Text Summarization

NLP research on aligning lexical representation spaces to one another has so far focused on aligning language spaces in their entirety. However, cognitive science has long focused on a local perspective, investigating whether translation equivalents truly share the same meaning or the extent that cultural and regional influences result in meaning variations. With recent technological advances and the increasing amounts of available data, the longstanding question of cross-lingual lexical alignment can now be approached in a more data-driven manner. However, developing metrics for the task requires some methodology for comparing metric efficacy. We address this gap and present a methodology for analyzing both synthetic validations and a novel naturalistic validation using lexical gaps in the kinship domain. We further propose new metrics, hitherto unexplored on this task, based on contextualized embeddings. Our analysis spans 16 diverse languages, demonstrating that there is substantial room for improvement with the use of newer language models. Our research paves the way for more accurate and nuanced cross-lingual lexical alignment methodologies and evaluation.

Proficiency in written English is crucial for academic and professional achievement. However, the competency of non-native speakers is an issue in syntax and semantics thus the coherency of their writings may be affected. Current grammar checkers like Grammarly as well as MS Word grammar and spelling checkers are normally able to only scan for simple grammar mishaps, for example, spelling errors and hence fail to flag complicated syntactical and/or semantic inaccuracies. The current research aims at creating a machine learning approach to error identification of syntax and semantic errors in English written work to improve feedback on language concerns. The study proposes a synergistic model that integrates rule-based methods for syntax error detection whereas the semantic error identification is done by large language model (GPT). This integration is designed to overcome the limitations of existing tools, providing a more comprehensive error detection system. The proposed framework has three main stages: (1) rule-based methods for detecting syntax errors, such as tense agreement and subject-verb concord, (2) a deep learning model (GPT) for identifying semantic errors related to word choice and context, and (3) integration of results to produce a detailed, categorized error report. These results indicate high performance with 98% of accuracy which means that the model helped to perform more accurate evaluation and revealed slightly increased clarity and coherence of the evaluated samples. This hybrid model provides a rich solution for identifying syntax and semantic errors and improving the quality of comments particularly to the non-English speaking students. This framework may be useful for learners in academic contexts to help those learning English gain greater abilities.

In the field of Neuro-Linguistic Programming (NLP), this study investigates the implementation of transformer-based language models to automate the extraction of personality insights from extensive textual corpora. With the use of transformer architectures such as GPT-3 and cuttingedge natural language processing methods, the work presents an innovative approach for identifying and analysing complex linguistic patterns present in large text corpora. In order to enable the models to capture subtle expressions associated with different aspects of personality, the process entails fine-tuning pre-trained transformer models on multiple datasets with annotated personality attributes. The primary goal is to create a reliable automated personality analysis system that can yield precise and comprehensive insights. Significant objectives include training and optimizing transformer models to enable personality trait prediction from textual input that has not been seen before. The GPT-3.5 transformer-based language model and the Hugging Face Transformers library were used to implement automated personality analysis. The model obtained an 85% accuracy, 86% precision, 84% recall, and 85% F1-score on the social media dataset. This research contributes to the growing field of personality analysis by combining NLP techniques with transformer-based language models, creating new opportunities for creative applications in the comprehension and interpretation of human conduct.

Essay grading is one of the most cumbersome tasks that, so far, have depended on human scores to evaluate linguistic proficiency, coherence, and argumentation. Yet this work aims to present an alternative way of addressing the automation of essay grading based on advanced techniques of NLP. In more detail, it utilizes a hybrid deep learning model that combines BERT with LSTM networks to capture semantic understanding and sequential context in student essays. Use BERT to extract rich contextual embeddings allowing for deeper text understanding, then further process them with LSTM for retaining long dependencies while offering strong predictions. It learns subtle nuances in how essays should be graded along dimensions such as grammar, coherence, or argumentation structure by learning from a dataset of student essays graded by human experts. This approach indicates significant improvement in grading accuracy beyond traditional machine learning models and offers more reliable and scalable solutions for automated essay scoring. Thus, BERT-LSTM displays high performance in the rating task of essays with minimal or no human interference in the case of the minority class as it reduces bias and maintains consistency in diverse topics and writing styles. The research contribution will be onto this newly emerging field of educational technology, providing an innovative tool that enhances the efficiency of grading an essay fairly and, therefore, supports educators and improves learning outcomes for students.

As the volume of text data continues to grow rapidly, efficient and accurate text summarization has become crucial for managing and processing this information. Traditional summarization methods, including extractive and abstractive approaches, face challenges in maintaining relevance, fluency, and factual consistency, particularly in complex domains such as legal documents. In this research, we propose an advanced summarization framework leveraging transformer networks, specifically designed to enhance the quality of English text summaries. The framework begins with the preprocessing of text data, including lowercasing, elimination of punctuation, HTML tags, stopwords, and stemming/lemmatization to standardize the input. Text feature extraction is then performed using the Term Frequency-Inverse Document Frequency (TF-IDF) method, converting textual data into numerical representations that highlight the significance of terms within documents. The core of the framework employs a transformer model that eschews traditional RNNs in favor of multiple attention layers and positional encoding to capture global dependencies and enable parallel processing. The model's encoder-decoder structure uses self-attention mechanisms and multi-head attention layers to effectively encode the sequence and context of words, leading to high-quality abstractive summaries. The proposed architecture's effectiveness is evaluated using the BillSum dataset, which includes over 22,000 U.S. Congressional bills and human-written summaries. The results highlight the superior performance of the proposed Transformer model, which achieved the highest ROUGE-1, ROUGE-2, and ROUGE-L scores, with values of 37.03, 16.89, and 28.34, respectively and the Proposed method is implemented using python software. Our approach demonstrates significant improvements in summary accuracy and relevance compared to existing methods, making it a promising solution for automated summarization tasks across various domains.

