Search

The use of the learning is to perform the text summarization capabilities of DistilBERT algorithm and compare it with the BERT algorithm and analyze the accuracy and the performance metrics of both the Algorithms. Text Summarization and Question - Answer generation is performed by using the pre trained models in the famous python transformers library. The corpus of at least 10-20 pages is taken and the operations are performed on that corpus. In the same way, 30 sets of the corpus are taken and they are divided into two groups. The two groups were BERT and DistilBERT algorithms. A comparison was made between the two algorithms' performance in terms of speed and accuracy. After calculating the means, standard deviations, as well as standard error means, we used the T-test for independent samples to see whether there was a statistically significant distinction between the groups. The significance level of P 0.001 (2-tailed) indicates a statistically significant difference between DistilBERT and BERT. The study proved that DistilBERT achieved better accuracy in a shorter amount of time compared to the BERT algorithm. Study shows that DistilBERT is more performant than BERT with fewer parameters. [ABSTRACT FROM AUTHOR]

Individuals in Academia have a critical requirement for several handy tools to facilitate learning, pedagogy, and collaboration. Offering frequently needed tools in one integrated destination to students, researchers, and professors alike can help fulfill this requirement. We propose the usage of state-of-the-art language and vision models to deploy this as a user interface containing three modules: technical text summarization, optical character recognition (OCR), and text-to-audio. The first module generates an effective summary, a set of keywords, and recommends similar publications. The second module splits any uploaded PDF into its constituent pages, parses the lines of text on them with image segmentation models, and outputs the text recognized using transformer architecture for further processing. Finally, the third module takes in a body of text and generates an audio file with user-set customizability. These modules work sequentially to provide an all-in-one streamlined solution to academics working with textual manuscripts. [ABSTRACT FROM AUTHOR]

This paper outlines methods for text summarizing on the text file and voice audio file conversion. In the first scenario, here converted the audio files to text format using Python tools. Text summarization is done for the latter scenario using modules from Natural Language Processing. The English data functions employ a Python framework called SpaCy. Important sentences discovered during the investigation of the extraction are used in the summarization approach. The number of times a word appears in the text file determines how much weight is given to that word. Using the main audio recording as a starting point, this method is used to create summaries. [ABSTRACT FROM AUTHOR]

The present expansion of unstructured, text-based data in the digital environment necessitates the development of automatic text summary systems that allow users to easily extract insights from them. We now have instant access to massive amounts of information. However, most of this data is superfluous, inconsequential, and may not convey the intended message. To locate the information needed from an online news piece, the user may have to sift through the content and spend a significant amount of time removing extraneous stuff. As a result, using computerized text summarizers that can extract vital information while removing inessential and unnecessary material is becoming increasingly crucial. Implementing summarizing can make documents simpler to read and reduce the amount of time spent searching for information and enable you to cram more details into a smaller area. In general, NLP uses two approaches to summarize texts: extraction and abstraction. ROUGE stands for Recall Oriented Understanding for Gist Evaluation. In this study, ROUGE measurements were used for the summary procedure. In this text summary regressive transformers with a bi-directional auto-encoder, BART has been used as the model. It is a de-noising auto-encoder for seq-to-seq model pre-training. A bidirectional (like BERT) encoder and an autoregressive (like GPT) decoder are combined in the transformer encoder-decoder (seq2seq) paradigm known as BART. The recommended model produced an output summary with a quality score of about 78 percent for a given input text. [ABSTRACT FROM AUTHOR]

The process of summarization is taking a longer piece of material and reducing it to a shorter version without losing any of the essential information or meaning. Due to the exponential increase in available information and data, automatic text summarizing has emerged as a useful alternative to the time-consuming and error-prone process of manual summation of massive amounts of texts. The summaries the algorithm generates help users better understand the material presented in the original document. There are two main types of summarization: abstract and extractive. The number of available automatic summarization tools for Indian languages is low. Our focus in this area has been on creating an automatic English-language text summarizer utilizing the T5 transformer model; we've employed a manual dataset for testing and training. Here we have used news summary dataset. [ABSTRACT FROM AUTHOR]

