In cross-lingual language models, representations for many different languages live in the same space. Here, we investigate the linguistic and non-linguistic factors affecting sentence-level alignment in cross-lingual pretrained language models for 1
01 languages and 5,050 language pairs. Using BERT-based LaBSE and BiLSTM-based LASER as our models, and the Bible as our corpus, we compute a task-based measure of cross-lingual alignment in the form of bitext retrieval performance, as well as four intrinsic measures of vector space alignment and isomorphism. We then examine a range of linguistic, quasi-linguistic, and training-related features as potential predictors of these alignment metrics. The results of our analyses show that word order agreement and agreement in morphological complexity are two of the strongest linguistic predictors of cross-linguality. We also note in-family training data as a stronger predictor than language-specific training data across the board. We verify some of our linguistic findings by looking at the effect of morphological segmentation on English-Inuktitut alignment, in addition to examining the effect of word order agreement on isomorphism for 66 zero-shot language pairs from a different corpus. We make the data and code for our experiments publicly available.
During the past few years, new developments have occurred in the field of 3D photogrammetric modeling of historical monuments. One of these developments is the expansion of 3D photogrammetric modeling cost-effective licensed software, such as Agisoft
Metashape, into the practical and affordable world. This type of SfM (Structure from Motion) software offers the world of 3D modelling of cultural heritage a powerful tool for documentation and visualization. This paper presents a methodology for evaluating the effectiveness of a range of commercial SfM photogrammetry systems when applied in the 3D modeling of historical monuments. The monument to which these systems were applied is the principal façade of Safita Tower, a medieval structure in Safita, north-western Syria.. The applied photogrammetric systems consist of the Nikon Coolpix P100 10 MP digital camera, the commercial software Agisoft Metashape, 3DF Zephyr Aerial and Pix4D Mapper. The resulted 3D point clouds were compared with an available dense point cloud acquired by a laser scanner. This comparison proved that the low-cost SfM photogrammetry is an accurate methodology to 3D modeling historical monuments.
Current sequence-to-sequence models are trained to minimize cross-entropy and use softmax to compute the locally normalized probabilities over target sequences. While this setup has led to strong results in a variety of tasks, one unsatisfying aspect
is its length bias: models give high scores to short, inadequate hypotheses and often make the empty string the argmax---the so-called cat got your tongue problem. Recently proposed entmax-based sparse sequence-to-sequence models present a possible solution, since they can shrink the search space by assigning zero probability to bad hypotheses, but their ability to handle word-level tasks with transformers has never been tested. In this work, we show that entmax-based models effectively solve the cat got your tongue problem, removing a major source of model error for neural machine translation. In addition, we generalize label smoothing, a critical regularization technique, to the broader family of Fenchel-Young losses, which includes both cross-entropy and the entmax losses. Our resulting label-smoothed entmax loss models set a new state of the art on multilingual grapheme-to-phoneme conversion and deliver improvements and better calibration properties on cross-lingual morphological inflection and machine translation for 7 language pairs.
Lemmatization is often used with morphologically rich languages to address issues caused by morphological complexity, performed by grammar-based lemmatizers. We propose an alternative for this, in form of a tool that performs lemmatization in the spa
ce of word embeddings. Word embeddings as distributed representations natively encode some information about the relationship between base and inflected forms, and we show that it is possible to learn a transformation that approximately maps the embeddings of inflected forms to the embeddings of the corresponding lemmas. This facilitates an alternative processing pipeline that replaces traditional lemmatization with the lemmatizing transformation in downstream processing for any application. We demonstrate the method in the Finnish language, outperforming traditional lemmatizers in example task of document similarity comparison, but the approach is language independent and can be trained for new languages with mild requirements.
The objective of the research is to complete a theoretical and practical study related to
coastal marine works in order to calculate the amounts of silt removal from harbor basins
and entrances, and to present the methods and devices used in the pe
rformance of
topographic survey and numerical methods in the calculation and comparison of
quantities.
