Topic Coherence Research Articles

Feature Engineering for Agile Requirement Management Using Semantic Analysis

Efficient management and prioritization of software requirements are critical challenges in agile projects, where requirements constantly evolve due to changing user needs, business goals, and regulatory updates. This paper explores the role of semantic feature extraction in enabling adaptive management strategies. Using the PROMISE Expanded Dataset and the Coquina Dataset, we employed TF-IDF weighted Word2Vec for advanced tokenization and feature extraction. Latent Dirichlet Allocation (LDA) was used to analyze how preprocessing steps like stop word removal impact topic representation, revealing that removing stop words improved topic specificity and coherence. To address class imbalance, Synthetic Minority Over-sampling Technique (SMOTE) was applied, enhancing the model's ability to handle underrepresented classes effectively. Principal Component Analysis (PCA) reduced the dimensionality of TF-IDF weighted Word2Vec embeddings from 100 features to 30, while Analysis of Variance (ANOVA) identified the most significant features for classification. The results obtained identified three features to have p-values below 0.05 as statistically significant, p-value = 0.0000605, p-value = 0.00000000469, and p-value = 0.0024. These extracted features could be used as input to training machine learning models for predicting and managing software requirements adaptively during agile development. With the reduction of ambiguities and sentiments of the user at the requirement phase, the development phase could be undertaken seamlessly with ease.

A method for exploring consumer satisfaction factors using online reviews: A study on anti-cold drugs

Anti-cold drugs prove effective in alleviating cold symptoms and minimizing complications. In recent years, there has been a growing tendency among the public to purchase these drugs online, driven by the prevalence of influenza. Maintaining a high level of consumer satisfaction is crucial for online pharmaceutical platforms. This study introduces a Multi-Attribute Decision Making (MADM) method to analyze online reviews from 31,392 consumers at JD Pharmacy. Unlike previous studies, multiple analytical tools are employed jointly to identify factors influencing consumer satisfaction. Four dosage forms of anti-cold drugs (granule, tablet, syrup, and spray) were analyzed using the Latent Dirichlet Allocation Model (LDA). Three tools (topic coherence curve, topic perplexity curve, and pyLDAvis visualization) were jointly applied. Then five satisfaction factors were identified as logistics distribution, drug efficacy, online consultation service, drug prices, and packaging quality. PROMETHEE II method was applied to rank the above factors. The results revealed variations in consumer satisfaction among different drug dosage forms. We conducted a detailed analysis of these distinctions. The results from this study will provide an effective reference to improve consumer satisfaction.

A data-driven analysis to determine the optimal number of topics 'K' for latent Dirichlet allocation model

Topic modeling is an unsupervised machine learning technique successfully used to classify and retrieve textual data. However, the performance of topic models is sensitive to selecting optimal hyperparameters, the number of topics 'K' and Dirichlet priors 'α' and 'β.' This data-driven analysis aims to determine the optimum number of topics, 'K,' within the latent Dirichlet allocation (LDA) model. This work utilizes three datasets, namely 20-Newsgroups news articles, Wikipedia articles, and Web of Science containing science articles, to assess and compare various 'K' values through the grid search approach. The grid search approach finds the best combination of hyperparameter values by trying all possible combinations to see which performs best. This research seeks to identify the 'K' that optimizes topic relevance, coherence, and model performance by leveraging statistical metrics, such as coherence scores, perplexity, and topic distribution quality. Through empirical analysis and rigorous evaluation, this work provides valuable insights for determining the ideal 'K' for LDA models.

