The news domain is an ever-evolving field, and it is more challenging to standardize text summarization for Indian low-resource languages because of the distinct syntax and semantics. It became very important to find an efficient method that could generate a concise summary. In this paper, we develop an evolutionary algorithm-based approach, namely, ILrLSUMM to generate concise extractive summaries for low-resource Indian languages, focusing on Hindi and Gujarati. To select the relevant sentences from a document to form a summary, our method employs a single-objective optimization process utilizing the efficacy of the differential evolutionary algorithm, which is a first of its kind as per knowledge. We investigate three key objectives: tf-idf score, sentence-to-title similarity, and thematic score. Our approach is purely unsupervised in nature; therefore, we utilized 500 articles from the M3LS dataset for a broader comparative analysis with the existing algorithms. Our evalu-ation was based on ROUGE scores, comparing our generated summaries with gold-standard summaries in the dataset. The results were promising in the sense that our method outperformed existing techniques, including large language models (LLMs) by 34% in Hindi and 53% in Gujarati on an average, according to the ROUGE-I Fl. This significant improvement highlights the effectiveness of our approach to handling text summarization for underrepresented languages.

Social media platforms like Twitter act as a medium for communication among people, government agencies, NGOs, and other relief providing agencies in widespread humanitarian havoc during a disaster outbreak when other communication means might not be available. Various agencies leverage twitter's open and public features to get timely and reliable updates, thus support agencies in communicating with the people on rescue and provide immediate relief. As situational updates are mixed in millions of other tweets, an efficient system is required to extract and summarize these tweets. We have developed a novel framework that uses a deep learning-based classification model to separate the informational tweets from others and summarizes them in the current paper. Non-situational tweets mostly comprise sentiments like grief, anger, sorrow, etc. Motivated by this observation, we have solved sentiment classification and informative tweet selection tasks simultaneously using a multi-task learning (MTL) in a deep-learning framework. Our summarization approach generates clustering solutions using various existing approaches and then ensembles cluster solutions using generative modelling. A summary is formulated by extracting tweets from different clusters. The proposed approach's superior performance on four disaster-related events indicates the developed framework's efficiency over state-of-the-art techniques.

In the modern era, the rapid expansion of social media and the proliferation of the internet community has led to a multi-fold increase in the richness and range of views and outlooks expressed by readers and viewers. To obtain valuable insights from this vast sea of opinions, we present an inventive and holistic procedure for multi-modal abstractive summarization with comment sensitivity. Our proposed model utilizes both textual and visual modalities and examines the remarks provided by the readers to produce summaries that apprehend the significant points and opinions made by them. Our model features a transformer-based encoder that seamlessly processes both news articles and comments, merging them before transmitting the amalgamated information to the decoder. Additionally, the core segment of our architecture consists of an attention-based merging technique which is trained adversarially by means of a generator and discriminator to bridge the semantic gap between comments and articles. We have used a Bi-LSTM-based branch for image pointer generation. We assess our model on the reader-aware multi-document summarization (RA-MDS) dataset which contains news articles, their summaries, and related comments. We have extended the dataset by adding images pertaining to news articles in the corpus to increase the richness and diversity of the dataset. Our comprehensive experiments reveal that our model outperforms similar pre-trained models and baselines across two of the four evaluated metrics, showcasing its superior performance.

Existing extractive summarization systems suffer from the problem of identifying correct aspects while generating summary from multiple documents on a particular topic. Online access to documents allows readers to post comments on available documents/blogs which may be helpful in correctly identifying the important aspects from a document in a crowd-sourcing fashion. However, comments can be unstructured and boisterous, making them difficult to manage. The current paper builds an unsupervised summarization system utilizing documents and the corresponding comments by carefully selecting a subset of sentences from the document based on document-comment similarities. A binary multiobjective optimization technique exploring the search capability of differential evolution is utilized for simultaneously optimizing different aspects of summarization like diversity between sentences of the summary, user-attention and density-based score providing importance to the comments with respect to the documents, user-attention with a syntactic score between the summary phrases. Extensive analysis is performed by varying the combinations of different objective functions, on a dataset containing 45 topics belonging to different news themes as well as in a new French dataset having documents, summaries, and comments. The experimental result validates the importance of all the objective functions and superiority over other methods.

