In the realm of collaborative filtering recommendation systems, Graph Neural Networks (GNNs) have demonstrated remarkable performance but face significant challenges in deployment on resource-constrained edge devices due to their high embedding parameter requirements and computational costs. Using common quantization method directly on node embeddings may overlooks their graph based structure, causing error accumulation during message passing and degrading the quality of quantized embeddings.To address this, we propose Graph based Node-Aware Dynamic Quantization training for collaborative filtering (GNAQ), a novel quantization approach that leverages graph structural information to enhance the balance between efficiency and accuracy of GNNs for Top-K recommendation. GNAQ introduces a node-aware dynamic quantization strategy that adapts quantization scales to individual node embeddings by incorporating graph interaction relationships. Specifically, it initializes quantization intervals based on node-wise feature distributions and dynamically refines them through message passing in GNN layers. This approach mitigates information loss caused by fixed quantization scales and captures hierarchical semantic features in user-item interaction graphs. Additionally, GNAQ employs graph relation-aware gradient estimation to replace traditional straight-through estimators, ensuring more accurate gradient propagation during training. Extensive experiments on four real-world datasets demonstrate that GNAQ outperforms state-of-the-art quantization methods, including BiGeaR and N2UQ, by achieving average improvement in 27.8% Recall@10 and 17.6% NDCG@10 under 2-bit quantization. In particular, GNAQ is capable of maintaining the performance of full-precision models while reducing their model sizes by 8 to 12 times; in addition, the training time is twice as fast compared to quantization baseline methods.

With the prevalence and rapid proliferation of SNS, dissemination of false information has become a big problem. In this paper, targeting Twitter, we propose a two-step approach for early detection of false information based on machine learning, which considers the event occurrence scale and the time series of tweets that compose the event. In Step 1, in the early stage of an event, whether it is false or true is decided if the prediction probability is high enough. In Step 2, the events whose authenticity cannot be determined in Step 1 are targeted for tracking, and their authenticity is ascertained as the tweets related to the events increase gradually. The experimental results comparing five machine learning models show that SVM is the optimal model for both steps and that our approach can achieve early detection of false information.

In recent years, the application of machine translation has become more and more widely. Currently, the neural multimodal translation models have made attractive progress, which combines images into deep learning networks, such as Transformer and RNN. When considering images in translation models, they directly apply gate structure or image attention to introduce image feature to enhance the translation effect. We argue that it may mismatch the text and image features since they are in different semantic space. In this paper, we propose a coordinated representation learning enhanced multimodal machine translation approach with multimodal attention. Our approach accepts the text data and its relevant image data as the input. The image features are fed into the decoder side of the basic Transformer model. Moreover, the Coordinated Representation Learning is utilized to map the different text and image modal features into their semantic representations. The mapped representations are linearly related in a shared semantic space. Finally, the sum of the image and text representations, called Coordinated Visual-Semantic Representation (CVSR), will be sent to a Multimodal Attention Layer (MAL) in our Transformer based translation approach. Experimental results show that our approach achieves the state-of-art performance on the public Multi30k dataset.

Task assignment is one of the important issues in crowdsourcing. Most existing schemes consider both of task-centric measures (e.g., required abilities) and worker-centric measures (e.g., worker preference) at local (each task) assignment level, but consider only task-centric measures at the global (the whole workflow) assignment level, such as productivity and throughput. This paper proposes to introduce Inclusiveness and Activity Degree as the worker-centric measures at the global level. It is not trivial whether we can find assignment that are good in terms of all of the two task-centric measures (productivity and throughput) and the two worker-centric measures (Inclusiveness and Activity Degree) at the global level. This paper explains five assignments that are expected to increase some measures and shows Activity Degree Conscious Assignment can produce high Inclusiveness and Activity Degree and not significantly reduce the productivity and throughput assignment through a simulation.

Worker-task assignments represent one of the critical issues in crowdsourcing, as they affect the quality of task results. This study addresses the problem of forming worker groups assigned to the same task in a task stream that requires more than one worker. We introduce a worker-group queue model that covers practical and common scenarios for task-stream crowdsourcing, and compare three strategies in terms of the skill balance among worker groups, the quality of the final outputs, the number of worker re-assignments of workers, and psychological stress felt by workers. We found that one of the compared strategies that employs multiple worker queues yields good results based on these measures.