Chatbots in the last few years have been the subject of growing interest as smart conversational agents for humancomputer interaction in various applications. Most current systems, however, have limitations like insufficient flexible voice control, limited document support, and minimal user-friendly capabilities. To overcome these drawbacks, this paper introduces a web-based chatbot system designed using Flask as the backend and JavaScript as the frontend. The system incorporates cuttingedge features like dark mode for user ease, text-to-speech with voice toggle control, selective text reading, microphone-based speech input, and document upload (PDF, DOCX, TXT) for content-based question answering. The backend utilizes the OpenRouter API to provide accurate responses, while the frontend offers an interactive and accessible interface. Experimental evaluation and user feedback emphasize that the suggested system provides greater usability, flexibility, and engagement over traditional chatbots. This research shows the possibility of integrating multi-modal capabilities with lightweight web technologies to enhance user experience in conversational AI systems.

The past two decades have seen impressive progress in a variety of areas of AI, particularly NLP, through the application of machine learning methods to a wide range of tasks. With the intensive use of deep learning methods in recent years this work has produced significant improvements in the coverage and accuracy of NLP systems in such domains as speech recognition, topic identification, semantic interpretation, and image description generation. While deep learning is opening up exciting new approaches to longstanding, difficult problems in computational linguistics, it also raises important foundational questions. Specifically, we do not have a clear formal understanding of why multi-level recursive deep neural networks achieve the success in learning and classification that they are delivering. It is also not obvious whether they should displace more traditional, logically driven methods, or be combined with them. Finally, we need to explore the extent, if any, to which both logical models and machine learning methods offer insights into the cognitive foundations of natural language. The aim of the Conference on Logic and Machine Learning in Natural Language (LAML) was to initiate a dialogue between these two approaches, where they have traditionally remained separate and in competition. It included invited talks by Marco Baroni (

The audio segmentation mismatch between training data and those seen at run-time is a major problem in direct speech translation. Indeed, while systems are usually trained on manually segmented corpora, in real use cases they are often presented with continuous audio requiring automatic (and sub-optimal) segmentation. After comparing existing techniques (VAD-based, fixed-length and hybrid segmentation methods), in this paper we propose enhanced hybrid solutions to produce better results without sacrificing latency. Through experiments on different domains and language pairs, we show that our methods outperform all the other techniques, reducing by at least 30% the gap between the traditional VADbased approach and optimal manual segmentation.

Due to the high volume of information and electronic documents on the Web, it is almost impossible for a human to study, research and analyze this volume of text. Summarizing the main idea and the major concept of the context enables the humans to read the summary of a large volume of text quickly and decide whether to further dig into details. Most of the existing summarization approaches have applied probability and statistics based techniques. But these approaches cannot achieve high accuracy. We observe that attention to the concept and the meaning of the context could greatly improve summarization accuracy, and due to the uncertainty that exists in the summarization methods, we simulate human like methods by integrating fuzzy logic with traditional statistical approaches in this study. The results of this study indicate that our approach can deal with uncertainty and achieve better results when compared with existing methods.

This paper explores abstractive and extractive Arabic text summarization using AI, employing fine-tuning and unsupervised machine learning techniques. We investigate the adaptation of pre-trained language models such as AraT5 through fine-tuning. Additionally, we explore unsupervised methods leveraging unlabeled Arabic text for generating concise and coherent summaries by utilizing different vectorizers and algorithms. The proposed models are rigorously evaluated using text-centric metrics like ROUGE . The research contributes to the development of robust Arabic summarization systems, offering culturally sensitive and contextually aware solutions. By bridging the gap between advanced AI techniques and Arabic language processing, this work fosters scalable and effective summarization in the Arabic domain.

Disasters strike communities around the world, with a reduced time-frame for warning and action leaving behind high rates of damage, mortality, and years in rebuilding efforts. For the past decade, social media has indicated a positive role in communicating before, during, and after disasters. One important question that remained un-investigated is that whether social media efficiently connect affected individuals to disaster relief agencies, and if not, how AI models can use historical data from previous disasters to facilitate information exchange between the two groups. In this study, the BERT model is first fine-tuned using historical data and then it is used to classify the tweets associated with hurricanes Dorian and Harvey based on the type of information provided; and alongside, the network between users is constructed based on the retweets and replies on Twitter. Afterwards, some network metrics are used to measure the diffusion rate of each type of disaster-motivated infor...

Nowadays, automatic multidocument text summarization systems can successfully retrieve the summary sentences from the input documents. But, it has many limitations such as inaccurate extraction to essential sentences, low coverage, poor coherence among the sentences, and redundancy. This paper introduces a new concept of timestamp approach with Naïve Bayesian Classification approach for multidocument text summarization. The timestamp provides the summary an ordered look, which achieves the coherent looking summary. It extracts the more relevant information from the multiple documents. Here, scoring strategy is also used to calculate the score for the words to obtain the word frequency. The higher linguistic quality is estimated in terms of readability and comprehensibility. In order to show the efficiency of the proposed method, this paper presents the comparison between the proposed methods with the existing MEAD algorithm. The timestamp procedure is also applied on the MEAD algori...

