Sequential recommender systems (SRSs) aim to recommend the next items to well match users' preferences. In addition to recommendation accuracy, diversity is another critical aspect in evaluating SRSs. Recently, the emerging diffusion models (DMs) have been widely adopted in SRSs. Their employed learning-to-generate paradigm allows them to cover a much broader range of users' preferences and thus generate more diversified items. However, existing DM-based SRSs still face two significant gaps that prevent them from further improving the recommendation diversity: (1) they often rely on non-diversified users' preferences as guidance to direct the training of diffusion networks, restricting networks' ability to generate diverse items; and (2) they are based on a homogeneous diffusion inference mechanism to generate the next items and thus can only accommodate users' major preferences. Such a practice neglects users' heterogeneous preferences towards various types of items, further limiting recommendation diversity. To bridge these two critical gaps and to further unleash the potential of DMs in enhancing the recommendation diversity of SRSs, we propose a novel diversity-guided diffusion model for sequential recommendations, called DiffDiv for short. To be specific, first, a new diversity-aware guidance learning mechanism is devised to direct the training of DMs to effectively capture users' diversified preferences from their historical interactions. Then, a novel heterogeneous diffusion inference mechanism is designed to generate diversified items to accommodate users' heterogeneous preferences, further boosting the recommendation diversity. Extensive experiments on real-world datasets validate the effectiveness of DiffDiv in terms of both recommendation accuracy and diversity.

Story Ending Generation is a task of generating a coherent and sensible ending for a given story. The key challenges of this task are i) how to obtain a good understanding of context, ii) how to capture hidden information between lines, and iii) how to obtain causal progression. However, recent machine learning models can only partially address these challenges due to the lack of causal entailment and consistency. The key novelty in our proposed approach is to capture the hidden story by generating transitional commonsense sentences between each adjacent context sentence, which substantially enriches causal and consistent story flow. Specifically, we adopt a soft causal relation using people’s everyday commonsense knowledge to mimic the cognitive understanding process of readers. We then enrich the story with causal reasoning and utilize dependency parsing to capture long range text relations. Finally, we apply multi-level Graph Convolutional Networks to deliver enriched contextual information across different layers. Both automatic and human evaluation results show that our proposed model can significantly improve the quality of generated story endings.

E-commerce platforms heavily rely on automatic personalized recommender systems, e.g., collaborative filtering models, to improve customer experience. Some hybrid models have been proposed recently to address the deficiency of existing models. However, their performances drop significantly when the dataset is sparse. Most of the recent works failed to fully address this shortcoming. At most, some of them only tried to alleviate the problem by considering either user side or item side content information. In this article, we propose a novel recommender model called Hybrid Variational Autoencoder (HVAE) to improve the performance on sparse datasets. Different from the existing approaches, we encode both user and item information into a latent space for semantic relevance measurement. In parallel, we utilize collaborative filtering to find the implicit factors of users and items, and combine their outputs to deliver a hybrid solution. In addition, we compare the performance of Gaussian distribution and multinomial distribution in learning the representations of the textual data. Our experiment results show that HVAE is able to significantly outperform state-of-the-art models with robust performance.

Although matrix factorization and its variants have proved their effectiveness in learning user preferences, they violate the triangle inequality and may fail to capture the inner grained preference information. Thus, metric learning based models have attracted increasing interests in recommender systems. However, most of these models only measure the distance in Euclidean space, though it is known to limit the prediction performance. To address this problem, this paper proposes curvilinear collaborative metric learning (CCML) as a solution. Rather than learning from Euclidean distance, CCML measures the “path” (i.e. curvilinear distance) between the user and items. Therefore, it is able to better infer information (in geometrical representation) and user-item relations behind the observed data points. Under this setting, the relations between the user and items are not linearly reflected by the relation vector. In order to effectively learn the representation of items, we propose to fuse macro features of categories and micro representations of items to form their latent vectors. Our model can be less prone to be affected by the noise commonly presented in datasets by separately learning macro and micro features. Experimental results on eight public recommendation datasets demonstrate that CCML produces a outstanding performance compared with five competitive baselines. Finally, quantitative analysis is included to verify the efficiency and effectiveness of CCML and the macro-micro module.