Learning to Write Argumentation Paragraphs Using the Problem Based Learning (PBL) Model

Background. Problem-based learning (PBL) has emerged as an innovative approach in education, particularly in developing critical thinking and writing skills. In the context of learning to write argumentative paragraphs, there is a need to explore the effectiveness of PBL in enhancing students' writing abilities and engagement. Purpose. This study aims to test and analyze the effectiveness of the PBL model in teaching students to write argumentative paragraphs, focusing on topic suitability and content coherence. Method. A qualitative research method with a descriptive approach was employed to identify, describe, and explain the research subjects. The study utilized multiple data collection instruments, including written tests, questionnaires, observations, and interviews. These tools were used to assess students' progress in essay writing and their perceptions of the learning experience. The research was conducted in two cycles to evaluate the implementation and impact of the PBL model. Results. The study found significant improvements in students' performance across the two cycles. In Cycle I, the highest score was 90, the lowest was 70, with an average score of 90. Cycle II showed further improvement, with the highest score reaching 100, the lowest score increasing to 90, and an average score of 100. Conclusion. The findings indicate that the PBL model is effective in enhancing students' ability to write argumentative paragraphs. The marked improvement in scores from Cycle I to Cycle II suggests that the PBL approach contributes to better topic selection, content development, and overall writing quality. This study underscores the potential of PBL as a valuable instructional strategy for developing argumentative writing skills in students.

Graph Contrastive Topic Model

Contrastive learning has recently been introduced into neural topic models (NTMs) to improve latent semantic discovery, but existing methods suffer from the sample bias problem owing to word frequency-based sampling strategy, which may result in false negative samples with similar semantics to the prototypes. We propose the novel graph contrastive neural topic model (GCTM), based on the graph-based sampling strategy, guided by the in-depth correlation and irrelevance information among documents and words. We model the input document as the document word bipartite graph (DWBG) and construct positive and negative word co-occurrence graphs (WCGs), to capture in-depth semantic correlation and irrelevance among words. Based on the DWBG and WCGs, we design the document-word information propagation (DWIP) process to perform the edge perturbation of DWBG, based on multi-hop correlations/irrelevance among documents and words. This yields the desired negative and positive samples, which are utilized for GCL together with the prototypes to improve learning document topic representations and latent topics. Experiments on several benchmark datasets demonstrate the effectiveness of our method for topic coherence and document representation learning compared with existing state-of-the-art methods.

Corporate financial distress prediction using the risk-related information content of annual reports

This study presents a financial distress prediction model focusing on the linguistic analysis of risk-related sections of corporate annual reports. Here, we introduce a novel methodology that leverages BERT-based contextualized embedding models for nuanced extraction of financial sentiment and topic coherence. This stands in contrast to existing research, which predominantly relies on dictionary-based or non-contextual word embeddings and addresses their limitations in context sensitivity. Furthermore, we apply an innovative financial distress prediction model that combines the robust XGBoost algorithm with unsupervised outlier detection techniques. This hybrid model is specifically designed to tackle the issue of class imbalance, a persistent challenge in financial distress prediction. The efficacy of the proposed model is empirically validated using a comprehensive dataset of 2545 companies listed on major global stock exchanges. Our findings indicate that the introduced model not only significantly outperforms most existing state-of-the-art financial distress prediction models in terms of predictive accuracy, but also significantly outperforms the Loughran & McDonald dictionary-based approach and the Word2Vec model, underlining its potential as a superior analytical tool for financial distress prediction.

Data-driven View of Customer Psychological Anxiety Attributes for Cooperate Social Responsibility in Supply Chain Management: a LDA Model

