An Empirical Comparison of Portuguese and Multilingual BERT Models for Auto-Classification of NCM Codes in International Trade

Classification problems are common activities in many different domains and supervised learning algorithms have shown great promise in these areas. The classification of goods in international trade in Brazil represents a real challenge due to the complexity involved in assigning the correct category codes to a good, especially considering the tax penalties and legal implications of a misclassification. This work focuses on the training process of a classifier based on bidirectional encoder representations from transformers (BERT) for tax classification of goods with MCN codes which are the official classification system for import and export products in Brazil. In particular, this article presents results from using a specific Portuguese-language-pretrained BERT model, as well as results from using a multilingual-pretrained BERT model. Experimental results show that Portuguese model had a slightly better performance than the multilingual model, achieving an MCC 0.8491, and confirms that the classifiers could be used to improve specialists’ performance in the classification of goods.

The ubiquity of selective attention in the processing of feedback during category learning

Feedback is essential for many kinds of learning, but the cognitive processes involved in learning from feedback are unclear. Models of category learning incorporate selective attention to stimulus features while generating a response, but during the feedback phase of an experiment, it is assumed that participants receive complete information about stimulus features as well as the correct category. The present work looks at eye tracking data from six category learning datasets covering a variety of category complexities and types. We find that selective attention to task-relevant information is pervasive throughout feedback processing, suggesting a role for selective attention in memory encoding of category exemplars. We also find that error trials elicit additional stimulus processing during the feedback phase. Finally, our data reveal that participants increasingly skip the processing of feedback altogether. At the broadest level, these three findings reveal that selective attention is ubiquitous throughout the entire category learning task, functioning to emphasize the importance of certain stimulus features, the helpfulness of extra stimulus encoding during times of uncertainty, and the superfluousness of feedback once one has learned the task. We discuss the implications of our findings for modelling efforts in category learning from the perspective of researchers trying to capture the full dynamic interaction of selective attention and learning, as well as for researchers focused on other issues, such as category representation, whose work only requires simplifications that do a reasonable job of capturing learning.

An Empirical Comparison of Portuguese and Multilingual BERT Models for Auto-Classification of NCM Codes in International Trade

The classification of goods involved in international trade in Brazil is based on the Mercosur Common Nomenclature (NCM). The classification of these goods represents a real challenge due to the complexity involved in assigning the correct category codes especially considering the legal and fiscal implications of misclassification. This work focuses on the training of a classifier based on Bidirectional En-coder Representations from Transformers (BERT) for the tax classification of goods with NCM codes. In particular, this article presents results from using a specific Portuguese Language tuned BERT model as well results from using a Multilingual BERT. Experimental results justify the use of these models in the classification process and also that the language specific model has a slightly better performance.

Morphological residual convolutional neural network (M-RCNN) for intelligent recognition of wear particles from artificial joints

Finding the correct category of wear particles is important to understand the tribological behavior. However, manual identification is tedious and time-consuming. We here propose an automatic morphological residual convolutional neural network (M-RCNN), exploiting the residual knowledge and morphological priors between various particle types. We also employ data augmentation to prevent performance deterioration caused by the extremely imbalanced problem of class distribution. Experimental results indicate that our morphological priors are distinguishable and beneficial to largely boosting overall performance. M-RCNN demonstrates a much higher accuracy (0.940) than the deep residual network (0.845) and support vector machine (0.821). This work provides an effective solution for automatically identifying wear particles and can be a powerful tool to further analyze the failure mechanisms of artificial joints.

The benefits of impossible tests: Assessing the role of error-correction in the pretesting effect

Relative to studying alone, guessing the meanings of unknown words can improve later recognition of their meanings, even if those guesses were incorrect – the pretesting effect (PTE). The error-correction hypothesis suggests that incorrect guesses produce error signals that promote memory for the meanings when they are revealed. The current research sought to test the error-correction explanation of the PTE. In three experiments, participants studied unfamiliar Finnish-English word pairs by either studying each complete pair or by guessing the English translation before its presentation. In the latter case, the participants also guessed which of two categories the word belonged to. Hence, guesses from the correct category were semantically closer to the true translation than guesses from the incorrect category. In Experiment 1, guessing increased subsequent recognition of the English translations, especially for translations that were presented on trials in which the participants’ guesses were from the correct category. Experiment 2 replicated these target recognition effects while also demonstrating that they do not extend to associative recognition performance. Experiment 3 again replicated the target recognition pattern, while also examining participants’ metacognitive recognition judgments. Participants correctly judged that their memory would be better after small than after large errors, but incorrectly believed that making any errors would be detrimental, relative to study-only. Overall, the data are inconsistent with the error-correction hypothesis; small, within-category errors produced better recognition than large, cross-category errors. Alternative theories, based on elaborative encoding and motivated learning, are considered.