In the theoretical part, the factors that lead to the formation of silt deposits in the port
basins, the methods of their removal and the deepening of the navigational pathways to
enter and exit the harbors were addressed.
In the practical part, the results, methods of measurements and topographic results were
presented during the stages of investment of the port, at least two stages, at the beginning
of the investment and before the process of direct withdrawal, and then calculating the
quantities of the implemented and comparing them, to obtain maritime plans and final
quantities.
The research concluded with specific proposals on the methods of calculating the
quantities of the isolated port, the method of constructing the measured geodetic networks,
the achievement of the topographic elevation under the water surface, and the
identification of the software parts related to the various marine works and ways of
benefiting from them.
Optimum Index Factor (OIF) technique of statical analysis is fusion with Decision Tree Classification (DTC) method in determine the spectral critical value for separation the features in the image processing programs, and the architecture of this app
roach is designed for accuracy extracting the area and distribution urban from space image. Accuracy assessment of that approach is tested by comparing the supervised classification results for these feature from both the original bands of image and synthetical bands/indices of the image upon OIF value. Applied results of the approach on certain district represent the north Swaidaa city by Quick Bird image are: 98% for the suggested approach opposite 93% for supervised classification method of the synthetical bands image and 82% for original bands image. Accuracy of the approach is derived from exact separation the urban feature than similar spectrally objects in the image as basalt exposures and roads, where achievement of the other processing methods are less.
Marine sources are of the important sources of coastal States, where a significant
portion of its economy is tied to people who earn their living from these sources [4]. After
the discovery of mineral and hydrocarbons ores at the seabed, and also b
ecause of the
disputes between States on fishing regions, it became necessary for some States to
reconsider their idea regarding maritime boundaries and open the disputed files with
neighboring States even it is over a small area, to demand re-delimitation of their maritime
boundaries, even if it required the intervention of the International Court of Justice. In this
research a way is purposed to draw he coast line by using Google Earth detailed photos for
the derivation of Lattakia coast baseline and studying its various special issues. The search
has found good results in identifying those lines allowing full generalization on the Syrian
coast, and also allowing for subsequent research linked to those lines.
Choosing of sprinkler trajectory angle and estimation of its
changing is an important factor in sprinkling irrigation systems.
Farmers choose sprinklers with small trajectory angle when they
used treatment wastewater, or water is spraying below tree canopy.
while sprinklers with high trajectory angle used when spraying
above canopy.
The research aims to complete the study of the theory and operation of networks
geodesic associated marine beach, whether horizontal or vertical, with display rods and
devices used in completion of topographic survey of networks in both land and ma
rine
parts.
In theoretical part it was addressed to specific shapes networks geodesic which
distinguish maritime works beach that takes the form of the morphological characteristics
of the natural land of each building proposed network, and the numerical characteristics of
each building proposed networks.
In the practical part the results of measurements and amendments was presented to
some forms of executed geodesic beach networks for several forms of works: (topographic
raise achievements within the existing and investor basin in order to establish a slider,
completion of the cadastral beach works in order to determine distortions in the existing
pier, clean the shipyard of silt and sediment.
Software has been reviewed to assist in completing marine works, utmost to obtain
the final marine charts.
The research concluded that specific proposals on vertical geodetic networks studies,
and identify the various parts of software related to marine works and ways to take
advantage of them.
The production of 3D models of urban areas, using aerial photographs, is of great benefit to companies and small engineering offices. But the major problem is the high cost of Digital Photogrammetry Workstations (DPWS) that are currently used for the
production of this kind of models. In addition, the use of these workstations requires long experience and good knowledge in photogrammetry. In this paper, we propose an alternative solution for 3D modeling of urban areas from a stereoscopic pair of aerial photos, a low cost close range photogrammetry software and the applications of 3D modeling available in some Geographic Information System (GIS) platforms. The close range photogrammetry software is a low coast system, compared to DPWS, and it doesn’t require any spatial background in photogrammetry. This software is used to extract the heights of elements that exist in the study area. GIS is used to produce the 2D map from the aerial photo. This map and the height data are used later to produce the 3D model of the study area.
