Learning Text-image Joint Embedding for Efficient Cross-modal Retrieval with Deep Feature Engineering

This article introduces a two-phase deep feature engineering framework for efficient learning of semantics enhanced joint embedding, which clearly separates the deep feature engineering in data preprocessing from training the text-image joint embedding model. We use the Recipe1M dataset for the technical description and empirical validation. In preprocessing, we perform deep feature engineering by combining deep feature engineering with semantic context features derived from raw text-image input data. We leverage LSTM to identify key terms, deep NLP models from the BERT family, TextRank, or TF-IDF to produce ranking scores for key terms before generating the vector representation for each key term by using Word2vec. We leverage Wide ResNet50 and Word2vec to extract and encode the image category semantics of food images to help semantic alignment of the learned recipe and image embeddings in the joint latent space. In joint embedding learning, we perform deep feature engineering by optimizing the batch-hard triplet loss function with soft-margin and double negative sampling, taking into account also the category-based alignment loss and discriminator-based alignment loss. Extensive experiments demonstrate that our SEJE approach with deep feature engineering significantly outperforms the state-of-the-art approaches.

Context Matters in Evidence Implementation Globally; Comment on "Stakeholder Perspectives of Attributes and Features of Context Relevant to Knowledge Translation in Health Settings: A Multi-Country Analysis"

Context influences the effectiveness of healthcare interventions and should be considered to inform their implementation. However, context remains poorly defined in the knowledge translation (KT) literature. The paper by Squires and colleagues constitutes a valuable contribution to the field of KT as it provides the basis for a comprehensive framework to assess the influence of context on implementation success. In their study, Squires et al. identified 66 context features, grouped into 16 attributes. Their findings highlight a great convergence in the context features mentioned by stakeholders across countries, experience levels and roles in KT. Thus, the proposed framework could eventually transfer to several implementation settings. However, all study participants were from high-income countries. It would therefore be important to replicate this research in low- and middle-income countries. A common understanding of what context means is essential to assessing its influence on the implementation of healthcare interventions globally.

Parallel Cleaning Algorithm for Similar Duplicate Chinese Data Based on BERT

Data is an important source of knowledge discovery, but the existence of similar duplicate data not only increases the redundancy of the database but also affects the subsequent data mining work. Cleaning similar duplicate data is helpful to improve work efficiency. Based on the complexity of the Chinese language and the bottleneck of the single machine system to large-scale data computing performance, this paper proposes a Chinese data cleaning method that combines the BERT model and a k-means clustering algorithm and gives a parallel implementation scheme of the algorithm. In the process of text to vector, the position vector is introduced to obtain the context features of words, and the vector is dynamically adjusted according to the semantics so that the polysemous words can obtain different vector representations in different contexts. At the same time, the parallel implementation of the process is designed based on Hadoop. After that, k-means clustering algorithm is used to cluster similar duplicate data to achieve the purpose of cleaning. Experimental results on a variety of data sets show that the parallel cleaning algorithm proposed in this paper not only has good speedup and scalability but also improves the precision and recall of similar duplicate data cleaning, which will be of great significance for subsequent data mining.

Building and Evaluating an Annotated Corpus for Automated Recognition of Chat-Based Social Engineering Attacks

Chat-based Social Engineering (CSE) is widely recognized as a key factor to successful cyber-attacks, especially in small and medium-sized enterprise (SME) environments. Despite the interest in preventing CSE attacks, few studies have considered the specific features of the language used by the attackers. This work contributes to the area of early-stage automated CSE attack recognition by proposing an approach for building and annotating a specific-purpose corpus and presenting its application in the CSE domain. The resulting CSE corpus is then evaluated by training a bi-directional long short-term memory (bi-LSTM) neural network for the purpose of named entity recognition (NER). The results of this study emphasize the importance of adding a plethora of metadata to a dataset to provide critical in-context features and produce a corpus that broadens our understanding of the tactics used by social engineers. The outcomes can be applied to dedicated cyber-defence mechanisms utilized to protect SME employees using Electronic Medium Communication (EMC) software.

DeepGCNs-Att: Point cloud semantic segmentation with contextual point representations

Graph Convolutional Networks are able to characterize non-Euclidean spaces effectively compared with traditional Convolutional Neural Networks, which can extract the local features of the point cloud using deep neural networks, but it cannot make full use of the global features of the point cloud for semantic segmentation. To solve this problem, this paper proposes a novel network structure called DeepGCNs-Att that enables deep Graph Convolutional Network to aggregate global context features efficiently. Moreover, to speed up the computation, we add an Attention layer after the Graph Convolutional Network Backbone Block to mutually enhance the connection between the distant points of the non-Euclidean space. Our model is tested on the standard benchmark S3DIS. By comparing with other deep Graph Convolutional Networks, our DeepGCNs-Att’s mIoU has at least two percent higher than that of all other models and even shows excellent results in space complexity and computational complexity under the same number of Graph Convolutional Network layers.

A habenula-insular circuit encodes the willingness to act

The decision that it is worth doing something rather than nothing is a core yet understudied feature of voluntary behaviour. Here we study “willingness to act”, the probability of making a response given the context. Human volunteers encountered opportunities to make effortful actions in order to receive rewards, while watching a movie inside a 7 T MRI scanner. Reward and other context features determined willingness-to-act. Activity in the habenula tracked trial-by-trial variation in participants’ willingness-to-act. The anterior insula encoded individual environment features that determined this willingness. We identify a multi-layered network in which contextual information is encoded in the anterior insula, converges on the habenula, and is then transmitted to the supplementary motor area, where the decision is made to either act or refrain from acting via the nigrostriatal pathway.

Research on the Characteristics and Usefulness of User Reviews of Online Mental Health Consultation Services: A Content Analysis

Online consultation based on Internet technology is gradually becoming the main way to seek health information and professional assistance. Online user reviews, such as content reviews and star ratings, are an important basis for reflecting users’ views on the effectiveness of health services. Here, we used user reviews related to online psychological consultation services for content feature mining and usefulness analyses. We used a professional online psychological counseling service platform in China to collect user reviews that were liked by users as a data sample for a content analysis. An LDA topic model, dictionary-based sentiment analysis, and the NRC Word-Emotion Association Lexicon were used to extract the topic, sentiment, and context features of the content of 4254 useful reviews, and the influence of these features on the usefulness of the reviews was verified by a multiple linear regression analysis. Our results show that the content of online reviews by psychological counseling users presented a positive emotional attitude as a whole and expressed more views on the process, effects, and future expectations of counseling than on other topics. There was a significant correlation between the topic, sentiment, and context features of a user review and its usefulness: reviews giving high scores and containing topics such as “ease emotions” and “consulting expectations” received more user likes. However, the usefulness of a review was significantly reduced if it was in existence for too long. This research provides valuable suggestions for understanding the needs and emotional attitudes of users with mental health problems in terms of online psychological consultation; identifying the factors that affect the number of likes a review receives can help platform users write better consultation evaluations and thereby provide greater usefulness. In addition, the use of online reviews generated by users for content analysis effectively supplements the current research on online psychological counseling in terms of data and methods.