Topic Features Research Articles

Improving document-level event detection with event relation graph

The correlation between events within the same document plays a crucial role in event detection. Most existing detection models often ignore event correlations, which is not applicable to multi-event detection at the document level. In the real world, it is a common phenomenon that the probability of correlated events occurring simultaneously is much greater than the probability of uncorrelated events occurring simultaneously. Based on this observation, we propose an event correlation-based document-level event detection model (EventCo-ED) to capture the document-level association between events. Specifically, EventCo-ED first constructs a novel event relation graph (ERG) to capture the correlation between events and uses this correlation to extract the topic features of a document. Secondly, DMBERT is employed to get sentence-level contextual representation as the local features. Finally, a gated feature fusion module is used to aggregate topic features and local features, and a correlation suppression module is used to increase the probability that related events are detected simultaneously and suppress the probability that unrelated events are detected simultaneously. Experimental results show that the proposed model can simultaneously improve the precision and recall of multi-event detection and achieve 1.56% and 3.63% F1 improvements on the LEVEN and MAVEN corpuses, respectively.

Semantic Tagging for the Urdu Language: Annotated Corpus and Multi-Target Classification Methods

Extracting and analysing meaning-related information from natural language data has attracted the attention of researchers in various fields, such as natural language processing, corpus linguistics, information retrieval, and data science. An important aspect of such automatic information extraction and analysis is the annotation of language data using semantic tagging tools. Different semantic tagging tools have been designed to carry out various levels of semantic analysis, for instance, named entity recognition and disambiguation, sentiment analysis, word sense disambiguation, content analysis, and semantic role labelling. Common to all of these tasks, in the supervised setting, is the requirement for a manually semantically annotated corpus, which acts as a knowledge base from which to train and test potential word and phrase-level sense annotations. Many benchmark corpora have been developed for various semantic tagging tasks, but most are for English and other European languages. There is a dearth of semantically annotated corpora for the Urdu language, which is widely spoken and used around the world. To fill this gap, this study presents a large benchmark corpus and methods for the semantic tagging task for the Urdu language. The proposed corpus contains 8,000 tokens in the following domains or genres: news, social media, Wikipedia, and historical text (each domain having 2K tokens). The corpus has been manually annotated with 21 major semantic fields and 232 sub-fields with the USAS (UCREL Semantic Analysis System) semantic taxonomy which provides a comprehensive set of semantic fields for coarse-grained annotation. Each word in our proposed corpus has been annotated with at least one and up to nine semantic field tags to provide a detailed semantic analysis of the language data, which allowed us to treat the problem of semantic tagging as a supervised multi-target classification task. To demonstrate how our proposed corpus can be used for the development and evaluation of Urdu semantic tagging methods, we extracted local, topical and semantic features from the proposed corpus and applied seven different supervised multi-target classifiers to them. Results show an accuracy of 94% on our proposed corpus which is free and publicly available to download.

Evaluation of different hemp (Cannabis sativa L.) progenies resulting from crosses with focus on oil content and seed yield

One of the main breeding objectives for hemp (Cannabis sativa L.) cultivation is higher seed yield important also for the oil yields. However, there has been limited research on hemp breeding that focuses on these topical features. In this study, 17 crosses were created using 12 different hemp varieties. For each variety, five female inflorescences were selected for targeted crossing and one for self-pollination. Seeds from the crosses were sown in separate plots with a seeding density of 30 kg/ha. The number of male and female plants, plant height, internode length, steam thickness, inflorescence length, number of inflorescences per plant, and plant number per plot were measured in crossbred family plants. The absolute weight of seeds per plot, the degree of marbling and oil content were determined. The progeny of the cross between Helena and Tiborszallasi were found to have the highest 100 seed mass (1.546 g, upper half). The offspring of self-pollinated varieties Finola, Tiborszallasi, Fedora 17 and the cross between Helena and Tiborszallasi produced the highest oil content. The progeny of the cross between Futura 75 and KC Dora showed the most promising results in terms of high yield production. This study provided the first important data pertaining to hemp breeding with the goal of increasing oil content and seed yield.