Text summarization is the task of reducing the amount of text from a document while keeping the meaning unchanged. Previously statistical extractive metrics have been used to rank summaries however the current summarization models use an abstractive approach. In an abstractive summarization model the main idea from the original document is taken and it is used to generate the summary. Hence in such cases an abstractive measure of comparison is hence needed. An abstractive measure will give better results over extractive measures when contextual meaning of the text bodies are being considered. In this paper a new approach for abstractive text summarization evaluation is proposed which is based on BERT embeddings and Principal Component Analysis (PCA). The pre-trained BERT model is used to generate embeddings for each sentence in the input text, after which PCA is applied for reducing the dimensionality of the embeddings while preserving the most informative features. The L2 distance metric, which is basically Euclidean distance, is used for comparing the similarity of the reduced dimensional embedding vectors. The proposed approach outperforms the ROUGE metric and shows higher scores for summaries which are contextually similar rather than being textually similar. This metric is used to select sentences from the document which is being summarized to create an improved prompt for state-of-the-art text summarization models, like T5 and GPT, to generate more informative summaries. The prompt based generated summary is validated using ROUGE and METEOR metrics. [ABSTRACT FROM AUTHOR]

Bidirectional Encoder Representations from Transformers (BERT) is one method technique in machine learning for solving various natural language processing (NLP) problems, including summarization. One general problem using BERT is getting much time and resources to train the model. This paper aims to implement the small architecture BERT, called Small2BERT, which is expected to summarize the data. Therefore, this study compares pre-trained Small2BERT and Summarization using Word Frequency (SWF) in extracting text summaries. This research uses the dataset of Indian News Summary, which includes news articles from the Hindu, Indian Times, and Guardian from February to August 2017. Small2BERT surpasses SWF in precision, whereas SWF gives a higher score in overall high f1-score. Small2BERT earns a 0.307 f1-score in Rogue-1, compared to 0.35 for SWF. Small2BERT scores 0.19 and 0.304 while SWF achieves 0.304 and 0.31 in f1-score for Rogue-2 and Rogue-L, respectively. [ABSTRACT FROM AUTHOR]

Text summarization is a great skill to have, particularly when it is required to check the understanding of a topic, support an argument, ensure the ideas make sense, and save time. The ability to locate the important points provided by summarization systems will help the user spend less time reading. Over the last decade, the amount of data on the Internet has grown at an exponential rate. As a result, a solution is required to convert this massive amount of raw data into valuable data that a human brain can comprehend. Text summarization is a frequent research approach for dealing with large amounts of data. An automatic summary is a well-known method for distilling a document's important points. It works by providing a reduced version of the text that preserves significant information. This paper examines the strategies, resulting in a summary that is less redundant and more focused. [ABSTRACT FROM AUTHOR]

From numerous sources, a vast volume of text data has been developing daily. This enormous volume of data contains crucial knowledge and information that must be skillfully condensed to be usable. In this era of digitalization, vast information is available to us, which are being uploaded online from all over the world in the form of the web pages, and social media etc. To expand quick knowledge from this massive content, the summary of text can help understand the concept fast. Whenever we talk about the text summarization in research article or paper, we must be aware that this colossal amount of data can be generic or scientific and with this we will have an astronomical number of publications. There would be great support to the researchers if the approach of automatically summarization were applied on the scientific articles to increase the speed of understanding. Due of its applicability in numerous fields, automatic summarization of natural language is a prominent academic and industrial focus in computer science. While doing text summarization the essence or the context of the text remains same this could be the primary requirement. It is due to the high usage of Internet as a medium text summarization is developed as an outstanding tool for interpreting text information. Extraction of important information from the source text material is known as text summarization. In this procedure, the extracted data is produced as a concise report and sent to the user as a succinct summary. Broadly text summarization is classified into two methods extractive and abstractive techniques for summarization. Defining a rank from the text file based on the word and features of the sentence, collate them to develop a summary where feature of sentences will be notified based on the statistical and linguistic trait, which we name as extractive summarization method. In other case first we have to understand the details and main points of the documents, after which we need to define these concept in terms of natural language is defined as Abstractive Summarization method. This paper will give you the comparative study of various Abstractive Summarization methods. [ABSTRACT FROM AUTHOR]

Understanding natural languages are a very complex task for machines. Named entity recognition (NER) helps to understand natural language. NER aims to find mentions of specific identifiers in the text that belong to present semantic kinds such as person, place, organisation, time, money etc. In this paper a BERT (Bidirectional Encoder Representation from Transformation) model is trained with NER that improved the understanding of NLP. BERT is a transformer-based technique. It is trained and tested on CoNLL-2003 English dataset. The Proposed BERT based model gives 98.52% on English CoNLL-2003 NER dataset that is 4% greater than baseline models. The proposed BERT based NER model performed better results than other existing models. So, this features can enhance NLP applications in future like automatic text summarization, information retrieval systems, machine translation, machine reading comprehension, text classification etc. [ABSTRACT FROM AUTHOR]