Background-Cooperate Social Responsibility (CSR) in supply chain management requires understanding customer psychological anxiety attributes. A data-driven approach, such as using a Latent Dirichlet Allocation (LDA) model, can provide insights. By recognizing and addressing customer psychological anxiety, CSR can provide better support, which leads to higher customer satisfaction. When customers feel understood and supported, they are more likely to have a positive perception of the company and its supply chain services. Subjects and methods- The study aims to use the LDA method to explore consumer psychology anxiety and its attributes for CSR in supply chain management. The corpus is collected from the Web of Science core collection with keywords “CSR” and “supply chain management”, and 965 articles related to the field from1990-2022. LDA is a natural language processing technique that uncovers thematic structures in textual data. By applying LDA to customer feedback, businesses can identify anxiety attributes. Steps include data collection, preprocessing, LDA model training, topic interpretation, and deriving business insights. Results- The study used the Python program to run LDA, after putting in the text data, the study identify number of 11 topics according to the value of topic coherence, then identified topics based on the most representative words or phrases within each topic combined CSR and supply chain related knowledge and fit the service quality model to find the customer psychological anxiety attributes, the results reveal that customer tend to be anxiety about the aspects of reliability and assurance for CSR in supply chain management, more concern about the environmental and social aspects responsibility. Conclusions- The study revealed customer psychological anxiety about CSR initiatives and strategies within their supply chain management. The results show that consumers feel most anxiety about the aspects of reliability and assurance of supply chain management. They are concerned about the environment and social responsibility that the supply chain enterprise is taking, especially in the food supply chain field. By addressing these concerns, organizations can enhance customer satisfaction, build stronger relationships, and improve their overall CSR performance. The findings of this study also contribute to the field by providing valuable guidance, further research can add other text mining methods like structural topic modelling, or add deeper quantitative research in the field and develop of a new customer-CSR service quality model in the service industry.

Applying short text topic models to instant messaging communication of software developers

When modeling topics from chat messages of developer instant messaging communication, individual chat messages are short text documents. Our study aims at understanding how short text topic models perform with conversations from developer instant messaging. We applied four models to nine Gitter chat rooms (with sizes ranging from ≈100 to ≈160,000 messages). To assess the quality of topics and identify the best performing models, we compared topics based on four metrics for topic coherence. Furthermore, for a subset of Gitter chat rooms we used two human-based assessments: intrusion tasks with 18 experts analyzing 40 topics each, and topic naming (assigning a name to a topic that summarizes its main concept) with eight additional experts naming 60 topics each. Models performed differently in terms of coherence metrics and human assessment depending on the corpus (small, medium or large chat room). Our findings offer recommendations for the selection and use of short text topic models with developer chat messages based on characteristics of models and their performance with different sizes of corpora, and based on different strategies to assess topic quality.

Investigating topic modeling techniques through evaluation of topics discovered in short texts data across diverse domains

The online channel has affected many facets of an individual's identity, commercial, social policy, and culture, among others. It implies that discovering the topics on which these brief writings are focused, as well as examining the qualities of these short texts is critical. Another key issue that has been identified is the evaluation of newly discovered topics in terms of topic quality, which includes topic separation and coherence. A topic modeling method has been shown to be an outstanding aid in the linguistic interpretation of quite tiny texts. Based on the underlying strategy, topic models are divided into two categories: probabilistic methods and non-probabilistic methods. In this research, short texts are analyzed using topic models, including latent Dirichlet allocation (LDA) for probabilistic topic modeling and non-negative matrix factorization (NMF) for non-probabilistic topic modeling. A novel approach for topic evaluation is used, such as clustering methods and silhouette analysis on both models, to investigate performance in terms of quality. The experiment results indicate that the proposed evaluation method outperforms on both LDA and NMF.

Exploring public voice on social media: Twitter Users' views on the circular economy

AbstractThis paper aims to explore all the ideas and topics associated with the discussion about the circular economy on social media sites. All tweets mentioning the CE were collected and subjected to latent Dirichlet allocation topic modelling to identify discussion themes manual validation refined topic coherence. Findings offer novel empirical insights into how the CE is currently framed in public discourse. Dominant themes emphasise strategic communication and technical CE activities rather than inclusive participation. Gaps were also identified, such as limited discussion of CE's social role and broader activities beyond recycling. By quantitatively examining large‐scale online discussions, this research makes an original contribution to establishing the neglected social dimension of CE transitions. The findings carry essential theoretical implications for refining CE conceptualisations better to integrate socio‐cultural factors critical for just, systemic change. Practically, the study equips decision‐makers with data‐driven guidance on developing more participatory communication approaches and policies promoting the CE. This study transitions research and advances the scholarship of the circular economy by using a mixed‐methods approach to analyse a crucial but little‐studied stakeholder perspective.