Interpretable fake news detection with topic and deep variational models

The growing societal dependence on social media and user generated content for news and information has increased the influence of unreliable sources and fake content, which muddles public discourse and lessens trust in the media. Validating the credibility of such information is a difficult task that is susceptible to confirmation bias, leading to the development of algorithmic techniques to distinguish between fake and real news. However, most existing methods are challenging to interpret, making it difficult to establish trust in predictions, and make assumptions that are unrealistic in many real-world scenarios, e.g., the availability of audiovisual features or provenance. In this work, we focus on fake news detection of textual content using interpretable features and methods. In particular, we have developed a deep probabilistic model that integrates a dense representation of textual news using a variational autoencoder and bi-directional Long Short-Term Memory (LSTM) networks with semantic topic-related features inferred from a Bayesian admixture model. Extensive experimental studies with 3 real-world datasets demonstrate that our model achieves comparable performance to state-of-the-art competing models while facilitating model interpretability from the learned topics. Finally, we have conducted model ablation studies to justify the effectiveness and accuracy of integrating neural embeddings and topic features both quantitatively by evaluating performance and qualitatively through separability in lower dimensional embeddings.

Analysis of customer satisfaction of HIMO Photo Studio based on Ordered Regresssion

With the upgrading of consumption and the improvement of living standards, a group of new type photo studios represented by HIMO Photo Studio have arisen and developed dramatically and become the benchmark of the era in the portrait photography industry. In this paper, we apply LDA topic classification and TF-IDF feature extraction method to conduct text feature extraction and encoding vectorization on the comment data of Beijing HIMO Photo Studio. An ordered regression model with customer satisfaction as the responding variable is established, and the interaction effect is also introduced into model to study the influencing factors of customer satisfaction creatively and quantitatively so that we can give suggestions for improving service quality. The empirical results show that factors such as photo season, shop type, waiting time, photo works, service attitude, photo subject, photo category and so on have a significant impact on customer satisfaction and there are significant differences in customer satisfaction with the same type of photos in different shops. Combining the business goals of HIMO Photo Studio and the industry needs of today's new type photo studios, this paper puts forward targeted suggestions for HIMO Photo Studio to further improve its services according to the conclusions of the model.

A Cross-Comparative Analysis of Transportation Safety Research

This article aims to identify the main trends of safety research in relation to different transportation modes (aviation, rail, road, and ship) based on the chosen methods, emphasizing the same and different features in the research topics. In the context of the main goal, quantitative content analysis was chosen, which identified the trends and domains of safety research, and text-mining software Leximancer was used for processing concept maps and reports showing semantic structures. The article found that the dominant topics in all modes of transportation are safety issues, mostly related to incidents/accidents. All modes of transportation have similarities in research (e.g., management, safety, and legislation are discussed), but differences regarding the specific operation prevail. The article emphasizes what trends in safety research dominate and how they differ. It has also been proven that, even though it is safety research, security issues can also appear because the two concepts form an indivisible complex in many ways.

Do learners use different topical progression patterns in L2 Chinese argumentative spoken and written discourse?

Abstract The current study examines how L2 Chinese learners develop their topics in spoken and written discourse. Currently, researchers have mainly analyzed topical progression features in written texts. Topical progression in speech is rarely examined. We know little about whether learners may advance their topics differently in spoken and written discourse. The current study investigates how advanced Chinese learners apply topical progressions, such as parallel or sequential progression, in argumentative spoken and written discourse, respectively. The study also examines how various topical progressions may relate to the holistic qualities of spoken and written discourse. Topical progression features were analyzed using the Topical Structure Analysis framework. The results showed significant differences in the use of new sequential and related sequential progressions between the two types of discourse. The case study analysis results revealed a multi-Z and single-Z topical movement pattern for higher-score and lower-score performances, respectively.