The objective of the study is to perform the text summarization capabilities of XL Net algorithm comparatively with the BERT algorithm, estimate, and analyses the accuracy and the performance of both the Algorithms. Text summarization and file transfer workflows are performed using the transformers pre-trained models. The corpus of at least 10-20 pages is taken and the operations are performed on that corpus. In the same way, 30 sets of the corpus are taken and they are divided into two groups. The two groups were BERT and XLNET algorithms. Both algorithms were compared for their accuracy and speed in their performance. The proposed system found that the accuracy of the XL Net algorithm for text summarization has produced a high of 95.59% with a significant value at 2-tailed (p < 0.05) compared with the BERT algorithm 76.63%. The mean, standard deviation and statistical significance were compared for different using independent T-Test. Statistical significance difference was observed between BERT and XL Net, p = 0.001 (2-tailed). The study proved that XL Net achieved better accuracy in a shorter amount of time compared to the BERT algorithm. XL Net has outperformed BERT on 20 different tasks. In contrast to BERT, present Novel XL Net is based on an auto Regressi ve Model, where the next token is dependent on previous tokens. Study shows that XL Net is more performant than BERT with fewer parameters. BERT has predicted the tokens independent of each other whereas XL Net being auto regressive in nature predicted all the tokens in sequence. [ABSTRACT FROM AUTHOR]

Text Summarization is one of the most experimental tasks in deep learning that decreases the size of a text while it's important to keep its core and key information. Previous work in this domain is going to give remarkable results by using highly dependent data. Natural language processing-related works are trending due to complications in the implementation. Abstractive text summarization is a critical idea in natural language processing and also it's a trending, hardest problem that varies the solution or outcome based on the dataset. The objective of the paper is to rewrite into well-simplified and core information in text format that is retrieved from files or documents or datasets. Abstractive text summarization using RNN is the methodology going to be implemented in our paper. Attention mechanism and LSTM are going to be add-on advantages to the model. Both the methodologies improve the performance of the model and are also used to solve the complications problem. Keyword extracting is the one of features of this paper. State-of-art is going to be used to verify the model's efficiency. BLEU and ROUGE are going to be used to identify the accuracy and validate the model. The greatest thing about the dataset is the relationship between columns is independent of each other. Finally, the aim of the paper is to generate shortened text from whole huge data (text format). These kinds of problems are often used to solve in sequence-to-sequence passion [ABSTRACT FROM AUTHOR]

The fast development of data technology has created a straight system to store and access large quantities of information. The created system has the drawback of extracting potentially valuable knowledge not only in an efficient manner but also in a manner that is easily understood by human beings. One solution to it is in linguistic summarization. This enables the clearing of coherent data summaries that are more consistent with the human cognitive system. The major drawback of the existing applications is in involving high dimensional and distributed info which makes it tough to capture the relevant data. This research focuses on two tasks: one is in selecting the most significant content from source documents and the other is in generating coherent summary by using lexical chaining. In this research paper, an automatic method of text summarization depending on fuzzy sets to extract diversity of structures has been proposed to identify more significant information in the documents. The summary generated by the system is compared to a summary created by domain experts. This method is completely different from other proposed methods described in the literature survey. The text summary is created in the proposed method by testing eight connected features via reduced dimensionality and less fuzzy set rules used for text summarization. In this method, the documents have been summarized by probing connected features and subsequently by different fuzzy sets. The DUC dataset is used during the training and testing phases of the planned summary system. Base Line, weight, accuracy, memory, and F-measure assess the planned system. The outcome of the experiments demonstrations that the proposed technique provides well f-measure than baseline and weight approaches [ABSTRACT FROM AUTHOR]

