Semantic Similarity Research Articles

CLSESSP: Contrastive learning of sentence embedding with strong semantic prototypes

By drawing inspiration from the computer vision domain, contrastive learning has emerged as another paradigm for unsupervised sentence embedding. By using contrastive learning, most recent sentence embedding methods have achieved promising results. However, previous methods focused on constructing positive and negative pairs through traditional data augmentation, which was unable to balance the loss of semantics and the additional features brought by aggressive data augmentation. In the present study, we propose a prototype contrastive learning framework based on a distribution divergence minimization loss function and introduce the concept of strong semantic prototypes. Specifically, with the aid of prompts, our method constructs three semantic prototypes for each instance: a basic semantic prototype, a strong semantic prototype and a negative prototype. By combining InfoNCE loss with distribution divergence minimization loss, Contrastive Learning of Sentence Embedding with Strong Semantic Prototypes (CLSESSP) creates another optimization objective that integrates strong augmentation. The experiments conducted on the semantic text similarity (STS) datasets demonstrate that the proposed CLSESSP surpasses the strong baseline by an average of 2.9 points in the BERT-base model. Extensive experimental results on transfer and clustering tasks demonstrate the effectiveness of our proposed model compared to strong baselines. The code is available at https://github.com/KCshen1125/CLSESSP.

Uniqueness meets Semantics: A Novel Semantically Meaningful Bag-of-Words Approach for Matching Resumes to Job Profiles

In an increasingly competitive world, automated screening of resumes of candidates is the need of the hour given the large numbers of such resumes in career portals on the World Wide Web. Resume classification is a subset of the document classification problem in which the keywords extracted from the resume play a significant role in determining the job profile. In this paper, we explore the novel combination of uniqueness in terms of the number of occurrences of a keyword in a resume class as compared to the other resume classes, and the concept of semantics by representing the filtered keywords using word embeddings that can be used to find semantic similarities between resume documents. The principle of maximum entropy partitioning is used to find the keywords unique to a particular class. The aim is to use semantic representations of only those keywords that occur more frequently in one class more than in any other class; these are then passed as input to a Bidirectional long short-term memory (LSTM) for classification. Our experiments on a benchmark dataset proves that the proposed approach outperforms the state of the art in text classification by a significant margin proving the efficacy of our approach.

A hybrid model to improve IC-related metrics of semantic similarity between words

This paper proposes a hybrid model to improve Information Content (IC) related metrics of semantic similarity between words, named IC+SP, based on the essential hypothesis that IC and the shortest path are two relatively independent semantic evidences and have approximately equal influences to the semantic similarity metric. The paradigm of IC+SP is to linearly combine the IC-related metric and the shortest path. Meanwhile, a transformation from the semantic similarity of the concepts to that of the words is presented by maximizing every component of IC+SP. 13 improved IC-related metrics based on IC+SP are formed and implemented on the experimental platform HESML Lastra-Díaz (Inf Syst 66:97–118, 2017). Pearson’s and Spearman’s correlation coefficients on well-accepted benchmarks for the improved metrics compare to those for the original ones to evaluate IC+SP. I introduce the Wilcoxon Signed-Rank Test needing no standard distribution hypothesis, while, this hypothesis is required by T-Test on the sample of small size. T-Test, as well as the Wilcoxon Signed-Rank Test, conduct on the differences of the correlative coefficients for improved and original metrics. It is expected that the improved IC-related metrics could significantly outperform their corresponding original ones, and the experimental results, including the comparisons of mean and maximum of correlation coefficients as well as the p-value and confidence interval of both tests, accomplish the anticipation in the vast majority of cases.

Trajectory privacy protection method based on sensitive semantic location replacement

