Twitter sentiment analysis provides valuable feedback from public emotion concerning certain events or products. Current research has been focused on obtaining sentiment features from vectorized lexical and syntactic feature from tweets, without further context. In this paper, we demonstrated how vectorized location information could be combined with word embeddings to produce a hybrid representation, which has resulted in an improvement on a tweet sentiment classification task. The location information of the geo-tagged tweets provided further context, which was useful for a sentiment classification task. The tweets investigated contained a set of geo-tagged tweets. The word embeddings of these tweets were combined with the geo-tagged tweets’ vectorized location features to form a sentiment feature set of geo-tagged tweets. The sentiment feature set was incorporated into a convolution neural network and a bi-directional long short-term memory network for the tasks of training and predicting of sentiment classification labels. This hybrid representation is compared with the baseline GloVe model through a few experiments, and the results have shown that the incorporation of vectorized location information has resulted in improvement of the accuracy for the twitter sentiment classification task.

The study of viral diversity is imperative in understanding sequence change and its implications for intervention strategies. The widely used alignment-dependent approaches to study viral diversity are limited in their utility as sequence dissimilarity increases, particularly when expanded to the genus or higher ranks of viral species lineage. Herein, we present an alignment-independent algorithm, implemented as a tool, UNIQmin, to determine the effective viral sequence diversity at any rank of the viral taxonomy lineage. This is done by performing an exhaustive search to generate the minimal set of sequences for a given viral non-redundant sequence dataset. The minimal set is comprised of the smallest possible number of unique sequences required to capture the diversity inherent in the complete set of overlapping k-mers encoded by all the unique sequences in the given dataset. Such dataset compression is possible through the removal of unique sequences, whose entire repertoire of overlapping k-mers can be represented by other sequences, thus rendering them redundant to the collective pool of sequence diversity. A significant reduction, namely ~44%, ~45%, and ~53%, was observed for all reported unique sequences of species Dengue virus, genus Flavivirus, and family Flaviviridae, respectively, while still capturing the entire repertoire of nonamer (9-mer) viral peptidome diversity present in the initial input dataset. The algorithm is scalable for big data as it was applied to ~2.2 million non-redundant sequences of all reported viruses. UNIQmin is open source and publicly available on GitHub. The concept of a minimal set is generic and, thus, potentially applicable to other pathogenic microorganisms of non-viral origin, such as bacteria.

Understanding liquid spray is essential for spray applications, including but not limited to designing fuel-efficient engines. Due to the challenges involved in collecting real-world liquid spray images, a synthetic liquid spray was generated using fluid simulation based on the atomization of the liquid jet. Semantic segmentation was chosen to analyze the liquid spray, as it reflects the precise location of the objects in the image. This paper presents a workflow to train a U-Net with a small sample (only 24 training images) dataset under the constraint that no ground truth is provided. An image is selected from the generated images of liquid spray and edited by randomly masking some objects. After the chosen image is annotated, a data augmentation technique that includes rotation and Gaussian smoothing is applied, resulting in 24 images available as the training set. The RGB original-sized images are fed to U-Net for training. Due to how the liquid spray images are obtained in the real world, Gaussian smoothing is explored as the inductive bias. Gaussian smoothing is incorporated between the convolutional layers of U-Net to enhance its feature extraction ability. The experiment results showed that the segmentation output improved when smoothing was incorporated into the U-Net. By visualizing the convolutional feature map of trained U-Net, we discover that smoothing makes convolution less biased to texture information. Going through this workflow, the trained U-Net is found to generalize well to the test images despite learning from few samples. Code is available at https://github.com/lynerlwl/spray-unet.

Pooling operations in convolutional neural networks often lead to the loss of fine-grained object features, particularly when those objects occupy only a few pixels (e.g., 5 × 5 or smaller). This problem is critical in droplet analysis, where pooling-induced feature loss causes deformation of finegrained droplet boundaries within the encoded representation. When fine-grained droplet features are structurally deformed and vanished in the encoder, the resulting ineffective representation impairs learning and leads to inaccurate droplet segmentation, which severely impairs downstream decision-making. We propose a data augmentation strategy using Gaussian blurring to mitigate this problem. By applying spatial blurring prior to encoding, the blur acts as a contour regularization, strengthening the network learning capability to retain fine-grained droplet structure through subsequent pooling layers and improving robustness against pooling deformation. To further strengthen spatial feature interaction, we integrate a graph convolution module at the bottleneck of a U-Net architecture. The graph is constructed from the bottleneck feature map, with edges weighted by raw intensity values of the corresponding feature elements. This non-parametric, intensity-based graph formulation captures structural relationships among features and complements spatial feature preservation. Our combined method achieves a 16.5% relative improvement in Recall (R), indicating that more fine-grained droplets are correctly segmented compared to the baseline U-Net. Ablation studies confirm that Gaussian blur augmentation is the dominant contributor to these gains, while the GraphSAGE provides complementary refinement. The code is available at: https://github.com/lynerlwl/blur-unet-segmentation.