In recent times, there is an extensive growth of multi model data holders like images, pdfs and documents holding a large amount of data due to rapid increase in usage of internet, social media and other information providing technologies. Manually extracting relevant information and discarding redundant information from the large sets of data is a prolonged work and enormous work. The automation of multi-modal data summarization is an important and necessary task. The rapid advancement in the deep learning techniques as neural networks shows the strength of Deep learning in text summarization. This paper proposes a neural-based abstractive multi-modal data summarization based on pre-trained modal approach in Transfer Learning. Existing models like TFRSP, RNN unable to get accurate results and consumes more GPU resources. This method first extracts the text from the documents, images and web-URL and encodes the extracted text from document, image and web-URLs with positional encoding and abstract the summary probability of sentence with unified Text-to-Text Transformer approach in transfer learning. This experiment is done by fine tuning the trained model on Colossal Clean Crawled Corpus(C4) dataset. Experiments show that our method outperforms the novel neural network architecture for language understanding methods for summarization and then translates to the local languages which can be saved as future references. [ABSTRACT FROM AUTHOR]

With the exponential growth in data, it is essential to analyse it. The data is a valuable source of information and knowledge that should be effectively interpreted to be helpful. The data should be interpreted in different forms based on the user requirements to get precise information and knowledge. Nowadays smartphones are the most commonly used electronic device. A smartphone is not only a communication device, it is also a powerful computing device. So it is possible to apply translation, text extraction, summary, and much more techniques, which require much computational work. This paper presents an application to analyse the text from documents on smartphones. However, it is challenging to interpret the documents on smartphones. The proposed application converts the documents or images to searchable and editable digital text, and further, it can be used to analyse them into different forms. The objective of this application is four-fold 1) To recognize text from documents or images by using optical Character Recognition 2) Summarization of the text 3) Translate the extracted text to different languages 4) Generation of speech from the text by using a text-to-speech algorithm. [ABSTRACT FROM AUTHOR]

In this paper we apply the generative deep learning language model to text summarization task. Because such large language models require a lot of resources to train, it is interesting to study how much long training affects the final result and at which point it gets saturated. The experiments are ran using mT5 model and 2 news corpora RIA and Lenta. During this research we achieve state-of-the-art results for the task of title prediction by news texts: 47.60 Rouge-L for RIA corpus and 40.69 Rouge-L for Lenta. It is shown that after 800k updates of the weights, the accuracy scores continue to grow, but their growth becomes more and more insignificant. [ABSTRACT FROM AUTHOR]

Natural Language Processing (NLP) is a field of computer science and linguistics that explores how computers interact with human (natural) language. NLP is frequently seen as a subfield of artificial intelligence, and its research area overlaps with computational linguistics. Machine translation, natural language text processing and summarization, user interfaces, speech recognition, and expert systems are all examples of NLP applications. Low-level NLP tasks and higher-level NLP tasks are two types of Natural Language Processing tasks. Tokenization, part-of-speech assignment to individual words (POS tagging), and shallow parsing are examples of low-level NLP activities (chunking). Meanwhile, higher-level NLP tasks, such as spelling/grammatical mistake detection and recovery, named entity recognition (NER), and information extraction, are constructed on top of low-level NLP tasks and used according to the problems found (IE). This study uses natural language processing to extract the location of congestion from tweets (twit). We get the location name and traffic conditions from the tweet by implementing a low-level NLP task in the form of tokenization, POS tagging, and chunking, followed by a higher-level NLP work in the form of named entity recognition. The types of words, such as nouns, verbs, adjectives, and descriptions, are taught at the introductory stage. Following POS tagging, word grouping (chunking) is carried out; if numerous consecutive words are verbs, they are classified as location names, whereas adjectives are grouped as traffic conditions. [ABSTRACT FROM AUTHOR]

Text Summarization is a process of shortening the longer text document(s) into a summary of informative sentences in an orderly way. Extractive summarization retrieves the important sentences without making any change to actual sentences. Each sentence in the document is represented as a vector of real numbers, which is termed as embedding. While embedding a sentence, values of specific features are analyzed and are plotted in n-dimensional space. This results in effective prediction of following and preceding sentences of the current sentence. Thus, the sentences with similar semantics lie closer to each other. Unsupervised Summarization groups the similar sentences by estimating the distance between the vectors and decides if sentence must be included in summary or not. Hierarchical Summarization constructs a tree-structure with the input text data, where the truncation of tree is done based on the number of clusters. Each cluster holds sentences that are semantically similar. By determining the nearest neighbor in each cluster, a specified number of sentences were retrieved from each cluster to be included in summary which holds at-least half the size of the input document(s). The performance of hierarchical summarization is measured on CNN/Daily Mail dataset and is determined using performance metrics. The evaluation score concludes that BIRCH algorithm performs better than the other Hierarchical Clustering techniques. [ABSTRACT FROM AUTHOR]