Correlation Aware Relevance-Based Semantic Index for Clinical Big Data Repository.

In this paper, we focus on indexing mechanisms for unstructured clinical big integrated data repository systems. Clinical data is unstructured and heterogeneous, which comes in different files and formats. Accessing data efficiently and effectively are critical challenges. Traditional indexing mechanisms are difficult to apply on unstructured data, especially by identifying correlation information between clinical data elements. In this research work, we developed a correlation-aware relevance-based index that retrieves clinical data by fetching most relevant cases efficiently. In our previous work, we designed a methodology that categorizes medical data based on the semantics of data elements and merges them into an integrated repository. We developed a data integration system for medical data sources that combines heterogeneous medical data and provides access to knowledge-based database repositories to different users. In this research work, we designed an indexing system using semantic tags extracted from clinical data sources and medical ontologies that retrieves relevant data from database repositories and speeds up the process of data retrieval. Our objective is to provide an integrated biomedical database repository that can be used by radiologists as a reference, or for patient care, or by researchers. In this paper, we focus on designing a technique that performs data processing for data integration, learn the semantic properties of data elements, and develop a correlation-aware topic index that facilitates efficient data retrieval. We generated semantic tags by identifying key elements from integrated clinical cases using topic modeling techniques. We investigated a technique that identifies tags for merged categories and provides an index to fetch data from an integrated database repository. We developed a topic coherence matrix that shows how well a topic is supported by a corpus from clinical cases and medical ontologies. We were able to find more relevant results using an annotation index from an integrated database repository, and there was a 61% increase in a recall. We evaluated results with the help of experts and compared them with naive index (index with all terms from the corpus). Our approach improved data retrieval quality by providing most relevant results and reduced data retrieval time as we applied correlation-aware index on an integrated data repository. Topic indexing approach proposed in this research work identifies tags based on a correlation between different data elements, improves data retrieval time, and provides most relevant cases as an outcome of this system.

Topic modeling for conversations for mental health helplines with utterance embedding

Conversations with topics that are locally contextual often produces incoherent topic modeling results using standard methods. Splitting a conversation into its individual utterances makes it possible to avoid this problem. However, with increased data sparsity, different methods need to be considered. Baseline bag-of-word topic modeling methods for regular and short-text, as well as topic modeling methods using transformer-based sentence embeddings were implemented. These models were evaluated on topic coherence and word embedding similarity. Each method was trained using single utterances, segments of the conversation, and on the full conversation. The results showed that utterance-level and segment-level data combined with sentence embedding methods performs better compared to other non-sentence embedding methods or conversation-level data. Among the sentence embedding methods, clustering using HDBScan showed the best performance. We suspect that ignoring noisy utterances is the reason for better topic coherence and a relatively large improvement in topic word similarity.

Conceptualizing Discussions on the Dark Web: An Empirical Topic Modeling Approach

Social networks on the Internet have become a home that attracts all types of human thinking to exchange knowledge and ideas and share businesses. On the other hand, it has also become a source for researchers to analyze this knowledge and frame it in patterns that define types of thoughts circulating on these networks and representing the communities around them. In particular, some social networks on the Dark Web attract a special kind of thinking centered around the malicious and illegal activities disseminated on websites and marketplaces on the Dark Web. These networks involve discussions to exchange and discourse information, tips, and advice on performing such business. Studying social networks on the Dark Web is still in its infancy. In this paper, we present a methodology for analyzing the content of social networks on the Dark Web using topic modeling methods. We demonstrate the needed stages for the topic modeling process, beginning with data preprocessing and feature extraction to topic modeling algorithms. We utilize and discuss the following four topic models: LDA, CTM, PAM, and PTM. We discuss the following four topic coherence measures as evaluation metrics: UMass, UCI, CNPMI, and CV, demonstrating the selection of the best number of topics for each model according to the most coherent produced topics. Furthermore, we discuss the limitations, challenges, and future work. Our proposed approach highlights the ability to discover the latent thematic patterns in conversations and messages in the common language used in social networks on the Dark Web, constructing topics as groups of terms and their associations. This paper provides researchers with a leading methodology for analyzing thought patterns on the Dark Web.

