Search

While collections of documents are often annotated with hierarchically structured concepts, the benefits of these structures are rarely taken into account by classification techniques. Within this context, hierarchical text classification methods are devised to take advantage of the labels' organization to boost classification performance. In this work, we aim to deliver an updated overview of the current research in this domain. We begin by defining the task and framing it within the broader text classification area, examining important shared concepts such as text representation. Then, we dive into details regarding the specific task, providing a high-level description of its traditional approaches. We then summarize recently proposed methods, highlighting their main contributions. We also provide statistics for the most commonly used datasets and describe the benefits of using evaluation metrics tailored to hierarchical settings. Finally, a selection of recent proposals is benchmarked against non-hierarchical baselines on five public domain-specific datasets. These datasets, along with our code, are made available for future research. [ABSTRACT FROM AUTHOR]

Politically polarizing issues are a growing concern around the world, creating divisions along ideological lines, which was also confirmed during the 2022 United States midterm elections. The purpose of this study was to explore the relationship between the results of the 2022 U.S. midterm elections and the topics that were covered during the campaign. A dataset consisting of 52,688 tweets in total was created by collecting tweets of senators, representatives and governors who participated in the elections one month before the start of the elections. Using unsupervised machine learning, topic modeling is built on the collected data and visualized to represent topics. Furthermore, supervised machine learning is used to classify tweets to the corresponding political party, whereas sentiment analysis is carried out in order to detect polarity and subjectivity. Tweets from participating politicians, U.S. states and involved parties were found to correlate with polarizing topics. This study hereby explored the relationship between the topics that were creating a divide between Democrats and Republicans during their campaign and the 2022 U.S. midterm election outcomes. This research found that polarizing topics permeated the Twitter (today known as X) campaign, and that all elections were classified as highly subjective. In the Senate and House elections, this classification analysis showed significant misclassification rates of 21.37% and 24.15%, respectively, indicating that Republican tweets often aligned with traditional Democratic narratives. [ABSTRACT FROM AUTHOR]

The detection and evaluation of fatigue levels in drivers play a crucial role in reducing traffic accidents and improving the overall quality of life. However, existing studies in this domain often focus on fatigue detection alone, with limited research on fatigue level evaluation. These limitations include the use of single evaluation methods and relatively low accuracy rates. To address these issues, this paper introduces an innovative approach for determining fatigue driving levels. We employ the Dlib library and fatigue state detection algorithms to develop a novel method specifically designed to assess fatigue levels. Unlike conventional approaches, our method adopts a multi-feature fusion strategy, integrating fatigue features from the eyes, mouth, and head pose. By combining these features, we achieve a more precise evaluation of the driver's fatigue state level. Additionally, we propose a comprehensive evaluation method based on the analytic hierarchy process (AHP) and fuzzy comprehensive evaluation, combined with situational prediction. This approach effectively evaluates the fatigue state level of drivers at specific moments or stages and provides accurate predictions. Furthermore, we optimize the gated recurrent unit (GRU) network using an enhanced marine predator algorithm (MAP), which results in significant improvements in predicting fatigue levels during situational prediction. Experimental results demonstrate a classification accuracy of 92% across various scenarios while maintaining real-time performance. In summary, this paper introduces a novel approach for determining fatigue driving levels through multi-feature fusion. We also incorporate AHP-fuzzy comprehensive evaluation and situational prediction techniques, enhancing the accuracy and reliability of fatigue level evaluation. This research holds both theoretical and practical significance in the field of fatigue driving. [ABSTRACT FROM AUTHOR]

The proliferation of short texts resulting from the rapid advancements of social networks, online communication, and e-commerce has created a pressing need for short text classification in various applications. This paper presents a novel approach for short text classification, which combines a hierarchical heterogeneous graph with latent Dirichlet allocation (LDA) fusion. Our method first models the short text dataset as a hierarchical heterogeneous graph, which incorporates more syntactic and semantic information through a word graph, parts-of-speech (POS) tag graph, and entity graph. We then connected the representation of these three feature maps to derive a comprehensive feature vector for the text. Finally, we used the LDA topic model to adjust the feature weight, enhancing the effectiveness of short text extension. Our experiments demonstrated that our proposed approach has a promising performance in English short text classification, while in Chinese short text classification, although slightly inferior to the LDA + TF-IDF method, it still achieved promising results. [ABSTRACT FROM AUTHOR]

