Large Language Models (LLMs) have revolutionized natural language processing, yet concerns persist regarding their tendency to reflect or amplify social biases. This study introduces a novel evaluation framework to uncover gender biases in LLMs: using free-form storytelling to surface biases embedded within the models. A systematic analysis of ten prominent LLMs shows a consistent pattern of overrepresenting female characters across occupations, likely due to supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF). Paradoxically, despite this overrepresentation, the occupational gender distributions produced by these LLMs align more closely with human stereotypes than with real-world labor data. This highlights the challenge and importance of implementing balanced mitigation measures to promote fairness and prevent the establishment of potentially new biases. We release the prompts and LLM-generated stories at GitHub.

Although one-hot encoding is commonly used for multiclass classification, it is not always the most effective encoding mechanism. Error Correcting Output Codes (ECOC) address multiclass classification by mapping each class to a unique codeword used as a label. Traditional ECOC methods rely on manually designed or randomly generated codebooks, which are labor-intensive and may yield suboptimal, dataset-agnostic results. This paper introduces three models for automated codebook learning based on contrastive learning, allowing codebooks to be learned directly and adaptively from data. Across four datasets, our proposed models demonstrate superior robustness to adversarial attacks compared to two baselines. The source is available at GitHub.

Spectral clustering is a popular method for effectively clustering nonlinearly separable data. However, computational limitations, memory requirements, and the inability to perform incremental learning challenge its widespread application. To overcome these limitations, this paper introduces a novel approach called parametric spectral clustering (PSC). By extending the capabilities of spectral clustering, PSC addresses the challenges associated with big data and real-time scenarios and enables efficient incremental clustering with new data points. Experimental evaluations conducted on various open datasets demonstrate the superiority of PSC in terms of computational efficiency while achieving clustering quality mostly comparable to standard spectral clustering. The proposed approach has significant potential for incremental and real-time data analysis applications, facilitating timely and accurate clustering in dynamic and evolving datasets. The findings of this research contribute to the advancement of clustering techniques and open new avenues for efficient and effective data analysis. We publish the experimental code at https://github.com/109502518/PSC_BigData.

Predicting an online user’s future behavior is beneficial for many applications. For example, online retailers may utilize such information to customize the marketing strategy and maximize profit. This paper aims to predict the types of webpages a user is going to click on. We observe that instead of building independent models to predict each individual type of web page, it is more effective to use a unified model to predict a user’s future clicks on different types of web pages simultaneously. The proposed model makes predictions based on the latent variables that represent possible interactions among the multiple targets and among the features. The experimental results show that this method outperforms the carefully tuned single-target training models most of the time. If the size of the training data is limited, the model shows a significant improvement over the baseline models, likely because the hidden relationship among different targets can be discovered by our model.

This study shows that simple supervised learning algorithms can easily predict a user’s personality and demographic information based on the features derived from the users’ browsing logs, even when the logs are not recorded with the finest granularity (i.e., each visited URL of a user). This is different from the analytical formula of Cambridge Analytica (CA), which reported that it needs to know each user’s detailed liked objects (e.g., articles, pages, etc.) on Facebook with a fine granularity (i.e., CA needs to know the liked articles, not only the types of the articles) to predict user information. However, we employed only the visited website categories to predict a user’s gender, age, relationship status, and big six personality scores, which is an authoritative index to represent an individual’s personality in six dimensions. We also show that applying simple clustering as a preprocessing step enhances the predictive power. As a result, the data collectors, even when storing only a coarse granularity of the visited URLs of the users, may leverage such information to identify a user’s preferences/tastes and her/his private information without notifying users.

Matrix factorization (MF) and its extended methodologies have been studied extensively in the community of recommender systems in the last decade. Essentially, MF attempts to search for low-ranked matrices that can (1) best approximate the known rating scores, and (2) maintain low Frobenius norm for the low-ranked matrices to prevent overfitting. Since the two objectives conflict with each other, the common practice is to assign the relative importance weights as the hyper-parameters to these objectives. The two low-ranked matrices returned by MF are often interpreted as the latent factors of a user and the latent factors of an item that would affect the rating of the user on the item. As a result, it is typical that, in the loss function, we assign a regularization weight λ p on the norms of the latent factors for all users, and another regularization weight λ q on the norms of the latent factors for all the items. We argue that such a methodology probably over-simplifies the scenario. Alternatively, we probably should assign lower constraints to the latent factors associated with the items or users that reveal more information, and set higher constraints to the others. In this article, we systematically study this topic. We found that such a simple technique can improve the prediction results of the MF-based approaches based on several public datasets. Specifically, we applied the proposed methodology on three baseline models -- SVD, SVD++, and the NMF models. We found that this technique improves the prediction accuracy for all these baseline models. Perhaps more importantly, this technique better predicts the ratings on the long-tail items, i.e., the items that were rated/viewed/purchased by few users. This suggests that this approach may partially remedy the cold-start issue. The proposed method is very general and can be easily applied on various recommendation models, such as Factorization Machines, Field-aware Factorization Machines, Factorizing Personalized Markov Chains, Prod2Vec, Behavior2Vec, and so on. We release the code for reproducibility. We implemented a Python package that integrates the proposed regularization technique with the SVD, SVD++, and the NMF model. The package can be accessed at https://github.com/ncu-dart/rdf.

