Unsupervised Text Classification Research Articles

Dynamic building defect categorization through enhanced unsupervised text classification with domain-specific corpus embedding methods

Large amounts of data are often categorized using different systems. In such cases, few-shot and unsupervised text classification are the two main approaches for dynamically classifying text into a single classification. Unsupervised text classification typically exhibits lower performance but requires significantly less data preparation effort and computing resources than the few-shot approach. This study proposes two methods to enhance unsupervised text classification for domain-specific non-English text using improved domain corpus embedding: 1) weighted embedding-based anchor word clustering (wean-Clustering), and 2) cosine-similarity-based classification using a defect corpus that is vectorized by fine-tuned pretrained language models (sim-Classification-ftPLM). The proposed methods were tested on 40,765 Korean building defect complaints and achieved F1 scores of 89.12% and 84.66% respectively, outperforming the state-of-the-art zero-shot (53.79%) and few-shot (72.63%) text classification methods, with minimal data preparation effort and computing resources.

Using Cluster Analysis for Author Classification of Albanian Texts: A Study on the Effectiveness of Stop Words

Cluster analysis is a statistical approach that identifies uniform clusters within data. The closeness of data is measured quantitatively using distance functions. Specifically for text data mining, clustering serves as a method of categorization of words based on the similarity of their occurrence within texts and classifying texts by topics or author. Hierarchical clustering is a powerful technique for identifying natural groupings within datasets, which can be especially useful for unsupervised text classification. This paper aims to utilize cluster analysis to establish Albanian texts clusters by authors. Using agglomerative hierarchical clustering we classify Albanian texts by authors according to the similarity of their word frequency. The similarity of texts is evaluated using cosine and Euclidean distances. Considering two study cases, respectively with and without Albanian stop words we conclude that the best clustering by authors of the Albanian documents is achieved with 87% accuracy using Ward’s method with cosine distance in the case of study by removing stop words.

Mapping indicators of cultural ecosystem services use in urban green spaces based on text classification of geosocial media data

In our increasingly urbanized world, the cultural ecosystem services (CES) provided by urban nature play a crucial role in enabling and maintaining the well-being of urban dwellers. Despite the increased number of studies leveraging geosocial media data for more efficient and socio-cultural-oriented CES assessment, the high complexity and costs associated with existing methods such as manual or automated image classification hinder their application in urban planning and ecosystems management. This study introduces a novel method that draws on the semantic similarity between word2vec word embeddings to classify large volumes of geosocial media textual metadata and quantify indicators of CES use. We demonstrated the applicability of our approach by quantifying spatial patterns of aesthetic appreciation and wildlife recreation in the green spaces of the city of Dresden based on the classification of >50,000 geotagged Instagram and Flickr posts. Moreover, we analyzed and mapped semantic patterns embedded in geosocial media and gained essential insights that can contribute toward a context-dependent assessment of CES use, which in turn can help inform decision making for more sustainable planning and management of urban ecosystems. The performance evaluation of the classification proves the validity of the proposed unsupervised text classification approach as a practical, reliable, and more efficient alternative to laborious and expensive annotation efforts required by manual or supervised classification methods.

Unsupervised Text Classification for Natural Language Interactive Narratives

Natural language interactive narratives are a variant of traditional branching storylines where player actions are expressed in natural language rather than by selecting among choices. Previous efforts have handled the richness of natural language input using machine learning technologies for text classification, bootstrapping supervised machine learning approaches with human-in-the-loop data acquisition or by using expected player input as fake training data. This paper explores a third alternative, where unsupervised text classifiers are used to automatically route player input to the most appropriate storyline branch. We describe the Data-driven Interactive Narrative Engine (DINE), a web-based tool for authoring and deploying natural language interactive narratives. To compare the performance of different algorithms for unsupervised text classification, we collected thousands of user inputs from hundreds of crowdsourced participants playing 25 different scenarios, and hand-annotated them to create a gold-standard test set. Through comparative evaluations, we identified an unsupervised algorithm for narrative text classification that approaches the performance of supervised text classification algorithms. We discuss how this technology supports authors in the rapid creation and deployment of interactive narrative experiences, with authorial burdens similar to that of traditional branching storylines.

Flow My FE the Vendor Said: Exploring Violent and Fraudulent Resource Exchanges on Cryptomarkets for Illicit Drugs

A growing share of illicit drug distribution takes place using cryptomarkets that use encryption and anonymization technologies. The risks of law enforcement intervention and violence are lower here than in off-line traditional drug markets, but with the technological innovations follow new opportunities for stealing and fraud. The sites themselves fall prey to theft and hacking attempts, administrators abscond with users’ funds, and malicious sellers regularly cheat buyers. In this study, we explore the types of theft and fraud that occur on cryptomarkets using multiple data sources: formalized community resources (e.g., guides, tutorials), ethnographic observations of user forums, thematic identification of forum posts using unsupervised text classification, and an expert interview. We find system-based violent predatory resource exchange similar to robberies and process-based fraudulent resource exchange similar to rip-offs. We discuss these offenses conceptually as extensions of common drug-related crimes in the digital world. This contributes to the research on how cryptomarkets work and can improve crime-prevention efforts.

Unsupervised Text Classification and Search using Word Embeddings on a Self-Organizing Map

Unsupervised Text Classification and Search using Word Embeddings on a Self-Organizing Map