Speedy increase in the employ of the content of multimedia is very commonly perceived in existing generations. Video is one of the extreme exceedingly recovered data of multimedia which can provide quick fixes and immediate gratification for whole the kinds of user information willingness. Detection of Key frame from Videos and classification of videos are the necessary protocols in the complements tasks of video retrieval. The aim of this research is to compare between histogram similarity and histogram differencing for more brief key frames extraction from video stream. The suggested system for key frame extraction has three steps. Firstly, the video frames series will take and the characteristics for points of interest are elicited utilizing SUSAN detector. After eliciting interest characteristics from all video frames images, secondly, K-Means clustering technique was utilized for these features to construct the clusters with a number of interest points. Thirdly, a histogram builds for each video frame based on numbers of features in each cluster. X-axis of a histogram represents the cluster number and y-axis represents the number of features in each cluster. For key frames extraction, numbers for each cluster can be offered and a query histogram can be constructed based on entered clusters' numbers. A query histogram was matched with every video frame histogram using Manhattan distance to discover the similar histogram to query histogram. After chosen similar histogram, it can extract all key frames from that video frames. The experimental results show that the number of key frames extracted from the video is very brief when using the histogram similarity compared to the histogram differencing. For example, the number of key frames extracted from a car video is 70 when using the histogram similarity, while it is 180 when using the histogram differencing.

 Abstract Introduction; Background  Introduction Research Design and Methods; Materials and Methods, Methods  Method Preliminary Results; Early Results; Results; Result  Result Conclusion; Summary  Conclusion Table 13: Mapping of section names in scholarly articles Although section name variants in Table 13 may not be exhaustive and may not include every variation under which a particular section name appears, this mapping strategy provided a relatively good coverage for a large number of articles in the TREC Genomics collection. Table 14 indicates the results of the mapping process on the three TREC 2006 Genomics

During maintenance developers cannot read the entire code of large systems. They need a way to get a quick understanding of source code entities (such as, classes, methods, packages, etc.), so they can efficiently identify and then focus on the ones related to their task at hand. Sometimes reading just a method header or a class name does not tell enough about its purpose and meaning, while reading the entire implementation takes too long. We study a solution which mitigates the two approaches, i.e., short and accurate textual descriptions that illustrate the software entities without having to read the details of the implementation. We create such descriptions using techniques from automatic text summarization. The paper presents a study that investigates the suitability of various such techniques for generating source code summaries. The results indicate that a combination of text summarization techniques is most appropriate for source code summarization and that developers generally agree with the summaries produced.

Realistic video generation from audio input is a challenging and emerging domain in the intersection of natural language processing, computer vision, and generative modeling. The ability to automatically generate coherent and visually compelling video content from raw audio has promising applications in media creation, virtual education, assistive technologies, and entertainment. Manual video creation remains time-consuming and skill-intensive, while automated solutions often lack semantic alignment and visual realism. To address this gap, this project proposes an end-to-end intelligent pipeline that synthesizes realistic video content from audio input using Generative Adversarial Networks (GANs). The system begins by transcribing the user's audio using OpenAI's Whisper ASR model, followed by extracting meaningful textual descriptions via a language model (e.g., Groq LLaMA or OpenAI GPT). The script is used to generate key search terms for visual content retrieval, sourcing high-quality imagery from the Pexels API. Speech is generated using Edge TTS, and synchronized subtitles are created.The images are compiled into a dynamic video using MoviePy, and visual quality is further enhanced using Real-ESRGAN for super-resolution. The final output is a short, high-resolution, contextually accurate video with natural narration and relevant imagery. This work demonstrates the effectiveness of combining audio processing, NLP, GAN-based enhancement, and open content APIs to automate realistic video generation from scratch.

As the World Wide Web continues to expand, the process of identifying pertinent information within its vast volume of documents becomes increasingly challenging. This complexity necessitates the development of efficient solutions, one of which is automatic text summarization; an active research area dedicated to extracting key information from extensive text. The difficulties are further compounded when addressing multi-document text summarization, due to the diversity of topics and sheer volume of information. In response to this issue, this study introduces a novel approach, the Binary Biology Migration Algorithm for Multi-Document Summarization (BBMA-MDS). Viewing multi-document summarization as a combinatorial optimization problem, this approach leverages the biology migration algorithm to select an optimal combination of sentences. Evaluations of the proposed algorithm's performance are conducted using the ROUGE metrics, which facilitate a comparison between the automatically generated summary and the reference summary, commonly known as the 'gold standard summary'. For a comprehensive evaluation, the well-established DUC2002 and DUC2004 datasets are employed. The results demonstrate the superior performance of the BBMA-MDS approach when compared to alternative algorithms, including firefly and particle swarm optimization, as indicated by the selected metrics. This study thus contributes to the field by proposing BBMA-MDS as an effective solution for the multi-document text summarization problem

Objective: The amount of information for clinicians and clinical researchers is growing exponentially. Text summarization reduces information as an attempt to enable users to find and understand relevant source texts more quickly and effortlessly. In recent years, substantial research has been conducted to develop and evaluate various summarization techniques in the biomedical domain. The goal of this study was to systematically review recent published research on summarization of textual documents in the biomedical domain. Materials and methods: MEDLINE (2000 to October 2013), IEEE Digital Library, and the ACM digital library were searched. Investigators independently screened and abstracted studies that examined text summarization techniques in the biomedical domain. Information is derived from selected articles on five dimensions: input, purpose, output, method and evaluation. Results: Of 10,786 studies retrieved, 34 (0.3%) met the inclusion criteria. Natural language processing (17; 50%) and a hybrid technique comprising of statistical, Natural language processing and machine learning (15; 44%) were the most common summarization approaches. Most studies (28; 82%) conducted an intrinsic evaluation. Discussion: This is the first systematic review of text summarization in the biomedical domain. The study identified research gaps and provides recommendations for guiding future research on biomedical text summarization. Recent research has focused on a hybrid technique comprising statistical, language processing and machine learning techniques. Further research is needed on the application and evaluation of text summarization in real research or patient care settings.