A GRU-CNN Algorithm Leveraging on User Reviews

Personalized recommendation systems learn user preference characteristics by analyzing behavioral data such as ratings and comments generated by users in the Internet, and provide precise recommendations for individual users accordingly. However, in real life, users often conduct group activities like group buying and traveling together. How to recommend for groups has become a heated research topic in recent years. Most existing group recommendation algorithms are recommended for given divided groups by collectively combining the preferences of members in the group. However, in most cases, users’ group properties are fickle. As the results of group detection are decisive to the performance of group recommendation, group detection is particularly important to the group recommendation algorithm. After analyzing problems of existing group recommendation algorithms, this paper proposes the density peak clustering group detection algorithm based on GRU-CNN and the group recommendation algorithm based on the mechanism. With respect to group detection, most of the existing group detection algorithms suffer from certain deficiencies: First, depending solely on the users’ static preference features while ignoring the variation of users’ interest over time when finding the group structure in the network; second, group division based on users’ topic features extracted from reviews is difficult to support further digging of the in-depth features in reviews. To address the above-mentioned problems, this paper proposes a density peak clustering group detection algorithm based on CNN-GRU. It would first extract representative keywords in the reviews with LDA topic model, and then model time series information based on GRU attaining users’ dynamic topic features. Coupling with deeper characteristics cored out by CNN, density peak clustering algorithm completes its group detection finally. Experiments on real dataset indicate that the features mined by the fusion depth neural network model effectively capture users’ dynamic preferences, and yield better results of group detection than that of existing algorithms.

Comprehensive commodity price forecasting framework using text mining methods

AbstractExploiting advanced and appropriate methods to construct high‐quality features from different types of data becomes crucial in agricultural futures price forecasting. Thus, this study develops a comprehensive commodity price forecasting framework using text mining methods. First, the modal features of the price series are extracted using the proposed Integrated‐EEMD‐VMD‐SE method, and dynamic topic sentiment features are constructed from Weibo texts using the proposed dynamic topic model joint sentiment analysis method. Second, combined with statistical variables, lag order selection and feature selection are performed on these comprehensive factors. Finally, 12 comparative prediction models are designed based on random forest (RF), long short‐term memory (LSTM), and multilayer perceptron (MLP), and empirical analysis is carried out on two cases of pork prices and soybean futures prices. The experimental results show that the proposed prediction framework has high prediction accuracy, and the mean absolute percentage error (MAPE) values are 1.00 and 0.92, respectively. The constructed time series modal features and dynamic topic sentiment features can significantly improve the performance of the prediction model.

Public Attitudes Toward Anxiety Disorder on Sina Weibo: Content Analysis.

