmark datasets for Arabic SA. Further, we examine the use of transformer-based language models for Arabic SA and show their superior performance compared to the existing approaches, where the best model achieves F-score scores of 0.69, 0.76, and 0.92 on the SemEval, ASTD, and ArSAS benchmark datasets. We also apply an extensive analysis of the possible reasons for failures, which show the limitations of the existing annotated Arabic SA datasets, and the challenge of sarcasm that is prominent in Arabic dialects. Finally, we highlight the main gaps in Arabic sentiment analysis research and suggest the most in-need future research directions in this area.

COMMUNICATIVE CONTEXT AND ITS ROLE IN INTERPRETING FAMILY DISCOURSE

Article describes the theoretical model of the contexts. An important role in the interpretation of discourse plays the communicative context in its broad understanding, indicating the statuses and roles of communicants, the peculiarities of their relationships, place, time, and topic of dialogue. It is not possible to evaluate the purpose of communication, its functional feature without exploring the context of discourse. A context is a piece of text that includes a unit selected for analysis that is necessary and sufficient to determine this unit. The non-apellative part of the speech act and the communicative context (macro and micro contexts) are necessary to interpret the discourse and utterance as a whole. It was determined that the priority of the communicative function performed by the appeal depends on the belonging of the appeal to a particular type of appellants of family discourse.

COMMUNICATIVE CONTEXT AND ITS ROLE IN INTERPRETING FAMILY DISCOURSE

Article describes the theoretical model of the contexts. An important role in the interpretation of discourse plays the communicative context in its broad understanding, indicating the statuses and roles of communicants, the peculiarities of their relationships, place, time, and topic of dialogue. It is not possible to evaluate the purpose of communication, its functional feature without exploring the context of discourse. A context is a piece of text that includes a unit selected for analysis that is necessary and sufficient to determine this unit. The non-apellative part of the speech act and the communicative context (macro and micro contexts) are necessary to interpret the discourse and utterance as a whole. It was determined that the priority of the communicative function performed by the appeal depends on the belonging of the appeal to a particular type of appellants of family discourse.

Comparison of Deep Learning Models and Various Text Pre-Processing Techniques for the Toxic Comments Classification

The emergence of anti-social behaviour in online environments presents a serious issue in today’s society. Automatic detection and identification of such behaviour are becoming increasingly important. Modern machine learning and natural language processing methods can provide effective tools to detect different types of anti-social behaviour from the pieces of text. In this work, we present a comparison of various deep learning models used to identify the toxic comments in the Internet discussions. Our main goal was to explore the effect of the data preparation on the model performance. As we worked with the assumption that the use of traditional pre-processing methods may lead to the loss of characteristic traits, specific for toxic content, we compared several popular deep learning and transformer language models. We aimed to analyze the influence of different pre-processing techniques and text representations including standard TF-IDF, pre-trained word embeddings and also explored currently popular transformer models. Experiments were performed on the dataset from the Kaggle Toxic Comment Classification competition, and the best performing model was compared with the similar approaches using standard metrics used in data analysis.

Semantic model to extract tips from hotel reviews

E-commerce is growing at a swift pace, and the related content on the web is exploding. This is due to the shift of a massive amount of sales and bookings to the online platform. A large number of customers now prefer e-commerce to buy products or online booking for stays. After their transactions, customers post their experiences in the form of text reviews. Further, a new customer usually goes through these reviews before making an online transaction. However, many of such reviews include less important and often redundant information. This work aims to generate short pieces of useful text (‘tip’) from the large number of reviews which portray not only the relevant and unique information but also the sentiment captured from the reviews. The main motivation to generate a set of tips is to enable new customers to differentiate between competing for similar businesses. Our Tip Extraction Algorithm builds upon the existing methods by including the sentiments captured from the reviews. The proposed algorithms also emphasize the number of reviews for similarity comparison, so that proper weight could be given to amenities or other reviews‘ content. Recommender systems do not consider most of the recent businesses due to the vastness of the number of reviews of well established businesses. We compare our proposed method with the state-of-the-art TipSelector Algorithm for hotel tip extraction, on hotel reviews obtained from the TripAdvisor website. The proposed method works well, even when the number of available reviews is very less. Experimental results show significant improvements over the current state-of-the-art.

Multi-Label Emotion Detection via Emotion-Specified Feature Extraction and Emotion Correlation Learning

Textual emotion detection is an attractive task while previous studies mainly focused on polarity or single-emotion classification. However, human expressions are complex, and multiple emotions often co-occur with non-negligible emotion correlations. In this paper, a Multi-label Emotion Detection Architecture (MEDA) is proposed to detect all associated emotions expressed in a given piece of text. MEDA is mainly composed of two modules: Multi-Channel Emotion-Specified Feature Extractor (MC-ESFE) and Emotion Correlation Learner (ECorL). MEDA captures underlying emotion-specified features through MC-ESFE module, which is composed of multiple channel-wise ESFE networks. Each channel in MC-ESFE is devoted to the feature extraction of a specified emotion from sentence-level to context-level through a hierarchical structure. With underlying features, emotion correlation learning is implemented through an emotion sequence predictor in ECorL. Furthermore, we define a new loss function: multi-label focal loss. With this loss function, the model can focus more on misclassified positive-negative emotion pairs and improve the overall performance by balancing the prediction of positive and negative emotions. The evaluation of proposed MEDA architecture is carried out on emotional corpus: RenCECps and NLPCC2018 datasets. The experimental results indicate that the proposed method can achieve better performance than state-of-the-art methods in this task.