The nanotechnologies are emerging in the field of future insulations for HVDC links. Innovative materials, usually called “nanodielectrics”, are associated with the incorporation of nanoparticles in a polymer matrix in order to improve their performances. In this work, after analyzing the dispersion of nanoparticles in the matrix with the SEM analysis, the HDPE based samples filled with nanosilica particles were electrically characterized by broadband dielectric spectroscopy, conduction current and space charge measurements. The weight %-content of the silica filler was varied as a parameter. Neat HDPE and nanofilled HDPE samples up to 4 %weight have been studied and a classification of these materials vs. their space charge build-up was done.

About the Book

This book serves as a structured and in-depth resource for engineering students, professionals, and researchers seeking to build expertise in Artificial Intelligence. Starting with core concepts in mathematics and statistics, the book gradually advances through Machine Learning (ML), Deep Learning (DL), Natural Language Processing (NLP), and Generative AI, ultimately leading to an in-depth exploration of Agentic AI and its potential impact on the future.

Key Features

Balanced Approach: Integrates theory with hands-on practice, including code snippets, exercises, and full project implementations.
Step-by-Step Learning Journey: Starts with foundational math and statistics, moving through ML, DL, NLP, Generative AI, and culminating in Agentic AI.
Future-Focused Insights: Explores emerging AI trends like Large Language Models (LLMs) and Agentic AI, emphasizing adaptability in a fast-evolving field.
From Basics to Cutting Edge: Covers content from grassroots fundamentals to advanced AI, making it a complete guide from ML to Agentic AI.

For the SERAPHIN project, we set up two assessment protocols m order to be able to more accurately assess the • quahty of abstracts -the FAN protocol and the MLffCE protocol, for which we provide the results The FAH protocol assesses the legibdtty of an abstract, independently from the source text The MLUCE protocol ls designed to allow users of automatic abstracts to assess then" quality These protocols were applied to a corpus of 27 texts which vaned m length from between three and twelve pages These texts were randomly chosen from EDF archives They include both sclenUfic and general press articles, extracts from books, and internal EDP notes The results of the FAlq protocol demonstrate the &fficulty of using surface lmgmsuc m&cators to assess the. quality of an abstract, the results of the MLUCE protocol illustrate the importance of user expectations

A new approach for textual analytics is introduced that creates a relationship structure for a document by using the relationship words in the system of basic English. Relationship words interlock concepts in sentences. The author of a document proposes certain relationships between concepts and uses relationship words in sentences to present a pattern of those relationships. In this paper, the concept of a document's relationship structure is used for automating summaries for documents. In addition to developing an automated approach to analyze, understand and present knowledge in documents, this article also considers how to validate the correctness of an artificial intelligence algorithm.

Question Answering (QA) is a specialized area in the field of Information Retrieval (IR). The QA systems are concerned with providing relevant answers in response to questions proposed in natural language. QA is therefore composed of three distinct modules, each of which has a core component beside other supplementary components. These three core components are: question classification, information retrieval, and answer extraction. Question classification plays an essential role in QA systems by classifying the submitted question according to its type. Information retrieval is very important for question answering, because if no correct answers are present in a document, no further processing could be carried out to find an answer. Finally, answer extraction aims to retrieve the answer for a question asked by the user. This survey paper provides an overview of Question-Answering and its system architecture, as well as the previous related work comparing each research against the oth...

Summarization is a technique in which the original text is summarized by selecting only what is important from the text. In today's world with the great advancement in the field of technology, huge amount of information is available online. In order to get only the necessary information from one or more sources available online, summarization is needed. By taking into the consideration the above problem, technique for hindi text summarization is proposed. Various documents related to sports, essays, news etc. which are available online were taken input for the system. The extraction of the summarized text was done on the basis of some important features of the text. It can be further used for news summary, summaries of books, etc.

A summary system comprises a subtraction of text document contents to generate a new form that delivers the essentials contents of the text documents. Due to the hassle of documents overload, getting the right information and effectivelydeveloped summaries are essential in retrieving information. Reduction of information allows users to find the information needed quickly without the need to read the full document collection. The purpose of this paper is to evaluate the performance of sentence scoring and clustering in the process of generating text summaries. Adaptive Neuro-Fuzzy Inference System (ANFIS) has been used to score the sentences and clustering were performed using K-Means and Hierarchical Clustering (HC) approaches. From the experimental findings, it was found that no improvements in the quality of the generated summaries obtained by simply performing clustering. This paper proposes some recommendations to improve text summarization.

The field of plant segmentation is of critical importance in advancing agricultural research, directly influencing plant health monitoring and yield optimisation. This study rigorously evaluates the performance of three state-of-the-art deep learning architectures – U-Net, Fully Convolutional Networks (FCN), and DeepLabV3+ – on the PlantVillage dataset, which encompasses diverse plant species and diseases. The analysis reveals that U-Net surpassed the other architectures, attaining an accuracy of 91.18% and an Intersection over Union (IoU) of 77.20%. This outcome underscores its efficacy in delineating plant structures with precision. DeepLabV3+ demonstrated balanced performance with an accuracy of 81.26% and an IoU of 68.10%, while FCN achieved moderate results with 77.02% accuracy and an IoU of 60.35%. The findings underscore the pivotal role of architecture selection in agricultural applications, offering insights for the optimization of plant health assessment systems and the advancement of precision agriculture.

