Sarcasm is a subtle form of language in which people express the opposite of what is implied. Previous works of sarcasm detection focused on texts. However, more and more social media platforms like Twitter allow users to create multi-modal messages, including texts, images, and videos. It is insufficient to detect sarcasm from multi-model messages based only on texts. In this paper, we focus on multimodal sarcasm detection for tweets consisting of texts and images in Twitter. We treat text features, image features and image attributes as three modalities and propose a multi-modal hierarchical fusion model to address this task. Our model first extracts image features and attribute features, and then leverages attribute features and bidirectional LSTM network to extract text features. Features of three modalities are then reconstructed and fused into one feature vector for prediction. We create a multi-modal sarcasm detection dataset based on Twitter. Evaluation results on the dataset demonstrate the efficacy of our proposed model and the usefulness of the three modalities.

High‐resolution digital terrain models (DTMs) generated from the Lunar Reconnaissance Orbiter Camera (LROC) Narrow Angle Cameras (NACs) provide an opportunity to study surface roughness of the Moon at meter scale. In this study, we mapped and investigated meter‐scale topographic roughness over 462 regions of the Moon using NAC DTMs. Our results show that, at meter to hectometer scales, there are obvious differences in median bidirectional slope, root‐mean‐square (RMS) height, and median absolute slope between maria and highlands. In terms of the median value, ratios of the bidirectional slope, RMS height, and median absolute slope within the maria and highlands are 1: 2.4, 1: 3.0, and 1: 2.7, respectively. However, up to a baseline of ∼0.2 km, no discernible differences in the Hurst exponent and median differential slope exist between the maria and highlands. The Hurst exponent varies from 0.7 to 0.95, with a median value of 0.9 within both the maria and highlands. To identify potential factors affecting meter‐scale roughness of the Moon, we compared the maria with volcanic features on Earth and Mars, as well as the lunar highlands with simulated cratered terrains. We found that the Hurst exponents within the lunar maria are much larger than those of the volcanic features on Earth and Mars, mainly because the lunar maria accumulated more impact craters with diameters smaller than 1 km. The Hurst exponents within the highlands are consistent with those of the simulated cratered terrains, whose surface roughness depends primarily on crater shape, number density, and stratigraphic age. All these results indicate that lunar surface roughness at meter to hectometer scales is mainly controlled by small degraded impact craters.

Partially buried craters on the Moon are those craters whose distal ejecta are covered by lava flows and where the crater rim crest still protrudes above the mare plain. Based on the difference in rim heights between a partially buried crater and an unburied crater, previous studies estimated the thicknesses of the lunar mare basalts. However, these studies did not consider the erosion of the crater rim height, which can result in an overestimate in the derived thickness. By using recent high‐resolution topographic data, we report a basalt thickness estimation method based on numerically modeling the topographic degradation of partially buried craters. We identified 661 buried craters over the lunar surface, and their spatial distribution suggests a preferential occurrence along the mare‐highland boundaries. An elevation model of fresh lunar craters was derived, and the topographic diffusion equation was used to model crater degradation. By modeling the formation, degradation, and flooding of partially buried craters, basalt thicknesses were estimated for 41 mare craters whose rims are completely exposed. The resulting mare basalt thicknesses vary from 33 to 455 m, with a median value of 105 m that is 95 m smaller than that derived when not considering crater degradation. The estimated eruption rate of lunar mare basalts is found to have peaked at 3.4 Ga and then decreased with time, indicating a progressive cooling of the lunar interior. As a by‐product from the crater degradation model, our results suggest that the topographic diffusivity of lunar craters increases with diameter.

[1] In an attempt to reduce the ambiguity on radar detection of water ice at the permanently shadowed regions near the lunar poles, radar echo strength and circular polarization ratio (CPR) of impact craters are analyzed using the Miniature Radio Frequency (Mini-RF) radar data from the Lunar Reconnaissance Orbiter mission. Eight typical craters, among over 70 craters, are selected and classified into four categories based on their locations and CPR characteristics: polar anomalous, polar fresh, nonpolar anomalous, and nonpolar fresh. The influences on CPR caused by surface slope, rocks, and dielectric constant are analyzed quantitatively using high-resolution topography data and optical images. A two-component mixed model for CPR that consists of a normal surface and a rocky surface is developed to study the effect of rocks that are perched on lunar surface and buried in regolith. Our analyses show that inner wall of a typical bowlshaped crater can give rise to a change of about 30 ı in local incidence angle of radar wave, which can further result in a CPR difference of about 0.2. There is a strong correlation between Mini-RF CPR and rock abundance that is obtained from high-resolution optical images, and predictions from the two-component mixed model match well with the observed CPRs and the estimated rock abundances. Statistical results show that there is almost no apparent difference in CPR characteristics between the polar and nonpolar anomalous craters, or between the polar and nonpolar fresh craters. The enhanced CPR in the interior of anomalous craters is most probably caused by rocks that are perched on lunar surface or buried in regolith, instead of ice deposits as suggested in previous studies.Citation : Fa, W., and Y. Cai (2013), Circular polarization ratio characteristics of impact craters from Mini-RF observations and implications for ice detection at the polar regions of the Moon,