Hasil untuk "Semantics"

R. Socher, A. Karpathy, Quoc V. Le et al.

Previous work on Recursive Neural Networks (RNNs) shows that these models can produce compositional feature vectors for accurately representing and classifying sentences or images. However, the sentence vectors of previous models cannot accurately represent visually grounded meaning. We introduce the DT-RNN model which uses dependency trees to embed sentences into a vector space in order to retrieve images that are described by those sentences. Unlike previous RNN-based models which use constituency trees, DT-RNNs naturally focus on the action and agents in a sentence. They are better able to abstract from the details of word order and syntactic expression. DT-RNNs outperform other recursive and recurrent neural networks, kernelized CCA and a bag-of-words baseline on the tasks of finding an image that fits a sentence description and vice versa. They also give more similar representations to sentences that describe the same image.

Elia Bruni, N. Tran, Marco Baroni

Distributional semantic models derive computational representations of word meaning from the patterns of co-occurrence of words in text. Such models have been a success story of computational linguistics, being able to provide reliable estimates of semantic relatedness for the many semantic tasks requiring them. However, distributional models extract meaning information exclusively from text, which is an extremely impoverished basis compared to the rich perceptual sources that ground human semantic knowledge. We address the lack of perceptual grounding of distributional models by exploiting computer vision techniques that automatically identify discrete "visual words" in images, so that the distributional representation of a word can be extended to also encompass its co-occurrence with the visual words of images it is associated with. We propose a flexible architecture to integrate text- and image-based distributional information, and we show in a set of empirical tests that our integrated model is superior to the purely text-based approach, and it provides somewhat complementary semantic information with respect to the latter.

B. Levin

Miriam R. L. Petruck

P. Mika

Ann A. Copestake, D. Flickinger, C. Pollard et al.

H. Kamp, Uwe Reyle

R. Jackendoff

M. H. Emden, R. Kowalski

Hassan Saif, Yulan He, Miriam Fernández et al.

We propose a semantic sentiment representation of words called SentiCircle.SentiCircle captures the contextual semantic of words from their co-occurrences.SentiCircle updates the sentiment of words based on their contextual semantics.SentiCircle can be used to perform entity- and tweet-level level sentiment analysis. Sentiment analysis on Twitter has attracted much attention recently due to its wide applications in both, commercial and public sectors. In this paper we present SentiCircles, a lexicon-based approach for sentiment analysis on Twitter. Different from typical lexicon-based approaches, which offer a fixed and static prior sentiment polarities of words regardless of their context, SentiCircles takes into account the co-occurrence patterns of words in different contexts in tweets to capture their semantics and update their pre-assigned strength and polarity in sentiment lexicons accordingly. Our approach allows for the detection of sentiment at both entity-level and tweet-level. We evaluate our proposed approach on three Twitter datasets using three different sentiment lexicons to derive word prior sentiments. Results show that our approach significantly outperforms the baselines in accuracy and F-measure for entity-level subjectivity (neutral vs. polar) and polarity (positive vs. negative) detections. For tweet-level sentiment detection, our approach performs better than the state-of-the-art SentiStrength by 4-5% in accuracy in two datasets, but falls marginally behind by 1% in F-measure in the third dataset.

Christin Wirth

Zhaochong An, Zirui Li, Mingqiao Ye et al.

Video understanding aims to enable models to perceive, reason about, and interact with the dynamic visual world. In contrast to image understanding, video understanding inherently requires modeling temporal dynamics and evolving visual context, placing stronger demands on spatiotemporal reasoning and making it a foundational problem in computer vision. In this survey, we present a structured overview of video understanding by organizing the literature into three complementary perspectives: low-level video geometry understanding, high-level semantic understanding, and unified video understanding models. We further highlight a broader shift from isolated, task-specific pipelines toward unified modeling paradigms that can be adapted to diverse downstream objectives, enabling a more systematic view of recent progress. By consolidating these perspectives, this survey provides a coherent map of the evolving video understanding landscape, summarizes key modeling trends and design principles, and outlines open challenges toward building robust, scalable, and unified video foundation models.

Vibha, Rajesh R. Pai, Sumith N.

This study proposes a semantic pipeline designed to generate domain-oriented and contextually relevant hypotheses by analyzing existing literature on mHealth applications in India. Using a corpus of mHealth texts, the framework extracts hidden semantics through TF-IDF, topic modeling, and contextual mapping with domain ontologies. It then employs prompt-based interactions with large language models (LLMs) to systematically generate and validate hypotheses aligned with identified topic-concept relationships. The results demonstrate the framework’s effectiveness in producing high-quality, structured hypotheses, as validated by expert ratings ranging from 4.2 to 4.6. Most hypotheses were found to be plausible or highly plausible, with low semantic redundancy indicating diversity across topics, except in stakeholder-related areas which showed moderate overlap. Although the inclusion of semantic augmentation increased processing time, it significantly enhanced interpretability and validity. The high lexical density observed (up to 0.90) further reflects the linguistic flexibility of the generated hypotheses. This approach underscores the potential of computational methods in automating hypothesis generation and enabling data-driven discoveries in the mHealth domain.

