NILC at CWI 2018: Exploring Feature Engineering and Feature Learning

Abstract

This paper describes the results of NILC team at CWI 2018. We developed solutions following three approaches: (i) a feature engineering method using lexical, n-gram and psycholinguistic features, (ii) a shallow neural network method using only word embeddings, and (iii) a Long Short-Term Memory (LSTM) language model, which is pre-trained on a large text corpus to produce a contextualized word vector. The feature engineering method obtained our best results for the classification task and the LSTM model achieved the best results for the probabilistic classification task. Our results show that deep neural networks are able to perform as well as traditional machine learning methods using manually engineered features for the task of complex word identification in English. * The opinions expressed in this article are those of the authors and do not necessarily reflect the official policy or position of the Itaú-Unibanco.

Lexical Simplification System to Improve Web Accessibility

“…In the BEA workshop [39], on the CWI task for uni/multiword phrase classification, most participant teams preferred ML approaches for their systems. For example, [40] presented three approaches for CWI, one using the traditional classification algorithms of ML based on lexical features (word length, number of syllables, and others) and N-gram features (probabilities of n-gram). Other works outside workshops have also been carried out, such as [41], which used the task dataset to train a convolutional neural network (CNN) with word embeddings and engineered features.…”

JCT at SemEval-2021 Task 1: Context-aware Representation for Lexical Complexity Prediction

“…Some of these works found WordNet (Miller, 1998) as a valuable source of lexical features. The main extracted feature is the number of synsets, but also information on hypernyms, hyponyms, holonym, and meronym is useful (Gooding and Kochmar, 2018;Hartmann and Dos Santos, 2018;Wani et al, 2018).…”

Adaptação Lexical Automática em Textos Informativos do Português Brasileiro para o Ensino Fundamental

“…Com base na nossa experiência no CWI 2018 14 (Hartmann & dos Santos, 2018), em que obtivemos a segunda melhor colocação na tarefa de classificação e terceira melhor colocação na tarefa de classificação probabilística para a língua inglesa (Yimam et al, 2018) Eventuais interseções entre os léxicos dos dicionários foram tratadas. Se uma palavraé complexa para um ano escolar T +2, ela naturalmenté e complexa para os anos escolares T e T + 1.…”

“…Em relaçãoàs etapas de Identificação de Palavras Complexas e Simplificação Lexical, trabalhos recentes têm mostrado que métodos que fazem uso de Feature Learning estão desempenhando melhor do que os métodos que utilizam Feature Engineering (Glavaš &Štajner, 2015;Paetzold & Specia, 2017;Hartmann & dos Santos, 2018;Štajner et al, 2019). Esse cenário está alinhado com os resultados obtidos nesta avaliação.…”