As Location-Based Services (LBSs) proliferate within the Social Internet of Vehicles (SIoVs), the collection and utilization of vehicle trajectories, which are rich in spatio-temporal and semantic informations, have intensified. Publishing these trajectories without adequate privacy safeguards significantly threatens user semantic trajectory privacy. Current methodologies for protecting semantic trajectory privacy do not adequately consider the similarity in query probability between sensitive and alternative semantic locations, thus undermining the efficacy of privacy protection. To address this critical gap, we introduces a novel trajectory privacy protection method based on sensitive semantic location replacement (SSLR), which strategically replaces sensitive semantic locations. This method consists of three key steps: first, semantically annotating trajectory sampling locations to identify sensitive semantic points; second, employing a unique double semicircular area to define replacement regions and selecting replacement Points of Interests (PoIs) that align in semantic attributes and query probabilities; third, reconstructing and publishing the modified trajectories. Our simulation results confirm that SSLR advances the state of the art by delivering superior performance in average recognition rate, geographic similarity, and semantic similarity compared to existing methods. Crucially, by incorporating considerations of query probability similarities, this paper not only enhances the effect of trajectory privacy protection but also preserves its utility. This dual enhancement represents a groundbreaking approach to trajectory privacy, with significant implications for advancing privacy strategies in LBSs within SIoVs.

Enhancing Medical Image Retrieval with UMLS-Integrated CNN-Based Text Indexing.

In recent years, Convolutional Neural Network (CNN) models have demonstrated notable advancements in various domains such as image classification and Natural Language Processing (NLP). Despite their success in image classification tasks, their potential impact on medical image retrieval, particularly in text-based medical image retrieval (TBMIR) tasks, has not yet been fully realized. This could be attributed to the complexity of the ranking process, as there is ambiguity in treating TBMIR as an image retrieval task rather than a traditional information retrieval or NLP task. To address this gap, our paper proposes a novel approach to re-ranking medical images using a Deep Matching Model (DMM) and Medical-Dependent Features (MDF). These features incorporate categorical attributes such as medical terminologies and imaging modalities. Specifically, our DMM aims to generate effective representations for query and image metadata using a personalized CNN, facilitating matching between these representations. By using MDF, a semantic similarity matrix based on Unified Medical Language System (UMLS) meta-thesaurus, and a set of personalized filters taking into account some ranking features, our deep matching model can effectively consider the TBMIR task as an image retrieval task, as previously mentioned. To evaluate our approach, we performed experiments on the medical ImageCLEF datasets from 2009 to 2012. The experimental results show that the proposed model significantly enhances image retrieval performance compared to the baseline and state-of-the-art approaches.

Discourse Fluctuation of “Chinese Aerospace” in the Mainstream Media

Drawing upon the principles of Critical Discourse Analysis and Usage Fluctuation Analysis, this study investigated the discourse fluctuation of “Chinese aerospace” in the mainstream media of China, the US, the UK, and Russia, covering the initiating, consistent, transient and terminating collocates within the discourse; and also examines the diachronic change of their semantic similarity. It is further explained with an in-depth investigation of the ideology and social causes behind. Results show that Chinese and Russian media’s attention to Chinese aerospace is on the rise, constructing an image of China as a strong nation leading in innovative aerospace technology; while British and American media’s attention fluctuates considerably, conveying an ambitious image. This study offers a new perspective for the discourse analysis of news reports and contributes to the discursive construction of “Chinese aerospace”.

Detecting coordinated and bot-like behavior in Twitter: the Jürgen Conings case

Social media platforms can play a pivotal role in shaping public opinion during times of crisis and controversy. The COVID-19 pandemic resulted in a large amount of dubious information being shared online. In Belgium, a crisis emerged during the pandemic when a soldier (Jürgen Conings) went missing with stolen weaponry after threatening politicians and virologists. This case created further division and polarization in online discussions. In this paper, we develop a methodology to study the potential of coordinated spread of incorrect information online. We combine network science and content analysis to infer and study the social network of users discussing the case, the news websites shared by those users, and their narratives. Additionally, we examined indications of bots or coordinated behavior among the users. Our findings reveal the presence of distinct communities within the discourse. Major news outlets, conspiracy theory websites, and anti-vax platforms were identified as the primary sources of (dis)information sharing. We also detected potential coordinated behavior and bot activity, indicating possible attempts to manipulate the discourse. We used the rapid semantic similarity network for the analysis of text, but our approach can be extended to the analysis of images, videos, and other types of content. These results provide insights into the role of social media in shaping public opinion during times of crisis and underscore the need for improved strategies to detect and mitigate disinformation campaigns and online discourse manipulation. Our research can aid intelligence community members in identifying and disrupting networks that spread extremist ideologies and false information, thereby promoting a more informed and resilient society.