Microblog summarization systems are gaining importance during natural disasters. A lot of tweets are posted along with multimedia content during the occurrence of any natural disaster event. Extracting relevant information/summary from these tweets is important for the smooth functioning of the rescue operation. Moreover, because of the limited size of the tweets, in many cases, tweets are associated with images. The current work is the first of its kind where both the image and the tweet text are utilized simultaneously to generate a summary from microblog data generated during a disaster event. Different aspects, such as syntactic similarity, the maximum length of the tweets, retweet score, and antiredundancy, are considered as objective functions and those are simultaneously optimized using a metaheuristic population-based evolutionary strategy to select a good set of tweets to form a good quality summary. In order to extract information from images, a dense captioning model is utilized and the dense captions are further utilized for calculating the antiredundancy measure. We employed word mover distance to capture the semantic similarity between two tweets. Due to the unavailability of the dataset for multimodal microblog summarization tasks in a disaster-event scenario, datasets are created and made openly available to the community. The obtained summarization results are evaluated using the well-known ROUGE measure.

Software developers frequently use existing bug text reports to help them in understanding the different aspects of the specific defects and changes made to resolve that bug. But, bug reports are usually lengthy in nature and require considerable effort in understanding. In this direction, summarization of bug reports seems to be useful, covering relevant and diversified information. In the current article, we investigate the use of Multi-objective Evolutionary Algorithm (MEA) for BRS and thus, we name our approach as MEABRS. For MEA, we utilize the search capability of multi-objective bi-nary differential evolution (MOBDE) where we simultaneously optimize different aspects of BRS including diversity among sentences, sentence relevance using term weighting scheme, and length of the sentence. A keyword-based objective function is also incorporated in our optimization process to improve the quality of a bug report’s summary and in order to do so, rapid automatic keyword extraction (RAKE) toolkit is utilized. The generated summaries are evaluated with the available gold summaries corresponding to the two benchmark datasets (ADS and SDS) in terms of precision, recall, F-measure, and ROUGE measure. Results obtained demonstrate the efficacy of MEABRS with an average improvement (over both datasets) of 4.5% in terms of F1-Measure. Further, results are also validated using a statistical significance t-test.

Nowadays, micro-blogging sites are getting popular due to the involvement of a large number of users. In the case of natural disasters, a significant amount of relevant information (giving crucial information) are present amongst the tweets. Therefore, there is a need to develop a system that summarizes relevant tweets by extracting informative tweets. In the current paper, we have proposed an unsupervised approach for summarizing the relevant tweets namely, MOOTweetSumm+, which automatically selects the informative tweets. Several tweet-scoring measures: (a) anti-redundancy measuring the dissimilarity between tweets; (b) similarity with outputs provided by LexRank (a graph-based method measuring tweet importance based on the concept of eigen-vector centrality in a graph); (c) BM25 based ranking function; (d) tf-idf based ranking function; (e) length of the tweet; (f) re-tweet count, are simultaneously optimized utilizing a binary differential evolution algorithm. Further, two different versions of the LexRank, utilizing syntactic and semantic similarity, have also been explored. For evaluation, four different disaster-event related datasets are used, and performance is measured in terms of ROUGE scores. An ablation study is also performed to determine which set of measures is best suited for different datasets. From the results obtained, it is clearly evident that our approach improves by 13.2% and 5.8% in terms of ROUGE-2 and ROUGE-L scores, over the existing approaches, respectively.

The rate of publishing scientific articles is increasing day by day which has created difficulty for the researchers to learn about the recent advancements in a faster way. Also, relying on the abstract of these published articles is not a good idea as they cover only broad ideas of the article. The summarization of scientific documents (SDS) addresses this challenge. In this paper, we propose a system for SDS having two components: identifying the relevant sentences in the article using citation context; generation of the summary by posing SDS as a binary optimization problem. For the purpose of optimization, a metaheuristic evolutionary algorithm is utilized. In order to improve the quality of summary, various aspects measuring the relevance of sentences are simultaneously optimized using the concept of multi-objective optimization. Inspired by the popularity of graph-based algorithms like LexRank which is popularly used in solving summarization problems of different real-life applications, its impact is studied in fusion with our optimization framework. An ablation study is also performed to identify the most contributing aspects for the summary generation. We investigated the performance of our proposed framework on two datasets related to the computational linguistic domain, CL-SciSumm 2016 and CL-SciSumm 2017, in terms of ROUGE measures. The results obtained illustrate that our framework effectively improves other existing methods. Further, results are validated using the statistical paired t-test.