Anxiety disorder has become a major clinical and public health problem, causing a significant economic burden worldwide. Public attitudes toward anxiety can impact the psychological state, help-seeking behavior, and social activities of people with anxiety disorder. The purpose of this study was to explore public attitudes toward anxiety disorders and the changing trends of these attitudes by analyzing the posts related to anxiety disorders on Sina Weibo, a Chinese social media platform that has about 582 million users, as well as the psycholinguistic and topical features in the text content of the posts. From April 2018 to March 2022, 325,807 Sina Weibo posts with the keyword "anxiety disorder" were collected and analyzed. First, we analyzed the changing trends in the number and total length of posts every month. Second, a Chinese Linguistic Psychological Text Analysis System (TextMind) was used to analyze the changing trends in the language features of the posts, in which 20 linguistic features were selected and presented. Third, a topic model (biterm topic model) was used for semantic content analysis to identify specific themes in Weibo users' attitudes toward anxiety. The changing trends in the number and the total length of posts indicated that anxiety-related posts significantly increased from April 2018 to March 2022 (R2=0.6512; P<.001 to R2=0.8133; P<.001, respectively) and were greatly impacted by the beginning of a new semester (spring/fall). The analysis of linguistic features showed that the frequency of the cognitive process (R2=0.1782; P=.003), perceptual process (R2=0.1435; P=.008), biological process (R2=0.3225; P<.001), and assent words (R2=0.4412; P<.001) increased significantly over time, while the frequency of the social process words (R2=0.2889; P<.001) decreased significantly, and public anxiety was greatly impacted by the COVID-19 pandemic. Feature correlation analysis showed that the frequencies of words related to work and family are almost negatively correlated with those of other psychological words. Semantic content analysis identified 5 common topical areas: discrimination and stigma, symptoms and physical health, treatment and support, work and social, and family and life. Our results showed that the occurrence probability of the topical area "discrimination and stigma" reached the highest value and averagely accounted for 26.66% in the 4-year period. The occurrence probability of the topical area "family and life" (R2=0.1888; P=.09) decreased over time, while that of the other 4 topical areas increased. The findings of our study indicate that public discrimination and stigma against anxiety disorder remain high, particularly in the aspects of self-denial and negative emotions. People with anxiety disorders should receive more social support to reduce the impact of discrimination and stigma.

One-class learning for fake news detection through multimodal variational autoencoders

Machine learning methods to detect fake news typically use textual features and Binary or Multi-class classification. However, accurately labeling a large news set is still a very costly process. On the other hand, one of the prominent approaches is One-Class Learning (OCL). OCL requires only the labeling of fake news, minimizing data labeling efforts. Although we eliminate the need to label non-interest news, the efficiency of OCL algorithms depends directly on the data representation model adopted. Most existing methods in the OCL literature explore representations based on one modality to detect fake news. However, different text features can be the reason for the news to be fake, such as topic or linguistic features. We model this behavior as different modalities for news to represent different textual feature sets. Thus, we present the MVAE-FakeNews, a multimodal method to represent the texts in the fake news detection through OCL that learns a new representation from the combination of promising modalities for news data: text embeddings, topic, and linguistic information. We used real-world fake news datasets in Portuguese and English in the experimental evaluation. Results show that MVAE-FakeNews obtained a better F1-Score and AUC-ROC, outperforming another fourteen methods in three datasets and getting competitive results on the other three. Moreover, our MVAE-FakeNews, with only 3% of labeled fake news, obtained comparable or higher results than other methods. To improve the experimental evaluation, we also propose the Multimodal LIME for OCL to identify how each modality is associated with the fake news class.

Comment-Guided Semantics-Aware Image Aesthetics Assessment

Existing image aesthetics assessment methods mainly rely on the visual features of images but ignore their rich semantics. Nowadays, with the widespread application of social media, the comments corresponding to images in the form of texts can be easily accessed and provide rich semantic information, which can be utilized to effectively complement image features. This paper proposes a comment-guided semantics-aware image aesthetics assessment method, which is built upon a multi-task learning framework for image aesthetics prediction and comment-guided semantics classification. To assist image aesthetics assessment, we first model the semantics of an image as the topic features of its corresponding comments using Latent Dirichlet Allocation. We then propose a two-stream multitask learning framework for both topic feature prediction and aesthetic score distribution prediction. Topic feature prediction task enables to infer the semantics from images, since the comments are usually unavailable during inference and comment-guided semantics can only serve as supervision during training. We further propose to deeply fuse aesthetics and semantic features using a layerwise feature fusion method. Experimental results demonstrate that the proposed method outperforms state-of-the-art image aesthetics assessment methods.

Understanding Mental Health Organizations’ Instagram Through Visuals: A Content Analysis

ABSTRACT This study analyzed the content, visual features, and audience engagement data of Instagram posts from two mental health organizations over one year (N = 725). For content features, mental health literacy and communicative strategies were examined. Posts that promoted knowledge of mental disorders and treatments, used information and community strategy were more likely to receive higher audience engagement. Visual features of demographic segments, visual composition, and visual framing topics were analyzed. Images that covered an unspecific population, used illustrated images, and focused on anti-stigma topical frames obtained more engagement. Theoretical contributions and practical applications regarding visual message design and management on social media to promote mental health are also offered.