Weighting Schemes and Incomplete Data: A Generalized Bayesian Framework for Chance-Corrected Interrater Agreement

Van Oest (2019) developed a framework to assess interrater agreement for nominal categories and complete data. We generalize this framework to all four situations of nominal or ordinal categories and complete or incomplete data. The mathematical solution yields a chance-corrected agreement coefficient that accommodates any weighting scheme for penalizing rater disagreements and any number of raters and categories. By incorporating Bayesian estimates of the category proportions, the generalized coefficient also captures situations in which raters classify only subsets of items; that is, incomplete data. Furthermore, this coefficient encompasses existing chance-corrected agreement coefficients: the S-coefficient, Scott’s pi, Fleiss’ kappa, and Van Oest’s uniform prior coefficient, all augmented with a weighting scheme and the option of incomplete data. We use simulation to compare these nested coefficients. The uniform prior coefficient tends to perform best, in particular, if one category has a much larger proportion than others. The gap with Scott’s pi and Fleiss’ kappa widens if the weighting scheme becomes more lenient to small disagreements and often if more item classifications are missing; missingness biases play a moderating role. The uniform prior coefficient usually performs much better than the S-coefficient, but the S-coefficient sometimes performs best for small samples, missing data, and lenient weighting schemes. The generalized framework implies a new interpretation of chance-corrected weighted agreement coefficients: These coefficients estimate the probability that both raters in a pair assign an item to its correct category without guessing. Whereas Van Oest showed this interpretation for unweighted agreement, we generalize to weighted agreement.

A Filtering Algorithm of Main Word Frequency for Online Commodity Subject Classification in E-Commerce

—Based on the traditional classification of plain text in E-Commerce, this article has put forward a processing method in accordance with semi-structured data and main information in web pages, which enhances the accuracy of the product distribution. On the basis of the traditional textmining, combined with the structure and links of web page, this article has proposed an improved web page text representation model in E-Commerce based on supporting vector machines and web text classification algorithm, but there are still a lot of shortcomings waiting for further improvement. According to the data contrast in precision ratio, recall ratio and F-measure, the effect of the improved experiment with LDF-IDF is comprehensively better than that of tf-idf. The precision rate in certain classification can reach 100%, but there is low precision rate caused by items with fewer samples or samples fuzziness. Therefore, the classification of the correct category will directly affect the effect of classification.

Analysis of sentiment based movie reviews using machine learning techniques

The decisions and approaches of renowned personality used to impress the real world are to a great extent adapted to how others have seen or assessed the world with opinion and sentiment. Examples could be any opinion and sentiment of people view about Movie audits, Movie surveys, web journals, smaller scale websites, and informal organizations. In this research classifies the movie review into its correct category, classifier model is proposed that has been trained by applying feature extraction and feature ranking. The focus is on how to examine the sentiment expression and classification of a given movie review on a scale of (–) negative and (+) positive sentiments analysis for the IMDB movie review database. Due to the lack of grammatical structures to comments on movies, natural language processing (NLP) has been used to implement proposed model and experimentation is performed to compare the present study with existing learning models. At the outset, our approach to sentiment classification supplements the existing movie rating systems used across the web to an accuracy of 97.68%.

Alphas: A Case study in International Institutional Mutual Funds

In this paper we show that not taking into account the fact that fund managers “deviate” from their stated categories biases upward their alphas. When evaluating fund managers most studies compare managers against the S&P 500 regardless of the sectors managers actually invest in. This procedure does not take into account that an important proportion of US stock managers invest in medium and small companies. This neglect biases performance results. In the international stock arena, not only do studies use the incorrect benchmark but they also neglect to take into account the fact that managers deviate from their stated sector. In this paper we not only employ the correct category the managers invest in but we also take into account the fact that managers systematically drift away from their stated category. This drift occurs for approximately half the funds examined and causes the estimated alpha of managers to be on average 45 basis points higher than it should be if we were to undertake the multiple regression that fund drift demands. In addition to using the right benchmarks, adjusting for “drift” in this paper we chose to use as “benchmarks” the ETF’s in each category so as to compare managers not against theoretical constructs, but against an actual investable vehicle in the corresponding category.

Pengembangan Media Pembelajaran Berbasis Flash Mx Pada Mata Pelajaran Muhadatsah

The research using the ADDIE went through these phases: (1) Analysis Phase was problem analysis, material analysis, learners character analysis, learning purpose, and learners need. (2) The design phase was making the media in Microsoft PowerPoint and steps of producing media in Flash MX. (3) Development Phase was validating the product to material expert and media expert. (4) The implementation phase was practicing for the class tent of MA Al-I'dadiyyah Bahrul Ulum learners with 22 learners to do effectiveness test, which is pre and post-test after Macromedia Flash MX based learning media was given. (5) Evaluation Phase was to see the effective and evaluation was extracted from learners response. The validated media according to assessment from material expert validation (Arab Language Department's Lecturer) was 4,1 with very valid category, evaluation from the material expert (muhadatsah course teacher) was 4,6 with very accurate class, whereas media expert was 3,7 with correct category and learners response as much as 87% with very agree to category. Effectiveness test was from normality test, homogeneity test, and simple paired t-test concluded that Ha accepted than Ho rejected means Macromedia Flash MX used media development was effective in muhadatsah learning in the tenth class of MA Al-I'dadiyyah.