Text summarisation is compressed or condensed version of any text document. Due to increasing use of digitisation, massive amount of information is available on internet. Text summarisation is an emerging alternative for users to find relative information in automated shortened versions. In this paper we propose single document summarisation technique using shallow features of sentence to generate summary. The weight of sentences is calculated by applying score of different words and sentence-based statistical features. Here, most salient sentences are selected based on weight of sentences and are put together to generate summary. This is modeled using fuzzy inference system. This approach utilises fuzzy inference and fuzzy measures to find most significant sentences. The result of our proposed method is compared with other methods using recall oriented understudy for Gisting evaluation (ROUGE-N) measures on document understanding conferences (DUC) 2002 dataset and results show that our proposed method outperforms a few baseline methods.

Requirement elicitation is an important and critical phase in the software development lifecycle, but it is exposed to inefficiencies and human error as it is dependent on manual methods such as interviews, workshops, and document analysis [1-4]. With the growing complexity of modern systems and the increasing volume of unstructured data, there is a need for innovative solutions to address these challenges. This paper explores the use of Natural Language Processing (NLP) to automate some important aspects of the requirement elicitation process by using advanced NLP techniques, such as the Named Entity Recognition (NER) framework, which aims to extract relevant business requirements from various unstructured sources, including emails, transcripts, and documents. This automation improves the accuracy and traceability of the elicited requirements by reducing the manual effort required. This paper, by taking an example of a banking use case, demonstrates how this framework helped in not only reducing the time spent on the requirement elicitation significantly but also discovering a lot of undocumented yet important requirements. This paper also highlights the challenges and importance of human oversight with this framework in regulated industries. Finally, the paper concludes by providing future direction/guidance, including how AI systems can be integrated into the core ecosystem and help with interactive requirement elicitation.

This research addresses the increasing trend of online news consumption within the Tamil-speaking community, driven by the growth of Internet technology. With the rise of online Tamil news platforms, readers can access a wide range of topics, but the absence of summarization in news articles poses challenges for efficient consumption. The research aims to employ machine learning techniques, specifically leveraging the XL-Sum model, for text summarization of Tamil news articles. The XL-Sum model comprises one million pairs of summaries and the news, annotated by professionals, covering 44 languages. Most research has focused on abstractive summarization in high-resource languages such as English, with limited work in lowresource languages like Tamil. The chosen XL-Sum model stands out as a large-scale multilingual abstractive summarization model and has been used with the custom dataset collected by the authors from selected Sri Lankan websites, consisting of 149 news and summary pairs annotated by professionals. The method involves gathering strategic data, selecting the XL-Sum pre-trained model, modifying and adapting it for Tamil, building a training dataset, retraining the model, and employing human evaluators to assess the quality of generated summaries, assigning ratings from high to medium. Key results include ROUGE scores comparing the new dataset with the existing model dataset, showing that the higher scores depend on the dataset's weight and finetuning the model. As an outcome, the authors can be able to release the new dataset for wider research community, providing news articles and their summarizations in Tamil from the Sri Lankan Community.

Text Summarization is condensing the source text into shorter form and retaining its information content and overall meaning. Punjabi text Summarization system is text extraction based summarization system which is used to summarize the Punjabi text by retaining the relevant sentences based on statistical and linguistic features of text. It comprises of two main phases: 1) Pre Processing 2) Processing. Pre Processing is structured representation of the original Punjabi text. In Processing, final score of each sentence is determined using feature-weight equation. Top ranked sentences in proper order are selected for final summary. This paper concentrates on complete pre processing phase of Punjabi text summarization system. Pre processing phase includes Punjabi words boundary identification, Punjabi sentences boundary identification, Punjabi stop words elimination, Punjabi language stemmer for nouns and proper names, allowing input in proper format and elimination of duplicate sentences.

Statistical-based collocation extraction approaches suffer from (1) low precision rate because high co-occurrence bi-grams may be syntactically unrelated and are thus not true collocations; (2) low recall rate because some true collocations with low occurrences cannot be identified successfully by statistical-based models. To integrate both syntactic rules as well as semantic knowledge into a statistical model for collocation extraction is one way to achieve a high precision while keeping a reasonable recall. This paper designs a cascade system which employs a hybrid model by integrating both syntactic and semantic knowledge into a statistical model for Chinese synonymous noun/verb collocations extraction. The grammatically bounded noun/verb collocations are extracted first from a syntactic-rule based module, which is then inputted to a semantic-based module for further retrieval of low frequent bi-gram collocations.

Natural Language Processing (NLP) has undergone a revolution with the emergence of large language models (LLMs) like GPT and BERT. Prompt engineering, the art of crafting effective inputs to guide LLMs, has become a cornerstone of modern NLP applications. This paper explores how prompt engineering is streamlining NLP processes, revolutionizing industries such as banking, and addressing the inherent risks associated with generative AI. Through practical examples and a discussion of potential pitfalls, we offer strategies for mitigating challenges and maximizing the benefits of prompt engineering in NLP.

Commented information is provided on the origins and nature of nanodielectrics, on recent developments relative to polymer nanocomposites, on future needs for cable developments and findings about nano-PE and XLPE, and preliminary results concerning PE coated with nanosilica are reported in turns, to give a feeling of the progress in the field of nanodielectrics and a sense of what could be the potentiality for HV cable applications.