Patient's Feedback Platform for Quality of Services via “Free Text Analysis” in Healthcare Industry

Data analysis of social media posting continues to offer a huge variety of information about the health situation faced by an individual. Social networking or social media websites provide us a wealth of information generated by users in a variety of domains, that generated information are unstructured and unlabeled and are not captured in an exceedingly systematic manner, as info generated is not humanly possible to process due to its size. One traditional way of collecting patients experience is by conducting surveys and questionnaires, as these methods ask fixed questions and are expensive to administer. In this paper, a patient feedback platform (PFP) using free text sentiment analysis is developed to computationally identify and categorize the polarity expressed in a piece of text. Six machine learning latest algorithms have been used as key evaluation for evaluating accuracy of the developed (PFP) model. Results achieved have shown 88 % accuracy on the basis of which it is recommended that developed (PFP) patient feedback platform could be used to improve E-health care services indeed.

Toward a better integration of requirements and model‐based specifications

AbstractAs of today, most specifications of technical systems still rely on requirements written in natural language. However, this approach is known to be problem‐prone, due to the inherent ambiguity of natural languages. On the other hand, fully formal or model‐based approaches seem to be out of reach in many practical cases, especially in early design phases of systems. In this article, we study how to combine in a pragmatic way natural language requirements with models. We propose to keep both formats and to link pieces of text in requirements with elements of models. In other words, corpuses of requirements are managed as hypertexts with links to models. For this approach to be fully efficient, the text of requirements is not free, but relies on controlled natural language techniques leading to a partial structuring of the text. We show that this makes it possible to design (semi)automatic verifications on requirements and models, which would be impossible with unconstrained natural language. We illustrate here our approach on a small illustrative example and we report results obtained on a full size industrial application.

How to Motivate Students to Study Online

Covid-19 pandemic has caused an unprecedented change in education. All the schools in Israel and other countries had to find all the possible solutions to convert their teaching and learning to remote online options. According to the research, distant learning with the help of online boards is supposed to catch students’ attention and make them interested and motivated (Newton, 2020). You can do it with the help of such smart boards as Zoom, Miro, Wakelet, Socrative and others. They provide teachers and their students with a wide range of eye-catching materials, such as highlighted pieces of text, pictures, videos which enhance general understanding of the material and enlarge students’ passive and active vocabulary. Moreover, kids can create their materials themselves and share them with their classmates. Sometimes it is the only way to keep them motivated and to prevent them from leaving “boring online lessons”. It also has a great educational effect on those students who struggle while using traditional textbooks and misunderstand the tasks. Besides, online Learning Technology (LT) helps them to develop the skills of collaboration, negotiation and leadership which is especially important in international communication.

Neural Co-training for Sentiment Classification with Product Attributes

Sentiment classification aims to detect polarity from a piece of text. The polarity is usually positive or negative, and the text genre is usually product review. The challenges of sentiment classification are that it is hard to capture semantic of reviews, and the labeled data is hard to annotate. Therefore, we propose neural co-training to learn the semantic representation of each review using the neural network model, and learn the information from unlabeled data using a co-training framework. In particular, we use the attention-based bi-directional Gated Recurrent Unit (Att-BiGRU) to model the semantic content of each review and regard different categories of the target product as different views. We then use a co-training framework to learn and predict the unlabeled reviews with different views. Experiment results with the Yelp dataset demonstrate the effectiveness of our approach.

Enhancing Deep Learning-Based Multi-label Text Classification with Capsule Network

Given a piece of text, multi-label text classification (MLTC) is designed to mark the most relevant one label or multiple labels for the text. Most of the existing MLTC models use convolutional neural network (CNN) as feature extractor, but CNN will lose information when dealing with MLTC task. In this paper, we explore the CNN combined with capsule network for MLTC. We use capsule network instead of pool layer in CNN to extract information related to classification results in high-dimensional features. We also explore the way of combining recurrent neural network (RNN) and CNN to model the characteristics of time and space for capsule network to complete classification. In two open MLTC datasets, our model achieves the better results as the baseline system, which shows the effectiveness of the combination of capsule network and CNN for MLTC.

Deep or simple models for semantic tagging?

Semantic tagging, which has extensive applications in text mining, predicts whether a given piece of text conveys the meaning of a given semantic tag. The problem of semantic tagging is largely solved with supervised learning and today, deep learning models are widely perceived to be better for semantic tagging. However, there is no comprehensive study supporting the popular belief. Practitioners often have to train different types of models for each semantic tagging task to identify the best model. This process is both expensive and inefficient. We embark on a systematic study to investigate the following question: Are deep models the best performing model for all semantic tagging tasks? To answer this question, we compare deep models against "simple models" over datasets with varying characteristics. Specifically, we select three prevalent deep models (i.e. CNN, LSTM, and BERT) and two simple models (i.e. LR and SVM), and compare their performance on the semantic tagging task over 21 datasets. Results show that the size, the label ratio, and the label cleanliness of a dataset significantly impact the quality of semantic tagging. Simple models achieve similar tagging quality to deep models on large datasets, but the runtime of simple models is much shorter. Moreover, simple models can achieve better tagging quality than deep models when targeting datasets show worse label cleanliness and/or more severe imbalance. Based on these findings, our study can systematically guide practitioners in selecting the right learning model for their semantic tagging task.