THE ROLE OF CLITIC DOUBLING IN THE INFORMATION STRUCTURE OF THE ALBANIAN LANGUAGE

This paper attempts to argue whether the repetition of the object or clitic doubling as a syntactic mechanism affects the encoding of the information structure in the Albanian language. In three previous studies by Kallulli (2000, 2008); Kapia (2010), it is reported that direct object clitic doubling is not an optional phenomenon in the Albanian language, that direct object clitic doubling systematically produces and encodes the information structure, and that clitic doubled objects are interpreted as topics. Thus, the object and the clitic have the status of topic in the sentence, as doubling occurs when the clitic refers to the already present discourse referent. That is, the NP is previously introduced in the discourse, thus becoming a specific referent available to be later selected via the doubling mechanism. In addition, according to them, when the object bears the topic feature, the direct object doubling is obligatory. Unlike Kallulli (2000, 2008) and Kapia (2010, 2012), Buchholz &amp; Fiedler (1987) draw different conclusions, according to which the direct object clitic doubling in the Albanian language may be optional if the word order is canonical (SVO). This paper, on the other hand, aims at proving whether the object clitic doubling in our corpus marks the information structure by giving the feature [+ topic] or [-focus].

News Recommendation System Based on Topic Embedding and Knowledge Embedding

News recommendation system is designed to deal with massive news and provide personalized recommendations for users. Accurately capturing user preferences and modeling news and users is the key to news recommendation. In this paper, we propose a new framework, news recommendation system based on topic embedding and knowledge embedding (NRTK). NRTK handle news titles that users have clicked on from two perspectives to obtain news and user representation embedding :1) extracting explicit and latent topic features from news and mining users' preferences for them in historical behaviors; 2) extracting entities and propagating users' potential preferences in the knowledge graph. Experiments in a real-world dataset validate the effectiveness and efficiency of our approach.

News Recommendation Based on User Topic and Entity Preferences in Historical Behavior

A news-recommendation system is designed to deal with massive amounts of news and provide personalized recommendations for users. Accurately modeling of news and users is the key to news recommendation. Researchers usually use auxiliary information such as social networks or item attributes to learn about news and user representation. However, existing recommendation systems neglect to explore the rich topics in the news. This paper considered the knowledge graph as the source of side information. Meanwhile, we used user topic preferences to improve recommendation performance. We proposed a new framework called NRTEH that was based on topic and entity preferences in user historical behavior. The core of our approach was the news encoder and the user encoder. Two encoders in NRTEH handled news titles from two perspectives to obtain news and user representation embedding: (1) extracting explicit and latent topic features from news and mining user preferences for them; and (2) extracting entities and propagating users’ potential preferences in the knowledge graph. Experiments on a real-world dataset validated the effectiveness and efficiency of our approach.

A cross-modal conditional mechanism based on attention for text-video retrieval.

Current research in cross-modal retrieval has primarily focused on aligning the global features of videos and sentences. However, video conveys a much more comprehensive range of information than text. Thus, text-video matching should focus on the similarities between frames containing critical information and text semantics. This paper proposes a cross-modal conditional feature aggregation model based on the attention mechanism. It includes two innovative modules: (1) A cross-modal attentional feature aggregation module, which uses the semantic text features as conditional projections to extract the most relevant features from the video frames. It aggregates these frame features to form global video features. (2) A global-local similarity calculation module calculates similarities at two granularities (video-sentence and frame-word features) to consider both the topic and detail features in the text-video matching process. Our experiments on the four widely used MSR-VTT, LSMDC, MSVD and DiDeMo datasets demonstrate the effectiveness of our model and its superiority over state-of-the-art methods. The results show that the cross-modal attention aggregation approach can effectively capture the primary semantic information of the video. At the same time, the global-local similarity calculation model can accurately match text and video based on topic and detail features.