Online advertising has become a popular method for companies to market their products and services to potential customers. The methods used by conventional web advertisement systems to decide on which advertisements to display to users in a real-time bidding environment generally do not consider the latent interests of users and as such it is difficult for advertisers to target and acquire new customers with potential interest in the product. Therefore, we proposed the development of a recommender system which could recommend advertisements to users based on their latent interests. In this paper, we outline two experiment studies related to the development of this system. The first study was carried out to examine the effect of using a long and short browsing history acquisition period to train the user model and predict user interests. The results suggested that a longer browsing history acquisition period did not necessarily result in better predictive performance. The second study examined the use of a categorized web browsing history to predict user interest. The results showed the accuracy of the classifiers increased when website categories were used instead of Fully Qualified Domain Names.

A common workflow to perform a continuous human task stream is to divide workers into groups, have one group perform the newly-arrived task, and rotate the groups. Usually, more than one worker belongs to each group for improving the quality of task results. We call this type of workflow the group rotation. This paper addresses the problem of how to manage Group Rotation Type Crowdsourcing, the group rotation in a crowdsourcing setting. In the group-rotation type crowdsourcing, we must change the group structure dynamically because workers come in and leave frequently. However, changing the group structure will give workers psychological stress, such as surprise, confusion or irritation. This paper explores a design space for group restructuring algorithms in the group rotation type crowdsourcing and compares implemented strategies in terms of the evaluation results on psychological stress with real-world crowd workers.

In order to realize smooth communication between people with or without difficulty in hearing, we aim to implement an information support system based on crowdsourcing, a problem-solving model, in which numerous people cooperate to accomplish a job. More specifically, we implement a crowdsourced captioning system that enables groups of deaf and hard-of-hearing people to collaboratively interpret a speech or lecture given by sign language to text in real-time. Deaf and hard-of-hearing people are generally considered as the ones who receive support from others, but our proposal helps them become the ones who offer support to others. In this paper, we explore the issues that need be resolved for the practical use of the proposed system through several experiments and succeed in verifying its feasibility.

Web advertising is watched with interest as an advertising method employed by companies to introduce their products and services. Web advertising includes listing advertisement, which shows advertisements related to a search keyword, and interest-matching advertising, which shows advertisements relevant to a user's search content and browsing history. However, it is difficult to show effective Web advertising to potential purchasers using the technique based on conventional keyword matching. In this paper, we consider a recommender system for Web advertising based on analysis of the user's potential interests. In particular, we focus on a user model with potential interest for a certain website by analyzing browsing history. We introduce a Web advertising recommender system that is based not only based on keyword matching, but also on reported learning results. In addition, we argue the influence of the period for acquisition of the browsing history, which is taken when the users' model is learned.

Identifying important users from social media has recently attracted much attention in information and knowledge management community. Although researchers have focused on users' knowledge levels on certain topics or influence degrees on other users in social networks, previous works have not studied users' prediction ability on future popularity. In this paper, we propose a novel approach to find important bloggers based on their buzzword prediction ability. We conduct a time-series analysis in the blogosphere considering four factors: post earliness, content similarity, entry frequency and buzzword coverage. We perform preparatory work in categorizing a blogger into knowledgeable categories, identifying past buzzwords, analyzing a buzzword's peak time content and growth period, and finally evaluate a blogger's prediction ability on a buzzword and on a category. Experimental results on real-world blog data consisting of 150 million entries from 11 million bloggers demonstrate that the proposed approach can find prophetic bloggers and outperforms others that do not take temporal features into account.

We present a novel system that combines the advantages of social communication and Web search by simultaneously discovering important pages and users. First, the system provides a communication interface attached to pages, which allows users to talk with each other in real time while browsing the same page, i.e., page-centric communication. Then, the system can efficiently provide two ranking lists of pages and users by analyzing a hybrid structure of hyperlinks (page-page relationship) and social links (page-user relationship and user-user relationship). Thus, users can efficiently search for important pages as well as important users related to their queries through the ranking function, and immediately obtain useful information or knowledge from not only pages themselves but also from other users.