M. Bermúdez-Edo, Tarek Elsaleh, P. Barnaghi et al.

Iva Pukleš, Csilla Páger, Nikola Sakač et al.

Amino acids (AAs) have broad nutritional, therapeutic, and medical significance and thus are one of the most common active ingredients of nutritional supplements. Analytical strategies for determining AAs are high-priced and often limited to methods that require modification of AA polarity or incorporation of an aromatic moiety. The aim of this work was to develop a new method for the determination of L-arginine, L-ornithine, and L-lysine on low-cost microchip electrophoresis instrumentation conjugated with capacitively coupled contactless conductivity detection. A solution consisting of 0.3 M acetic acid and 1 × 10⁻⁵ M iminodiacetic acid has been identified as the optimal background electrolyte, ensuring the shortest possible analysis time. The short migration times of amino acids (t ≤ 64 s) and method simplicity resulted in high analysis throughput with high precision and linearity (R²<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mo>≥</mo></mrow></semantics></math></inline-formula> 0.9971). The limit of detection values ranged from 0.15 to 0.19 × 10⁻⁶ M. The accuracy of the proposed method was confirmed by recovery measurements. The results were compared with CE-UV-VIS and HPLC-DAD methods and showed good agreement. This work represents the first successful demonstration of the ME-C⁴D analysis of L-arginine, L-ornithine, and L-lysine in real samples.

Lei Wang, Xiaomin Liu, Chenye Tian et al.

Here, the single-target parameterization of alternatives to leading-edge noise is carried out using analytical models based on the Wiener–Hopf technique. Four leading-edge serration profiles with different frequencies, amplitudes, and phases are implemented to aid the understanding of sound suppression mechanisms. The effects of the serrated shape factor, wavelength, and amplitude are analyzed at tip-to-root ratios of 0.5, 1, and 2, respectively. An effective double-wavelength sinusoidal serration design can substantially reduce the noise emissions of 5.2 dB at <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mover accent="true"><mrow><mi>h</mi></mrow><mo>¯</mo></mover></mrow></semantics></math></inline-formula> = 2. Additionally, compared to single-wavelength serrations, an additional 1.47 dB noise reduction effect can be obtained by double-wavelength serrations under the appropriate design parameters. The surface pressure and phase distribution of different spanwise-varying leading edges indicate that the phase interference effect affected by source-radiated noise reduction is enhanced by this serration at the hills for serrations with a small curvature, and noise emission in the low-frequency band is more effectively suppressed. The sharper the serration is, the more conducive it is to a reduction in high-frequency noise. Nevertheless, the effectiveness of serrations is usually partially limited by the non-negligible trailing-edge self-noise. The sound source intensity of the root is decreased by the ogee-shaped serrations with a large curvature transition. A secondary noise reduction mechanism with a local source cut-off effect caused by nonlinearity is demonstrated.

Yasena Chantova

This article presents a discussion on polysemy in a crossword by Perec. We apply Culioli’s modelling approach in terms of operation of location and construction of a notion and its notional domain. Polysemy is seen as part of the intangible cultural heritage within languages.

E. Pacuit

Zeyd Boukhers, Christoph Lange, Oya Beyan

Our vision paper outlines a plan to improve the future of semantic interoperability in data spaces through the application of machine learning. The use of data spaces, where data is exchanged among members in a self-regulated environment, is becoming increasingly popular. However, the current manual practices of managing metadata and vocabularies in these spaces are time-consuming, prone to errors, and may not meet the needs of all stakeholders. By leveraging the power of machine learning, we believe that semantic interoperability in data spaces can be significantly improved. This involves automatically generating and updating metadata, which results in a more flexible vocabulary that can accommodate the diverse terminologies used by different sub-communities. Our vision for the future of data spaces addresses the limitations of conventional data exchange and makes data more accessible and valuable for all members of the community.

Tomoya Kurosawa, Hitomi Yanaka

Even in the era of massive language models, it has been suggested that character-level representations improve the performance of neural models. The state-of-the-art neural semantic parser for Discourse Representation Structures uses character-level representations, improving performance in the four languages (i.e., English, German, Dutch, and Italian) in the Parallel Meaning Bank dataset. However, how and why character-level information improves the parser's performance remains unclear. This study provides an in-depth analysis of performance changes by order of character sequences. In the experiments, we compare F1-scores by shuffling the order and randomizing character sequences after testing the performance of character-level information. Our results indicate that incorporating character-level information does not improve the performance in English and German. In addition, we find that the parser is not sensitive to correct character order in Dutch. Nevertheless, performance improvements are observed when using character-level information.