Artificial intelligence in forensic medicine and related sciences – selected issues = Sztuczna inteligencja w medycynie sądowej i naukach pokrewnych – wybrane zagadnienia

Aim. The aim of the work is to provide an overview of the potential application of artificial intelligence in forensic medicine and related sciences, and to identify concerns related to providing medico-legal opinions and legal liability in cases in which possible harm in terms of diagnosis and/or treatment is likely to occur when using an advanced system of computer-based information processing and analysis. Materials and methods. The material for the study comprised scientific literature related to the issue of artificial intelligence in forensic medicine and related sciences. For this purpose, Google Scholar, PubMed and ScienceDirect databases were searched. To identify useful articles, such terms as „artificial intelligence,” „deep learning,” „machine learning,” „forensic medicine,” „legal medicine,” „forensic pathology” and „medicine” were used. In some cases, articles were identified based on the semantic proximity of the introduced terms. Conclusions. Dynamic development of the computing power and the ability of artificial intelligence to analyze vast data volumes made it possible to transfer artificial intelligence methods to forensic medicine and related sciences. Artificial intelligence has numerous applications in forensic medicine and related sciences and can be helpful in thanatology, forensic traumatology, post-mortem identification examinations, as well as post-mortem microscopic and toxicological diagnostics. Analyzing the legal and medico-legal aspects, artificial intelligence in medicine should be treated as an auxiliary tool, whereas the final diagnostic and therapeutic decisions and the extent to which they are implemented should be the responsibility of humans.

Artificial intelligence in forensic medicine and related sciences - selected issues.

The aim of the work is to provide an overview of the potential application of artificial intelligence in forensic medicine and related sciences, and to identify concerns related to providing medico-legal opinions and legal liability in cases in which possible harm in terms of diagnosis and/or treatment is likely to occur when using an advanced system of computer-based information processing and analysis. The material for the study comprised scientific literature related to the issue of artificial intelligence in forensic medicine and related sciences. For this purpose, Google Scholar, PubMed and ScienceDirect databases were searched. To identify useful articles, such terms as "artificial intelligence," "deep learning," "machine learning," "forensic medicine," "legal medicine," "forensic pathology" and "medicine" were used. In some cases, articles were identified based on the semantic proximity of the introduced terms. Dynamic development of the computing power and the ability of artificial intelligence to analyze vast data volumes made it possible to transfer artificial intelligence methods to forensic medicine and related sciences. Artificial intelligence has numerous applications in forensic medicine and related sciences and can be helpful in thanatology, forensic traumatology, post-mortem identification examinations, as well as post-mortem microscopic and toxicological diagnostics. Analyzing the legal and medico-legal aspects, artificial intelligence in medicine should be treated as an auxiliary tool, whereas the final diagnostic and therapeutic decisions and the extent to which they are implemented should be the responsibility of humans.

Lexical pathway from L2 to L1 activation in intermediate proficient bilinguals: behavioral and ERP evidence.

Numerous studies have demonstrated that second language (L2) comprehension is often accompanied by activations in the first language (L1). Using both behavioral measurement and event-related potential (ERP), this study conducted two experiments to investigate whether a direct activation pathway exists from L2 lexical representation to L1 lexical representation (the lexical pathway) in intermediate proficient bilinguals. In Experiment 1, we designed a vowel letter search task on English word pairs, which enables bilinguals to prevent semantic priming in the first 300 ms processing stage after the words' onset. In Experiment 2, Mandarin-English bilinguals were recruited to complete this task on English word pairs with occasional first character repetition between the Chinese counterparts of a word pair. Results showed a significant main effect within both the P200 and N400 time windows, indicating the activation of bilinguals' L1 lexical representation during these intervals. However, the main effect of semantic relatedness was only significant in the N400 time window. These results suggest that bilinguals can activate their L1 lexical representation directly before engaging in conceptual representation. This finding supported a lexical pathway of activation from L2 lexical representation to L1 lexical representation during visual-word recognition in intermediate proficient bilinguals.