An important supply chain research problem is the bullwhip effect: demand fluctuations increase as one moves up the supply chain from retailer to manufacturer. It has been recognized that demand forecasting and ordering policies are two of the key causes of the bullwhip effect. In this paper we present a spreadsheet application, which explores a series of replenishment policies and forecasting techniques under different demand patterns. It illustrates how tuning the parameters of the replenishment policy induces or reduces the bullwhip effect. Moreover, we demonstrate how bullwhip reduction (order variability dampening) may have an adverse impact on inventory holdings. Indeed, order smoothing may increase inventory fluctuations resulting in poorer customer service. As such, the spreadsheets can be used as an educational tool to gain a clear insight into the use or abuse of inventory control policies and improper forecasting in relation to the bullwhip effect and customer service.

In this paper we describe and analyze the performance of a speech summarization prototype based in classical text summarization techniques, mainly lexical chains. We show that shallow text-oriented techniques are robust enough to identify relevant information and basic units of content in spontaneous speech. Moreover, the performance recovering the most relevant segments does not drop when the input is highly noisy, as in automatic transcripts. Segments from transcripts are extracted taking into account information provided by lexical chains, shallow parsing, collocations and discourse markers. Two different approaches to segmentation are evaluated. In the PreSeg approach, as in classical textual summarizing, the segmentation is part of the preprocessing step, while in the PostSeg approach, summary text segments are found after the detection of relevant document parts.

This paper documents a submission to the shared task on scene segmentation hosted at KONVENS 2021 (Zehe et al., 2021b). The aim of this shared task was to find methods for segmenting narrative texts into different scenes -segments of text where location, time and the constellation of characters stay more or less coherent. This task is formulated as a sentence classification task where sentences bordering the scenes have to be distinguished from in-scene sentences. The approach presented in this paper is based on two steps. In the first one, a twin BERT training setup is used to learn a sentence embedding space in which sentences functioning as scene borders are well-separated from ones that are in-scene. In the second one, the sentence embeddings generated by this model are used as feature vectors to feed a gradient-boosted decision tree ensemble which conducts final predictions. In the shared task leaderboard, the system ranked second in track 1 and first in track 2.

During maintenance developers cannot read the entire code of large systems. They need a way to get a quick understanding of source code entities (such as, classes, methods, packages, etc.), so they can efficiently identify and then focus on the ones related to their task at hand. Sometimes reading just a method header or a class name does not tell enough about its purpose and meaning, while reading the entire implementation takes too long. We study a solution which mitigates the two approaches, i.e., short and accurate textual descriptions that illustrate the software entities without having to read the details of the implementation. We create such descriptions using techniques from automatic text summarization. The paper presents a study that investigates the suitability of various such techniques for generating source code summaries. The results indicate that a combination of text summarization techniques is most appropriate for source code summarization and that developers generally agree with the summaries produced.

Polymers are required to use in radiation environments as insulation materials. However, they often suffer from tracking failure. There is an increasing demand to evaluate radiation effects on dielectric performance. This paper presents a recurrence plot (RP) approach to analyse surface discharge sequences of gamma-ray irradiated polymer materials based on tracking test. Studying the non-linear characteristics of discharge sequences can assist in understanding the underlying mechanism of the discharge process. Discharge sequences of the test are extended to m-dimensional phase space by using the phase space reconstructed method. As test samples, polybutylene terephthalate (PBT) and polybutylene naphthalate (PBN) were irradiated to 100 kGy and then up to 1 MGy with a dosage rate of 10 kGy h -1 by using a 60 Co gamma source. The tracking tests were carried out according to the test method described in IEC60112. It is found that the RPs can give visual recurrent patterns of discharge sequences for identification of the effects of gamma-ray radiation dosage on the resistance to tracking of the polymers. The detection of recurrent patterns together with comparative tracking index value results indicate that with the increase in the radiation dosage, the resistance to tracking of PBT decreases, but increases for PBN.

This paper proposes a novel extractive summarization method for contact center dialogues. We use a particular type of hidden Markov model (HMM) called Class Speaker HMM (CSHMM), which processes operator/caller utterance sequences of multiple domains simultaneously to model domain-specific utterance sequences and common (domainwide) sequences at the same time. We applied the CSHMM to call summarization of transcripts in six different contact center domains and found that our method significantly outperforms competitive baselines based on the maximum coverage of important words using integer linear programming.

Abstract—This paper aims on addressing the issue of irrelevant news content and information overloading among users
by providing a personalized summarization model. To achieve
this we have developed a personalized news summarization
system which provides concise summaries of news articles by
extracting keywords according to the user preference, this process
analyses the user specific interest like preferred topics, categories
and keywords .The summarization system uses NewsAPI to
fetch real time data (i.e recent news articles) and uses trained
Natural Language Processing(NLP) that contains a hugging face
transformers library for text summarization which ensures that
the content remains up to date and relavent to the user preference
.The architecture of this system is designed to be modular and
scalable which enables integration of various components such
as data collection, preprocessing, personalization, summarization,
and keyword extraction. This system in this research is efficient, reliable and user-friendly which provides a concise and
personalised summarization with extracted keywords for user
personalized news topics. This research reduces users’ burden
by making news consumption more efficient and relevant. This
also ensures that the user gets focused, personalised and easy to
understand summaries by connecting the gap between too much
information and the need for their simplicity.
Index Terms—Personalized news summarizer, Keyword extraction, News API.

In this paper, we attempt to summarize online discussions by filtering posts. Selecting the highly related posts from the discussion boards leads to a summarized version of the discussion. Online Discussion Summarizer (ODS) is based on unsupervised information retrieval techniques. Four features are used in the summarization function; which are the term frequency inverse post frequency, title term frequency, description term frequency and author reputation. This paper shows that combining the four features in the same function results in higher accuracy than using each alone. ODS was able to summarize online discussions with an accuracy of 72%, precession of 83% and recall of 62%.