Topic Modeling: A Consistent Framework for Comparative Studies

In recent years, the field of Topic Modeling (TM) has grown in importance due to the increasing availability of digital text data. TM is an unsupervised learning technique that helps uncover latent semantic structures in large sets of documents, making it a valuable tool for finding relevant patterns. However, evaluating the performance of TM algorithms can be challenging as different metrics and datasets are often used, leading to inconsistent results. In addition, many current surveys of TM algorithms focus on a limited number of models and exclude state-of-the-art approaches. This paper has the objective of addressing these issues by presenting a comprehensive comparative study of five TM algorithms across three different benchmark datasets using five different metrics. We offer an updated survey of the latest TM approaches and evaluation metrics, providing a consistent framework for comparing different algorithms while introducing state-of-the art approaches that have been disregarded in the literature. The experiments, which primarily use Context Vectors (CV) Topic Coherence as an evaluation metric, show that Top2Vec is the best-performing model across all datasets, disrupting the tendency for Latent Dirichlet Allocation to be the best performer. Doi: 10.28991/ESJ-2024-08-01-09 Full Text: PDF

MIT: Mutual Information Topic Model for Diverse Topic Extraction.

To automatically mine structured semantic topics from text, neural topic modeling has arisen and made some progress. However, most existing work focuses on designing a mechanism to enhance topic coherence but sacrificing the diversity of the extracted topics. To address this limitation, we propose the first neural-based topic modeling approach purely based on mutual information maximization, called the mutual information topic (MIT) model, in this article. The proposed MIT significantly improves topic diversity by maximizing the mutual information between word distribution and topic distribution. Meanwhile, MIT also utilizes Dirichlet prior in latent topic space to ensure the quality of mined topics. The experimental results on three publicly benchmark text corpora show that MIT could extract topics with higher coherence values (considering four topic coherence metrics) than competitive approaches and has a significant improvement on topic diversity metric. Besides, our experiments prove that the proposed MIT converges faster and more stable than adversarial-neural topic models.

Decoding Multilingual Topic Dynamics and Trend Identification through ARIMA Time Series Analysis on Social Networks: A Novel Data Translation Framework Enhanced by LDA/HDP Models

In this study, the authors present a novel methodology adept at decoding multilingual topic dynamics and identifying communication trends during crises. We focus on dialogues within Tunisian social networks during the coronavirus pandemic and other notable themes like sports and politics. We start by aggregating a varied multilingual corpus of comments relevant to these subjects. This dataset undergoes rigorous refinement during data preprocessing. We then introduce our No‐English‐to‐English Machine Translation approach to handle linguistic differences. Empirical tests of this method show high accuracy and F1 scores, highlighting its suitability for linguistically coherent tasks. Delving deeper, advanced modeling techniques, specifically LDA and HDP models, are employed to extract pertinent topics from the translated content. This leads to applying ARIMA time series analysis to decode evolving topic trends. Applying our method to a multilingual Tunisian dataset, we effectively identify key topics mirroring public sentiment. Such insights prove vital for organizations and governments striving to understand public perspectives during crises. Compared to standard approaches, our model outperforms, as confirmed by metrics like coherence score, U‐mass, and topic coherence. Additionally, an in‐depth assessment of the identified topics reveals notable thematic shifts in discussions, with the proposed trends’ identification indicating impressive accuracy, backed by RMSE‐based analysis.