Due to the popularity of web services on the Internet, it is important to have a clear view of their utilization behaviors. Despite asynchronous service invocations and distributed executions can provide better user experience, our views are blurred by out-of-order and fragmented service logs. Researchers have been trying various methods to reveal emerging service orchestration patterns, but nearly all of them have taken deterministic approaches. Hence, they do not natively cater for incomplete data and noises. In this paper, we propose to address these problems by using topic models aiming to reveal service orchestration patterns from sparse service logs. Probabilistic approaches do not only tolerate data defects, but their associated approximation methods also overcome combinatorial explosion. We first investigate the implications of sparsity on topic models. Secondly, we propose an extended time-series form of susceptible-infectious-recovered model to monitor the dynamics of emerging service orchestrations. We quantify their emerging-potential by estimated effective-reproduction-number, which is obtained incrementally by Bayesian parameter estimations. Guided by our proposed emerging-potential measure, one can profile and categorize emerging service orchestration patterns, and generate automated alerts on upcoming consumption peaks. In practice, our model enables service providers to better allocate their resources to meet demands dynamically. While our findings affirm that biterm topic model can be applied to service logs with short and sparse log entries, the effectiveness of our proposed monitoring solutions is also shown by experiments.

A recent study on collective attention in Twitter shows that an epidemic spreading of hashtags is predominantly driven by external factors. We extend a time-series form of susceptible-infectious-recovered (SIR) model to monitor microblog emerging outbreaks by considering both endogenous and exogenous drivers. In addition, we adopt partially labeled Dirichlet allocation (PLDA) model to generate both background latent topics and hashtag topics. It overcomes the problem of small available samples in hashtag analysis by including related but unlabeled tweets through inference. We standardize hashtag topic contagiousness measure as the estimated effective-reproduction-number R derived from epidemiology. It is obtained by Bayesian parameter estimation. Guided by R, one can profile and categorize emerging topics, and generate alerts on potential outbreaks. Experiment results confirm the effectiveness of this approach.

Due to the popularity of using web services to deliver services on the Web, a clear view of how they are being consumed is becoming critical. Researchers have been trying multiple methods to reveal actual service orchestration patterns from service logs. However, most of the discovery methods have taken deterministic approaches, and hence, they do not provide enough allowance to cater for incomplete data and noises. On the other hand, most investigations do not take combinatorial explosion into consideration leading to scalability problem. Moreover, asynchronous web service invocations and distributed executions also make it difficult to identify service patterns due to the randomness in log record generation. In this paper, probabilistic topic mining class of solutions are applied to reveal web service orchestration patterns from service logs, in which robust approximation methods are available to provide scalability. Data sparsity problem in service log is also investigated by using biterm topic model (BTM) and comparing its results with traditional latent Dirichlet allocation (LDA) model. In addition, a topic matching method is introduced based on the Hungarian method on Jensen-Shannon divergence matrix, whilst notions of aggJSD and autoJSD are also introduced to measure topic diversity between matched topic sets and within a single topic set respectively. Experiment results confirm that BTM can be used for service logs with short log entries and with sparsity larger than 90% approximately.

Due to the popularity of smartphones, finding and recommending suitable services on mobile devices are increasingly important. Recent research has attempted to use role-based approaches to recommend mobile services to other members among the same group in a context dependent manner. However, the traditional role mining approaches originated from the domain of security control tend to be rigid and may not be able to capture human behaviors adequately. In particular, during the course of role mining process, these approaches easily result in over-fitting, i.e., too many roles with slightly different service consumption patterns are found. As a result, they fail to reveal the true common preferences within the user community. This paper proposes an online role mining algorithm with a residual term that automatically group users according to their interests and habits without losing sight of their individual preferences. Moreover, to resolve the over-fitting problem, we relax the role mining mechanism by introducing quasi-roles based on the concept of quasi-bicliques. Most importantly, the new concept allows us to propose a monitoring framework to detect and correct over-fitting in online role mining such that recommendations can be made based on the latest and genuine common preferences. To the best of our knowledge, this is a new area in service recommendation that is yet to be fully explored.