The paper deals with the analysis of conversation transcriptions between customers and agents in a call center of a customer care service. The objective is to support the analysis of text transcription of human-to-human conversations, to obtain reports on customer problems and complaints, and on the way an agent has solved them. The aim is to provide customer care service with a high level of efficiency and user satisfaction. To this aim, topic modeling is considered since it facilitates insightful analysis from large documents and datasets, such as a summarization of the main topics and topic characteristics. This paper presents a performance comparison of four topic modeling algorithms: (i) Latent Dirichlet Allocation (LDA); (ii) Non-negative Matrix Factorization (NMF); (iii) Neural-ProdLDA (Neural LDA) and Contextualized Topic Models (CTM). The comparison study is based on a database containing real conversation transcriptions in Italian Natural Language. Experimental results and different topic evaluation metrics are analyzed in this paper to determine the most suitable model for the case study. The gained knowledge can be exploited by practitioners to identify the optimal strategy and to perform and evaluate topic modeling on Italian natural language transcriptions of human-to-human conversations. This work can be an asset for grounding applications of topic modeling and can be inspiring for similar case studies in the domain of customer care quality. [ABSTRACT FROM AUTHOR]

With artificial intelligence permeating conversations and marketing interactions through digital technologies and media, machine learning models, in particular, natural language processing (NLP) models, have surged in popularity for analyzing unstructured data in marketing. Yet, we do not fully understand which NLP models are appropriate for which marketing applications and what insights can be best derived from them. We review different NLP models and their applications in marketing. We layout the advantages and disadvantages of these models and highlight the conditions under which different models are appropriate in the marketing context. We introduce the latest neural autoencoder NLP models, demonstrate these models to analyze new product announcements and news articles, and provide an empirical comparison of the different autoencoder models along with the statistical NLP models. We discuss the insights from the comparison and offer guidelines for researchers. We outline future extensions of NLP models in marketing. [ABSTRACT FROM AUTHOR]

Highlights: What are the main findings? First, the combination of the methods of machine learning with psychological methods to predict the user's OCEAN personality model could achieve a higher accuracy; Second, the model proposed in this paper that is a combination of LDA plus BP neural network is generally superior to the combination model of the same type. What is the implication of the main finding? First, through digital footprints and understanding the rules of user behavior, user behavior can be predicted and targeted recommendations made; Second, it was found that predicting the user's OCEAN personality model and then their behavior can provide an effective method for micro-directional recommendations in network communication. In the era of big data, the Internet is enmeshed in people's lives and brings conveniences to their production and lives. The analysis of user preferences and behavioral predictions of user data can provide references for optimizing information structure and improving service accuracy. According to the present research, user's behavior on social networking sites has a great correlation with their personality, and the five characteristics of the OCEAN (Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism) personality model can cover all aspects of a user's personality. It is important in identifying a user's OCEAN personality model to analyze their digital footprints left on social networking sites and to extract the rules of users' behavior, and then to make predictions about user behavior. In this paper, the Latent Dirichlet Allocation (LDA) topic model is first used to extract the user's text features. Second, the extracted features are used as sample input for a BP neural network. The results of the user's OCEAN personality model obtained by a questionnaire are used as sample output for a BP neural network. Finally, the neural network is trained. A mapping model between the probability of the user's text topic and their OCEAN personality model is established to predict the latter. The results show that the present approach improves the efficiency and accuracy of such a prediction. [ABSTRACT FROM AUTHOR]

In this paper, we present a method to identify forum posts expressing user intentions in online discussion forums. The results of this task, for example buying intentions, can be exploited for targeted advertising or other marketing tasks. Our method utilizes labeled data from other domains to help the learning task in the target domain by using a Naive Bayes (NB) framework to combine the data statistics. Because the distributions of data vary from domain to domain, it is important to adjust the contributions of different data sources when constructing the learning model, to achieve accurate results. Here, we propose to adjust the parameters of the NB classifier by optimizing an objective, which is equivalent to maximizing the between-class separation, using stochastic gradient descent. Experimental results show that our method outperforms several competitive baselines on a benchmark dataset consisting of forum posts from four domains: Cellphone, Electronics, Camera, and TV. In addition, we explore the possibility of combining NB posteriors computed during the optimization process with another classifier, namely Support Vector Machines. Experimental results show the usefulness of optimized NB class posteriors when using as features for SVMs in the cross-domain settings. [ABSTRACT FROM AUTHOR]