Most studies on recommender systems target at increasing the click through rate, and hope that the number of orders will increase as well. We argue that clicking and purchasing an item are different behaviors. Thus, we should probably apply different strategies for different objectives, e.g., increase the click through rate, or increase the order rate. In this article, we propose to generate the distributed representations of users’ viewing and purchasing behaviors on an e-commerce website. By leveraging on the cosine distance between the distributed representations of the behaviors on items under different contexts, we can predict a user’s next clicking or purchasing item more precisely, compared to several baseline methods. Perhaps more importantly, we found that the distributed representations may help discover interesting analogies among the products. We may utilize such analogies to explain how two products are related, and eventually apply different recommendation strategies under different scenarios. We developed the Behavior2Vec library for demonstration. The library can be accessed at https://github.com/ncu-dart/behavior2vec/.

In this article, we explore the relationships among digital objects in terms of their similarity based on vertex similarity measures. We argue that SimRank—a famous similarity measure—and its families, such as P-Rank and SimRank++, fail to capture similar node pairs in certain conditions, especially when two nodes can only reach each other through paths of odd lengths. We present new similarity measures ASCOS and ASCOS++ to address the problem. ASCOS outputs a more complete similarity score than SimRank and SimRank’s families. ASCOS++ enriches ASCOS to include edge weight into the measure, giving all edges and network weights an opportunity to make their contribution. We show that both ASCOS++ and ASCOS can be reformulated and applied on a distributed environment for parallel contribution. Experimental results show that ASCOS++ reports a better score than SimRank and several famous similarity measures. Finally, we re-examine previous use cases of SimRank, and explain appropriate and inappropriate use cases. We suggest future SimRank users following the rules proposed here before naïvely applying it. We also discuss the relationship between ASCOS++ and PageRank.

Source retrieval for plagiarism detection involves using a search engine to retrieve candidate sources of plagiarism for a given suspicious document so that more accurate comparisons can be made. An important consideration is that only documents that are likely to be sources of plagiarism should be retrieved so as to minimize the number of unnecessary comparisons made. A supervised strategy for source retrieval is described whereby search results are classified and ranked as potential sources of plagiarism without retrieving the search result documents and using only the information available at search time. The performance of the supervised method is compared to a baseline method and shown to improve precision by up to 3.28%, recall by up to 2.6% and the F1 score by up to 3.37%. Furthermore, features are analyzed to determine which of them are most important for search result classification with features based on document and search result similarity appearing to be the most important.

Discovering similar objects in a social network has many interesting issues. Here, we present ASCOS, an Asymmetric Structure COntext Similarity measure that captures the similarity scores among any pairs of nodes in a network. The definition of ASCOS is similar to that of the well-known SimRank since both define score values recursively. However, we show that ASCOS outputs a more complete similarity score than SimRank because SimRank (and several of its variations, such as P-Rank and SimFusion) on average ignores half paths between nodes during calculation. To make ASCOS tractable in both computation time and memory usage, we propose two variations of ASCOS: a low rank approximation based approach and an iterative solver Gauss-Seidel for linear equations. When the target network is sparse, the run time and the required computing space of these variations are smaller than computing SimRank and ASCOS directly. In addition, the iterative solver divides the original network into several independent sub-systems so that a multi-core server or a distributed computing environment, such as MapReduce, can efficiently solve the problem. We compare the performance of ASCOS with other global structure based similarity measures, including SimRank, Katz, and LHN. The experimental results based on user evaluation suggest that ASCOS gives better results than other measures. In addition, the asymmetric property has the potential to identify the hierarchical structure of a network. Finally, variations of ASCOS (including one distributed variation) can also reduce computation both in space and time.

With the popularity of online social network services, influence maximization on social networks has drawn much attention in recent years. Most of these studies approximate a greedy based sub-optimal solution by proving the submodular nature of the utility function. Instead of using the analytical techniques, we are interested in solving the diffusion competition and influence maximization problem by a data-driven approach. We propose Information Propagation Game (IPG), a framework that can collect a large number of seed picking strategies for analysis. Through the IPG framework, human players are not only having fun but also helping contributing the seed picking strategies. Preliminary experiment suggests that centrality based heuristics are too simple for seed selection in a multiple player environment.

We introduce the graph vertex similarity measure, Relation Strength Similarity (RSS), that utilizes a network's topology to discover and capture similar vertices. The RSS has the advantage that it is asymmetric; can be used in a weighted network; and has an adjustable "discovery range" parameter that enables exploration of friend of friend connections in a social network. To evaluate RSS we perform experiments on a coauthorship network from the CiteSeerX database. Our method significantly outperforms other vertex similarity measures in terms of the ability to predict future coauthoring behavior among authors in the CiteSeerX database for the near future 0 to 4 years out and reasonably so for 4 to 6 years out.

To improve the search results for socially-connect users, we propose a ranking framework, Social Network Document Rank (SNDocRank). This framework considers both document contents and the similarity between a searcher and document owners in a social network and uses a Multi-level Actor Similarity (MAS) algorithm to efficiently calculate user similarity in a social network. Our experiment results based on YouTube data show that compared with the tf-idf algorithm, the SNDocRank method returns more relevant documents of interest. Our findings suggest that in this framework, a searcher can improve search by joining larger social networks, having more friends, and connecting larger local communities in a social network.