Морфофункциональные особенности ягодичной группы мышц разгибателей и ротаторов тазобедренного сустава у овцы

The article presents anatomical features of the gluteal group of hip extensors in domestic sheep that are absent in the available literature. It was found that the muscles of the gluteal extensor group of sheep belong to different morphological types: the middle gluteal is dynamic, the additional gluteal is statodynamic, and the deep gluteal is dynamostatic. The material for the research was a sectional material - pelvic limbs (n=10), selected from adult sheep of the Dagestan mountain breed, without external signs of pathologies of the musculoskeletal system. We used methods of fine macro- and microanatomic dissection under the control of a binocular magnifier «Micromed HR 350 S», followed by functional analysis of the studied structures. Shown.that the main extensor of the hip joint is the middle gluteal muscle, which is confirmed by its topical features, the orientation of its bundles of muscle fibers and belonging to the dynamic type in its internal structure. The topical features of anatomical formations on the pelvic (nameless) and femoral bones are clarified – «the line of the double muscle», «the line of the external locking muscle» on the ventral surface of the sciatic bone; «the site of the additional gluteal muscle», «the tubercle of the deep gluteal muscle» on the large trochanter of the femur.

Анатомо-топографические особенности заднебедренной группы мышц тазобедренного сустава у овцы

The article presents anatomical and topographic features of the posterior femoral group of hip extensors in small cattle – sheep of Dagestan mountain breed, and also clarified the topical features of anatomical bone formations of the sciatic hillock, in particular the presence of "dorsal, lateral and medial outgrowths" on it. The material for the research was a sectional material - pelvic limbs (n=8), selected from adult sheep of the Dagestan mountain breed, without external signs of pathologies of the musculoskeletal system. The methods of fine macro- and microanatomic dissection under the control of a binocular magnifier «Micromed HR 350 S» were used, followed by functional analysis of the studied structures and skeletotopic projection of points of attachment of parts of the muscle. The established sites of attachment of the semi-tendon muscle confirm its functions as an optional extensor of the hip and metatarsal joints, and the main function is associated with flexion, pronation and adduction of the knee joint and, accordingly, the proximal part of the limb. The revealed anatomical features of the studied muscles may reflect their functional relationship. The data obtained are reference in assessing the structural and functional state of the muscles of the posterior femoral group of hip extensors in domestic sheep.

Анатомо-топографическая характеристика мышц тазовой конечности у тигра

The aim of the present study was to establish anatomical and topographic features of the muscles of the thigh region of the Amur tiger and to assess the functional significance of the studied structures. The research was carried out at the Department of Anatomy and Histology of Animals named after Professor A. F. Klimov of the Moscow State Academy of Veterinary Medicine and Biotechnology – MBA named after K.I. Scriabin. The object of study was the Amur tiger. The material is pelvic limbs of mature animals, without external signs of pathologies. We used methods of conventional and fine anatomical dissection of muscles under the control of a binocular magnifier «Micromed HR 350 S», as well as biomechanical modeling of a static-locomotor act. The specific anatomical and topographic features of skeletal thigh muscles in the Amur tiger are presented. Thus, the first muscle layer, located between the leaves of the deep fascia, is represented from the lateral surface by the biceps femoris muscle, with the medial – tailor's and slender muscles. The topical features of the fixation of muscles acting on the knee and metatarsal joints, which are absent in the available literature, are presented. The features indicating the anatomical and functional integration of the organs of the musculoskeletal system of the pelvic limb in a representative of the feline family (the tiger), which must be taken into account when assessing the status-locomotor act, in the diagnosis of arthro-desmo and myopathologies, are revealed.