Low Resource Neural Machine Translation: Assamese to/from Other Indo-Aryan (Indic) Languages

Machine translation (MT) systems have been built using numerous different techniques for bridging the language barriers. These techniques are broadly categorized into approaches like Statistical Machine Translation (SMT) and Neural Machine Translation (NMT). End-to-end NMT systems significantly outperform SMT in translation quality on many language pairs, especially those with the adequate parallel corpus. We report comparative experiments on baseline MT systems for Assamese to other Indo-Aryan languages (in both translation directions) using the traditional Phrase-Based SMT as well as some more successful NMT architectures, namely basic sequence-to-sequence model with attention, Transformer, and finetuned Transformer. The results are evaluated using the most prominent and popular standard automatic metric BLEU (BiLingual Evaluation Understudy), as well as other well-known metrics for exploring the performance of different baseline MT systems, since this is the first such work involving Assamese. The evaluation scores are compared for SMT and NMT models for the effectiveness of bi-directional language pairs involving Assamese and other Indo-Aryan languages (Bangla, Gujarati, Hindi, Marathi, Odia, Sinhalese, and Urdu). The highest BLEU scores obtained are for Assamese to Sinhalese for SMT (35.63) and the Assamese to Bangla for NMT systems (seq2seq is 50.92, Transformer is 50.01, and finetuned Transformer is 50.19). We also try to relate the results with the language characteristics, distances, family trees, domains, data sizes, and sentence lengths. We find that the effect of the domain is the most important factor affecting the results for the given data domains and sizes. We compare our results with the only existing MT system for Assamese (Bing Translator) and also with pairs involving Hindi.

Similarity Comparisons of SARS-CoV-2 Samples between Wuhan and G20

To analyze SARS CoV-2 genomes, this paper uses a viral genome collected in Wuhan as a basic sequence to select a set of genomes from G20 countries/regions for comparison. Two methods are applied in visualization followed in the A1 and C1 modules of the MAS to provide 1D similarity projection and information entropy approaches. For a pair of two genomes segmented as M segments to calculate base differences corresponding to each segment, a measuring vector can be generated to be projected as a visual map. Refined information will be provided to calculate the information entropy corresponding to base differences. This provides quantitative measures to organize these genomes arranged into a linear order. The first method uses a line chart representation to illustrate relevant differences among genomes 1 by 1 between Wuhan and other regions. The second method uses a histogram representation to observe the information entropy projection results. Significant similarities are observed among the differences with approximately equal entropies with higher similarity. Based on the comparison of gene sequence diagrams between Wuhan and various regions, a difference analysis was carried out.

Similarity Comparisons of SARS-CoV-2 Samples between Wuhan and G20

To analyze SARS CoV-2 genomes, this paper uses a viral genome collected in Wuhan as a basic sequence to select a set of genomes from G20 countries/regions for comparison. Two methods are applied in visualization followed in the A1 and C1 modules of the MAS to provide 1D similarity projection and information entropy approaches. For a pair of two genomes segmented as M segments to calculate base differences corresponding to each segment, a measuring vector can be generated to be projected as a visual map. Refined information will be provided to calculate the information entropy corresponding to base differences. This provides quantitative measures to organize these genomes arranged into a linear order. The first method uses a line chart representation to illustrate relevant differences among genomes 1 by 1 between Wuhan and other regions. The second method uses a histogram representation to observe the information entropy projection results. Significant similarities are observed among the differences with approximately equal entropies with higher similarity. Based on the comparison of gene sequence diagrams between Wuhan and various regions, a difference analysis was carried out.

Reconstruction of Simulated Cylindrical Defects in Acrylic Glass Plate Using Pulsed Phase Thermography

The results of testing of acrylic glass, in which cylindrical defects were simulated at different depths by applying Pulsed Phase Thermography, are presented in the paper. To ensure better visibility of the simulated defects, suitable thermal images were selected and then processed by using two different procedures. In the first procedure, reduced thermal image sequences were generated by uniform extraction from the basic sequence, to enable analysis at different sampling frequencies. The second procedure was based on the application of a window function, which ensured that only uniformly selected thermal images took part in the evaluation of the basic sequence. The remaining thermal images were not used, but they did participate in the determination of the length of the analyzed sequence; in other words, their existence was registered through the number of samples used in Fast Fourier Transformation. The second procedure yielded much better results with regard to the estimation of the shape of a defect and the depth at which it was located. To provide better insight into the development of the thermal process in the defect area, an additional analysis of pixel intensity variation in the time domain was undertaken.

Înmormântarea în Maramureș după răspunsurile la chestionarele lui Ion Mușlea

Our work is based on the material provided by the answers to Ion Muşlea's questionnaires on the funeral in Maramureş, which raises a special interest to the researcher due to the rich ritual and ceremonial acts, beliefs and superstitions that give complexity, individuality and also the status of regional variant to the funeral in this area. After brief general considerations about the funeral, in which we emphasized its characteristic traits, especially the historical evolution of this custom, we proceeded to synthesize and systematize the information obtained in the localities Vadu Izei, Rona de Jos, Hărniceşti and Giuleşti, following the succession of the sequences in the funeral ceremony. The first part of our paper was reserved for the signs of death, then we talked about the period before the agony, and finally about the agony of the dying person. The largest part of our work concerns the burial itself, as it is the basic sequence of the funeral ceremony and it has a very rich repertoire, made up of all kinds of ritual-magical acts related to different sequences of the custom, some of them complex - such as the wake ceremony. We included information about the post-burial sequence, where the ritual funeral repasts are predominant. Our work ends with the beliefs and superstitions that once circulated intensely during the funeral, the emphasis being on the superstitions - reproduced separately by the investigator - about male or female ghosts; then the beliefs related to the soul, to the afterlife, to apparitions and to death. In the context of using the material investigated by Ion Muşlea's correspondents, we have made some interpretations focusing on decoding the symbols of some magical instruments or vegetal elements used in the magical practices and ritual acts of the funeral in Maramures. Our guidelines throughout the work was our concern to show the entire complexity, richness and variety of the funeral habits in this area.

Derivational affixes in Japanese and Indonesian

Japanese and Indonesian seen from morphological typology is an agglutinative language in which the morphological processes are done by affixation, ie by adding prefixes, suffixes and infixes. While the basic sequence sentence structure has a SOV sequence pattern for Japanese and SVO for Indonesian language. The predicate filled by the verb is capable of binding arguments in constructing the clause structure, so that there are verbs with one, two and three arguments, it depends on the type of verb. This study is a preliminary study of Japanese and Indonesian derivative verbs: the study of linguistic typology. The Theory of Linguistic Typology is used to analyze the formation of Japanese and Indonesian derivative verbs in which the verb serves as the core of the predicate to bind the argument in constructing the clause structure. From the perspective of linguistic typology, the results of the analysis show that (1) the basic form of Japanese derivative verb formers are adjectives (keiyoushi) and noun verbs, whereas Indonesian derivative verbs are derived from adjectives, nouns and pre-categorical. (2) The Japanese derivation affixes joining the adjective (keiyoushi) are -める meru, -まるmaru, -がるgaru’, -むmu and which joins the noun verb is -するsuru. While the derivational affix of the Indonesian language that joined the nouns are meng-, ber-, ter-, ke-an, ber-an, ber-kan, per-, -i, per-i, per-kan, the affix that joins the adjective are meng-, ber-, ter-, ke-an, ber-an, ber-kan, per-, -kan, per-i, dan –i and the affixes that join the pre-categorical are meng-, ter-, ber-, ber-an, -i,-kan.