The Web 2.0 has resulted in a shift as to how users consume and interact with the information, and has introduced a wide range of new textual genres, such as reviews or microblogs, through which users comunicate, exchange, and share opinions. The explotation of all this user-generated content is of great value both for users and companies, in order to assist them in their decision-making processes. Given this context, the analysis and development of automatic methods that can help manage online information in a quicker manner are needed. Therefore, this article proposes and evaluates a novel concept-level approach for ultra-concise opinion abstractive summarization. Our approach is characterized by the integration of syntactic sentence simplification, sentence regeneration and internal concept representation into the summarization process, thus being able to generate abstractive summaries, which is one the most challenging issues for this task. In order to be able to analyze different settings for our approach, the use of the sentence regeneration module was made optional, leading to two different versions of the system (one with sentence regeneration and one without). For testing them, a corpus of 400 English texts, gathered from reviews and tweets belonging to two different domains, was used. Although both versions were shown to be reliable methods for generating this type of summaries, the results obtained indicate that the version without sentence regeneration yielded to better results, improving the results of a number of stateof-the-art systems by 9%, whereas the version with sentence regeneration proved to be more robust to noisy data.

An image captioning system involves modules on computer vision as well as natural language processing. Computer vision module is for detecting salient objects or extracting features of images and Natural Language Processing (NLP) module is for generating correct syntactic and semantic image captions. Although many image caption datasets such as Flickr8k, Flickr30k and MSCOCO are publicly available, most of the datasets are captioned in English language. There is no image caption corpus for Myanmar language. Myanmar image caption corpus is manually built as part of the Flickr8k dataset in this current work. Furthermore, a generative merge model based on Convolutional Neural Network (CNN) and Long-Short Term Memory (LSTM) is applied especially for Myanmar image captioning. Next, two conventional feature extraction models Visual Geometry Group (VGG) OxfordNet 16-layer and 19-layer are compared. The performance of this system is evaluated on Myanmar image caption corpus using BLEU score...

Bad weather can collect a lot of dust, contaminating the high voltage glass insulator, which is a significant issue in the electricity distribution field due to the high cost of the cleaning process . For that reason, academic and industrial researchers are interested in the coating technique for obtaining new superhydrophilic materials. Glass coating with a superhydrophobic nanocomposite material has a wide range of applications due to it is unique characteristics, such as a self-cleaning technique with low adhesion surface , corrosion resistance , antiicing and excellent antibacterial properties . To achieve a self-cleaning glass insulator, a coating of superhydrophobic nanocomposite materials can be applied, which has the ability to reduce not only the dust accumulation but also the complex rain salts that accumulate on its surface. The superhydrophobic nanocomposite material can be defined as a unique material that is extremely difficult to wet with a contact angle of over 150 o . Many researchers have been investigating the coating techniques of the high voltage glass insulator via the spin coating [1], the deposition coating [11], the sol-gel method , and the laser etching technique [13]. However, these techniques have low efficiencies, and the flashover can occur after a while, even using composite materials. Coating glass with new transparent superhydrophobic materials has attracted a lot of attention in recent years due to its unique properties, such as low viscosity, excellent flow-ability, high temperature resistance, and ageing resistance. Additionally, roughness and low surface energy are important factors influencing the coating technique of the superhydrophobic materials. Furthermore, the high roughness is mainly attributed to the inhomogeneity of the coating surface, which may hinder the optical clarity Abstract: A new nanocomposite superhydrophobic of the RTV (room temperature vulcanized) silicon rubber reinforced with a different percentage of nanosilica was prepared by a two-stage sol-gel route to obtain a superhydrophobic surface coating on high voltage glass insulator, preventing the dust-water droplet from adhering to its surface. The cold spraying technique was utilized to build up a thin nanocomposite superhydrophobic layer on the glass insulator containing different percentages of the nanosilica particles, such as 23 wt %, 33 wt %, and 44 wt % with RTV silicon substrate. The synthesized nanocomposite was analyzed using the contact angle, roughness, adhesion, hardness, and dielectric strength tests. Moreover, the prepared RTV silicon rubber/nanosilica superhydrophobic nanocomposite layer was characterized using the field emission scanning electron microscope (FESEM), X-ray diffraction (XRD), Fourier transform infrared spectroscopy (FTIR), and the particle size analysis test. Based on the results, the nanosilica particles were well-incorporated into the RTV silicon rubber, obtaining an excellent homogenous distribution thin layer on its surface, supporting its capability to be a superior superhydrophobic surface. The results reveal that the RTV silicon rubber/ 33wt % nanosilica was the best as a superhydrophobic behavior with a contact angle reaching higher than 158˚± 3; also, a significant change in the dielectric strength was obtained to be 25.5 kV (using a speed voltage of 5.0 kV/s). Importantly, the flashover test was also conducted, and it was found that there was a significant change in the leak current between the coated and uncoated samples. The leak current of the coated sample with a superhydrophobic nanocomposite was reduced to 2.5 mA, while the uncoated sample became 3.2 mA using a voltage load value of 60 kV. The results presented here may improve the nanocomposite material as an antiweathering superhydrophobic thin layer supported by the prepared nano-SiO 2 particles against the dust-water droplets which may be adhesive to the high voltage glass insulator.

The Sky-Hub Project is a video conferencing app designed for seamless remote communication through video, audio,
and chat. It supports virtual meetings, collaborations, and events. Key features include dynamic room creation, real-time call
quality monitoring, and adaptive video quality based on network speed. Unique tools like Video Summarization provide meeting
highlights, while Call Latency Monitoring ensures connection reliability. With a responsive design for all devices, the app offers
low-latency communication and scales for multiple users. Customizable and cost-effective, it’s a robust solution for personal,
professional, and educational use.