Large language models for precision oncology: Clinical decision support through expert-guided learning.

e13609 Background: Precision oncology revolutionized cancer treatment by identifying molecular biomarkers to guide personalized care. The ever-growing body of medical literature presents a challenge for oncologists researching targeted therapies. While recent studies investigated large language models (LLMs) to streamline this process, LLM reliance on general rather than medical knowledge limits clinical relevance and trustworthiness. To address these limitations, we developed a retrieval augmented generation (RAG) system that integrates PubMed clinical studies, trial databases and oncological guidelines with LLMs to support targeted treatment recommendations. The Molecular Tumor Board (MTB) at the Center of Personalized Medicine (ZPMTUM) guided and evaluated treatment options proposed by the LLM to assess their applicability for clinical decision support. Methods: We used 10 publicly accessible fictional patient cases with 7 tumor types and 59 distinct molecular alterations. Our LLM system MEREDITH (Medical Evidence Retrieval and Data Integration for Tailored Healthcare) consists of Google's Gemini Pro, enhanced with RAG and Chain-of-Thought (CoT) prompting. To establish a benchmark, clinical experts at ZPMTUM manually annotated the cases. Informed by MTB expert feedback, we iteratively improved our LLM system from a draft system relying on PubMed-indexed data to an enhanced system, which replicated expert annotation processes by incorporating oncology guidelines, drug availability and trial databases (ClinicalTrials.gov, QuickQueck.de). ZPMTUM assessed credibility and clinical relevance of manually annotated and LLM-generated recommendations. Patient-level data on (likely) pathogenic molecular alterations and recommended treatment options were summarized using median and interquartile range (IQR). Semantic similarity between LLM and clinician responses was assessed using cosine similarity of text vector embeddings; paired t-test evaluated significance. Results: The median of (likely) pathogenic molecular alterations per patient was 2.5 (IQR: 2-3). ZPMTUM identified a median of 2 treatment options per patient (IQR: 1-3), while the enhanced LLM identified a median of 4 (IQR: 3-5). MEREDITH proposed multiple relevant treatment suggestions, including therapies based on preclinical studies, and molecular interactions, for further assessment by the MTB. ZPMTUM prioritized the most suitable clinical option. The mean semantic textual similarity of LLM responses increased significantly from 0.69 in the draft system to 0.76 in the enhanced system (p <0.001). Thus, feedback from ZPMTUM enhanced the model's ability to align its responses with clinician thought processes. Conclusions: Leveraging expert thought processes to instruct LLMs holds promise as a novel decision support tool for precision oncology.

Adaptive resonance demodulation semantic-induced zero-shot compound fault diagnosis for railway bearings

For the challenges of diverse compound faults and low identification accuracy of railway bearings, a new zero-shot diagnosis model based on adaptive resonance demodulation semantic is proposed for the compound fault diagnosis of railway bearings. The model adopts adaptive resonance demodulation to identify the optimal resonance frequency band rich in fault information in bearing signals, and constructs the single and compound fault semantics of samples without separating the compound fault signals, thus improving the interpretability of semantic features. Moreover, spatial Euclidean distance is used to measure the similarity of features and semantics in the mapping space, which enables the identification of unknown compound faults by single faults. Verification through railway bearing data shows that this model effectively improves the compound fault identification accuracy under zero samples and is better than the comparison models. The research results can provide theoretical reference for the research and application of railway bearing fault diagnosis.

Co-augmentation of structure and feature for boosting graph contrastive learning

Graph Contrastive Learning (GCL) learns invariant representation by maximizing the consistency between different augmented graphs that share the same semantics. However, the performance of existing GCL methods is inseparable from varied manually designed augmentation techniques that randomly perturb edges/nodes/features, which unexpectedly change the semantic similarity and lead to biased structure and feature augmentation. In this paper, we propose a co-augmentation strategy for STRucture and FEature (STRFE) to eliminate augmentation bias. Specifically, we construct a relatively unbiased augmented graph by amplifying or suppressing graph frequency in the spectral domain, which promises structural consistency and feature diversity between the augmented and original graph. Moreover, we investigate external and internal contrastive loss to balance the consistency and diversity between the original and augmented graph, which facilitates preserving semantic similarity and encourages relatively unbiased structure and feature augmentation to enhance the performance of GCL. Theoretical analysis proves why our proposed structure and feature co-augmentation strategy can perform well. Extensive experiments show that STRFE achieves competitive results in three real-world datasets on different downstream tasks compared with more than ten benchmarks.