This article proposes a novel unsupervised approach (FigSum++) for automatic figure summarization in biomedical scientific articles using a multi-objective evolutionary algorithm. The problem is treated as an optimization problem where relevant sentences in the summary for a given figure are selected based on various sentence scoring features (or objective functions), such as the textual entailment score between sentences in the summary and a figure’s caption, the number of sentences referring to that figure, semantic similarity between sentences and a figure’s caption, and the number of overlapping words between sentences and a figure’s caption. These objective functions are optimized simultaneously using multi-objective binary differential evolution (MBDE). MBDE consists of a set of solutions, and each solution represents a subset of sentences to be selected in the summary. MBDE generally uses a single differential evolution variant, but in the current study, an ensemble of two different differential evolution variants measuring diversity among solutions and convergence toward global optimal solution, respectively, is employed for efficient search. Usually, in any summarization system, diversity among sentences (called anti-redundancy ) in the summary is a very critical feature, and it is calculated in terms of similarity (like cosine similarity) among sentences. In this article, a new way of measuring diversity in terms of textual entailment is proposed. To represent the sentences of the article in the form of numeric vectors, the recently proposed BioBERT pre-trained language model in biomedical text mining is utilized. An ablation study has also been presented to determine the importance of different objective functions. For evaluation of the proposed technique, two benchmark biomedical datasets containing 91 and 84 figures are considered. Our proposed system obtains 5% and 11% improvements in terms of the F -measure metric over two datasets, compared to the state-of-the-art unsupervised methods.

In recent years, social networking sites such as Twitter have become the primary sources for real-time information of ongoing events such as political rallies, natural disasters, and so on. At the time of occurrence of natural disasters, it has been seen that relevant information collected from tweets can help in different ways. Therefore, there is a need to develop an automated microblog/tweet summarization system to automatically select relevant tweets. In this article, we employ the concept of multiobjective optimization in microblog summarization to produce good quality summaries. Different statistical quality measures namely, length, tf-idf score of the tweets, antiredundancy, measuring different aspects of summary, are optimized simultaneously using the search capability of a multiobjective differential evolution technique. Different types of genetic operators including recently developed self-organizing map (a type of neural network) based operator, are explored in the proposed framework. To measure the similarity between tweets, word mover distance is utilized which is capable of capturing the semantic similarity between tweets. For evaluation, four benchmark data sets related to disaster events are used, and the results obtained are compared with various state-of-the-art techniques using ROUGE measures. It has been found that our algorithm improves by 62.37% and 5.65% in terms of ROUGE-2 and ROUGE-L scores, respectively, over the state-of-the-art techniques. Results are also validated using statistical significance t-test. At the end of the article, extension of the proposed approach to solve the multidocument summarization task is also illustrated.

The self-organizing map (SOM) helps in exploratory phase of data mining by projecting the input data into a lower dimensional map. In recent years SOM has also been applied for classification of data points. The prominent utility of SOM based classification is evident from the use of no labeled data during training. In a multi-class classification problem where classes have high degree of overlap, it would be difficult to design a single-level SOM based classification system which can perform well for all the classes. In order to deal with the multi-class classification efficiently, the current paper proposes to develop a Cascaded SOM based architecture where classes are handled in a hierarchical way. Also, it can be applied for solving any multi-class classification problem where labeled data is limited. As a case study, in the first part of the paper results are shown for single label version of complex email classification problem where classes are highly overlapping to each other and in the second part, results are shown for some multi-labeled data sets. Different representation schemas for emails and a large set of features are also adopted for the purpose of experiment. Proposed Cascaded SOM based classification model performs well in email-classification compared to standard classification approaches and classical SOM based model.