Je tiens tout d'abord à remercier les Profs. Mirta Gordon et Phillipe Langlais d'avoir accepté d'être les rapporteurs de cette thèse. Ils ont contribué par leurs nombreuses remarques et suggestions à améliorer la qualité de ce mémoire. Je remercie les Profs. Horacio Saggion et Eva Buchi pour participer au Jury de soutenance. Également le Prof. Daniel Malterre pour présider ce Jury. Je remercie le Consejo Nacional de Ciencia y Tecnología (CONACYT) du Mexique pour le financement de cette thèse. Aussi les laboratoires LPM de Nancy et LIA d'Avignon pour leurs supports. Je voudrais souligner les r ôles des Profs. Luis Martínez et Daniel Malterre dans le choix qui est devenu finalement mon chemin scientifique, même si, comme ils les savent bien, j'aurai préféré autrement. Je remercie spécialement le Prof. Marc El-Bèze de m'avoir si gentiment accueilli au LIA pendant deux ans et demi. Merci à mes trois directeurs de m'avoir guidé pendant la élaboration de cette thèse : à Juan Manuel Torres, je le remercie de m'avoir confié un projet si intéressante et original, de m'avoir toujours donné la liberté d'action et les outils TAL nécessaires pour développer ce travail. À Eric SanJuan, toujours gentil et humble, je le remercie d'avoir mis la main à la pâte pour améliorer ma rudimentaire fac ¸on de programmer et surtout pour les nombreuses discussions dans lesquelles nous avons d û systématiquement diviser le tableau en deux pour confronter nos différents points de vue (le prix de la pluridisciplinarité !). Les meilleures idées ont venu, bien s ûr, quand il avait de la bière et des cacahuètes ! Je tiens à remercier tout spécialement Bertrand Berche pour avoir été à mon c ôté pendant la révision, en temps record, de ce manuscrit, pour ses idées et précisions en quant les modèles de spins utilisés, et surtout pour son support sincère dans les moments difficiles et son humanité (de la vraie). Sans l'un d'entre vous, ce travail serait un arc-en-ciel sans couleurs. Je suis très reconnaissante à Patricia Velázquez, Iria Dacuna, Sonia Mandin et Fidelia Ibekwe, avec qui j'ai eu la fortune de collaborer, pour partager avec moi la richesse de leurs recherches, leurs esprits et leurs coeurs. À mes principaux relecteurs, Rémi Lavalley et Raphael Rubino, je dois tout ce qu'il est bien écrit dans ce manuscrit (le reste c'est moi !). Merci aux 3-Florians du LIA : Boudin, Pinault et Verdet ; toujours prêts à aider une mexicaine en détresse, soit pour la relecture, pour apprivoiser le linux ou pour installer des outils TAL ; mais surtout pour des choses plus sérieuses comme les dégustations de chocolat suisse, les cafés faits machine IUT ou le très attendu atelier tarte ! À tous le trois, merci d'être de très bons camarades. Aux personnes dont leur travail professionnel et gentillesse ont fait spécialmente agréable les séjours à Nancy et Avignon.

Türk milleti eşsiz bir kültürel mirasa sahiptir. Gerek sözlü gerek yazılı kaynaklardan gelen eserler eğitim-öğretimin her kademesindeki öğrenci ile buluşturulmalıdır. Bu da ancak ders kitaplarında yer alan metinler aracılığıyla" yapılacaktır. "Çalışmanın amacı ortaokul Türkçe ders kitaplarında yer alan halk edebiyatı anlatı türlerinin kullanım sıklığını tespit etmektir. Çalışmada halk edebiyatı anlatı türlerinin sınıflandırılması yapılmış, Türkçe ders kitapları incelenerek kullanılan metin türleri ortaya konmuştur. Daha sonra ise ders kitaplarındaki halk edebiyatı anlatı türlerinin kullanım sıklığı değerlendirilmiştir. "Nitel araştırma yöntemi kullanılan çalışmada elde edilen veriler doküman incelemesi" ile tablolaştırılmış ve yorumlanmıştır. Araştırmanın örneklem grubunu Türkiye'de "okutulan ortaokul Türkçe ders kitapları" oluşturmaktadır." Bu ders kitapları 5. sınıflarda (Anıttepe Yay).; 6. sınıflarda (MEB &Ata Yay.); 7. sınıflarda (MEB ve Özgün Yay.); 8. sınıflarda (MEB) tarafından hazırlanmıştır. Serbest okuma metinlerinin de dâhil edildiği toplamda 240 okuma ve dinleme metni incelenmiştir. İncelemeler sonucunda sınıflar bazında hikâye edici metin türleri içerisinde en çok yer verilen türün hikâye olduğu sonucuna ulaşılmıştır. En az yer verilen hikâye edici metin türleri ise destan ve fıkra türüdür. Türkçe ders kitaplarındaki metinler arasında türkü, mâni, ninni ve tekerleme türlerinin bulunmadığı görülmüştür. Yapılan çalışmayla Türkçe ders kitaplarında halk edebiyatı anlatı türlerine az yer verildiği görülmüştür. Verilen türlerin "dengeli bir dağılma sahip olmadığı" tespit edilmiştir.

On the basis of the above analysis, a prototype implementation of an experimental system, called importance evaluator, has been developed. This system is a Functional module of SUSY and concentrates on the importance evaluation task only. It receives in input the internal ...