Experimental study on short-text clustering using transformer-based semantic similarity measure

Sentence clustering plays a central role in various text-processing activities and has received extensive attention for measuring semantic similarity between compared sentences. However, relatively little focus has been placed on evaluating clustering performance using available similarity measures that adopt low-dimensional continuous representations. Such representations are crucial in domains like sentence clustering, where traditional word co-occurrence representations often achieve poor results when clustering semantically similar sentences that share no common words. This article presents a new implementation that incorporates a sentence similarity measure based on the notion of embedding representation for evaluating the performance of three types of text clustering methods: partitional clustering, hierarchical clustering, and fuzzy clustering, on standard textual datasets. This measure derives its semantic information from pre-training models designed to simulate human knowledge about words in natural language. The article also compares the performance of the used similarity measure by training it on two state-of-the-art pre-training models to investigate which yields better results. We argue that the superior performance of the selected clustering methods stems from their more effective use of the semantic information offered by this embedding-based similarity measure. Furthermore, we use hierarchical clustering, the best-performing method, for a text summarization task and report the results. The implementation in this article demonstrates that incorporating the sentence embedding measure leads to significantly improved performance in both text clustering and text summarization tasks.

Improving few-shot named entity recognition via Semantics induced Optimal Transport

Named entity recognition (NER) is to identify and categorize entities in unstructured text, which serves as a fundamental task for a variety of natural language processing (NLP) applications. In particular, emerging few-shot NER methods aim to learn model parameters well with few samples and have received considerable attention. The dominant few-shot NER methods usually employ pre-trained language models (PLMs) as their basic architecture and fine-tune model parameters with few NER samples. Since the sample size is small and there are a large number of parameters in PLMs, fine-tuning may result in the parameters of PLMs being highly biased. To address this issue, this study introduces the semantic distribution distance constraints to optimize the fine-tuning process of few-shot NER models and develops a framework named Semantic Constraints on few-shot Named Entity Recognition (SCNER). Specifically, the framework formulates the general knowledge transfer of PLMs as an optimal transport procedure with a semantic prior. And, a Semantics-induced Optimal Transport (SOT) regularizer is developed to utilize the importance and similarities of tokens within sentences. SOT builds the semantic distribution of the sentence and defines the transport costs between tokens to achieve the token-level optimal transport procedures. Finally, SOT is employed as a regularization term of few-shot NER to introduce the semantic distribution distance constraint for effectively transferring general knowledge from PLMs. The experiments on four public datasets demonstrate that the proposed method significantly improves the performance of NER models in both few-shot and fully supervised scenarios. SCNER is a common framework that can be applied to a variety of models without adding additional learning parameters, and can be used to enhance the generalization ability and adaptability of various few-shot NER models.

Weight Saliency search with Semantic Constraint for Neural Machine Translation attacks

Text adversarial attack is an effective way to improve the robustness of Neural Machine Translation (NMT) models. Existing NMT attack tasks are often completed by replacing words. However, most of previous works pursue a high attack success rate but produce semantic inconsistency sentences, leading to wrong translations even for humans. In this paper, we propose a Weight Saliency search with Semantic Constraint (WSSC) algorithm to make semantic consistency word modifications to the input sentence for black-box NMT attacks. Specifically, our WSSC has two major merits. First, it optimizes the word substitution with a word saliency method, which is helpful to reduce word replacement rate. Second, it constrains the objective function with a semantic similarity loss, ensuring every modification does not lead to significant semantic changes. We evaluate the effectiveness of the proposed WSSC by attacking three popular NMT models, i.e., T5, Marian, and BART, on three widely used datasets, i.e., WMT14, WMT16, and TED. Experimental results validate that our WSSC improves Attack Success Rate (ASR) by 4.02% and Semantic Similarity score (USE) by 1.28% on average. Besides, our WSSC also shows good properties in keeping grammar correctness and transfer attack.