Exploring Topic Coherence with PCC-LDA and BERT for Contextual Word Generation

Exploring Topic Coherence with PCC-LDA and BERT for Contextual Word Generation

Aspect Based Sentiment Analysis: Feature Extraction using Latent Dirichlet Allocation (LDA) and Term Frequency - Inverse Document Frequency (TF-IDF) in Machine Learning (ML)

The growth and development of social networks, blogs, forums, and e-commerce websites has produced a number of data, notably textual data, which has increased tremendously. Twitter is one of the most popular media social platforms; during the COVID-19 pandemic, people all around the world use social media to share their opinions or concerns about the pandemic that has changed their lives. It revealed a significant rise in tweets on coronavirus, including positive, negative, and neutral tweets about the virus's impact. Sentiment analysis faces challenges: sparse data limits understanding, while topic coherence and interpretability demand improvement for clearer insights. The primary goal of this paper is to improve the accuracy and effectiveness of sentiment analysis during the COVID-19 pandemic through the application of advanced techniques and classifiers. In this article, we experiment with such Support Vector Machines (SVM) and Naive Bayes (NB) on Twitter data for high-accuracy machine learning models. Using Latent Dirichlet Allocation (LDA)for feature extraction, we aim to capture comprehensive aspects and topics for sentiment analysis. Additionally, we explore Count Vectorizer and Term Frequency - Inverse Document Frequency (TF-IDF) as word embedding techniques. The main objectives are to extract topics, understand public concerns about Covid-19, and compare classifier performance in Aspect-Based Sentiment Analysis on Covid-19 tweets. This paper introduces advanced sentiment analysis techniques, such as LDA, Count Vectorizer, and SVM, enhancing nuanced sentiment analysis during the COVID-19 pandemic with notable 85% accuracy in SVM classification.

Data Lake Management System based on Topic Modeling

In an environment full of competitiveness, data is a valuable asset for any company looking to grow. It represents a real competitive economic and strategic lever. The most reputable companies are not only concerned with collecting data from heterogeneous data sources, but also with analyzing and transforming these datasets into better decision-making. In this context, the data lake continues to be a powerful solution for storing large amounts of data and providing data analytics for decision support. In this paper, we examine the intelligent data lake management system that addresses the drawbacks of traditional business intelligence, which is no longer capable of handling data-driven demands. Data lakes are highly suitable for analyzing data from a variety of sources, particularly when data cleaning is time-consuming. However, ingesting heterogeneous data sources without any schema represents a major issue, and a data lake can easily turn into a data swamp. In this study, we implement the LDA topic model for managing the storage, processing, analysis, and visualization of big data. To assess the usefulness of our proposal, we evaluated its performance based on the topic coherence metric. The results of these experiments showed our approach to be more accurate on the tested datasets

PEMODELAN TOPIK TERKAIT BANJIR PADA TWITTER DENGAN MENGGUNAKAN LATENT DIRICHLET ALLOCATION

In this background discusses the topic of tweet about Flooding on Twitter using the keyword "Flood". Tweet data was taken from June 1, 2021 to June 2, 2021 with the number of tweet data obtained, which was 2000 tweets. The number of tweets related to flooding has not been analyzed so that the topics contained in it are not yet known. Research . Modeling topics related to floods in Indonesia on Twitter social media with the LDA method. Research. This study uses experimental methods with several variables to test hypotheses. Then the data is processed with stages, namely web data extraction, preprocessing, feature extraction, topic modeling using latent dirichlet allocation algorithms, visualization, and analysis. Research. The results of the topic coherence stage were carried out a search for the most optimal topic from the 20 topics that had been determined at the beginning. The results of topic coherence for 20 topics concluded that for topic 10 it has a total topic value of 0.41 and has an ideal topic modeling result and is in accordance with the provisions. Conclusion : Based on the results of the discussion of topic coherence, it can be concluded that the most ideal number of topics is topic 10 because it has the highest value compared to other topics. The advice here is to be able to display or get flood information in Indonesia in real time and accurately.