A framework for trust-related knowledge transfer in human–robot interaction

Trustworthy human–robot interaction (HRI) during activities of daily living (ADL) presents an interesting and challenging domain for assistive robots, particularly since methods for estimating the trust level of a human participant towards the assistive robot are still in their infancy. Trust is a multifaced concept which is affected by the interactions between the robot and the human, and depends, among other factors, on the history of the robot’s functionality, the task and the environmental state. In this paper, we are concerned with the challenge of trust transfer, i.e. whether experiences from interactions on a previous collaborative task can be taken into consideration in the trust level inference for a new collaborative task. This has the potential of avoiding re-computing trust levels from scratch for every new situation. The key challenge here is to automatically evaluate the similarity between the original and the novel situation, then adapt the robot’s behaviour to the novel situation using previous experience with various objects and tasks. To achieve this, we measure the semantic similarity between concepts in knowledge graphs (KGs) and adapt the robot’s actions towards a specific user based on personalised interaction histories. These actions are grounded and then verified before execution using a geometric motion planner to generate feasible trajectories in novel situations. This framework has been experimentally tested in human–robot handover tasks in different kitchen scene contexts. We conclude that trust-related knowledge positively influences and improves collaboration in both performance and time aspects.

MMGPL: Multimodal Medical Data Analysis with Graph Prompt Learning

Prompt learning has demonstrated impressive efficacy in the fine-tuning of multimodal large models to a wide range of downstream tasks. Nonetheless, applying existing prompt learning methods for the diagnosis of neurological disorder still suffers from two issues: (i) existing methods typically treat all patches equally, despite the fact that only a small number of patches in neuroimaging are relevant to the disease, and (ii) they ignore the structural information inherent in the brain connection network which is crucial for understanding and diagnosing neurological disorders. To tackle these issues, we introduce a novel prompt learning model by learning graph prompts during the fine-tuning process of multimodal models for diagnosing neurological disorders. Specifically, we first leverage GPT-4 to obtain relevant disease concepts and compute semantic similarity between these concepts and all patches. Secondly, we reduce the weight of irrelevant patches according to the semantic similarity between each patch and disease-related concepts. Moreover, we construct a graph among tokens based on these concepts and employ a graph convolutional network layer to extract the structural information of the graph, which is used to prompt the pre-trained multimodal models for diagnosing neurological disorders. Extensive experiments demonstrate that our method achieves superior performance for neurological disorder diagnosis compared with state-of-the-art methods and validated by clinicians.

The formation of inaccurate cognitive destination image from a memory process perspective

ABSTRACT People often harbour misconceptions when forming their perceptions of destinations. Via a top-of-mind association task, this paper investigates the cognitive mechanisms behind the formation of inaccurate destination images of several Chinese metropolises. The results reveal that the primary factors leading to inaccurate cognitive destination images are proximity and stereotyping. Proximity includes geographic, cultural, sensory, semantic, and phonetic similarities among mental representations; whereas, stereotyping that tends to resist change can lead to systematic biases. This paper enriches our understanding of destination image and offers novel perspectives for destination branding and management.

Lightweight railroad semantic segmentation network and distance estimation for railroad Unmanned aerial vehicle images

Derailments significantly harm railroads in terms of severity and fatality rates. Manually monitoring railway tracks is a tedious and often insufficient task to prevent derailment accidents. This necessitates the development of an automated monitoring system to oversee the condition of railway tracks. In the foreseeable future, Unmanned Aerial Vehicles (UAVs), artificial intelligence, and computer vision will play a crucial role in periodically assessing the railway environment to ensure passenger safety. This paper proposes a lightweight and efficient railroad semantic segmentation network (Lite-RSNet), which segments real-time aerial railroad images into rails, gauges, and backgrounds. The segmented regions then compute the distance between rails and obstacles, identifying potential derailment hazards in railroad applications. Lite-RSNet incorporates a multi-level feature fusion framework that includes a residual layer, a multi-head attention layer, a Concurrent Spatial and Channel Squeeze Excitation (CSSE) block, and a Convolutional Block Attention Module (CBAM). The designed approach enhances onboard processing efficiency by streamlining the model to use fewer parameters and reduce storage demands. Performance assessment of the model utilized the Rail Segmentation Dataset (RSD), which consists of high-resolution, low-altitude aerial photographs of Indian railroads. The Lite-RSNet model outperforms current leading methods, maintaining a lean structure with only 6 million parameters and achieving notable metrics of 0.971 for the Dice Coefficient (DC) and 0.947 for the Jaccard Index (JI) on the test dataset. Furthermore, distance estimation between an obstacle and rail has been implemented using image processing techniques as an additional work of this proposal to know about the caution of an obstacle.