2017/07/06 by José Camacho-Collados, Camacho-Collados, Jose, Mohammad Taher Pilehvar +1 · 2 citations
Computer Science · #Computation and Language (cs.CL) #FOS: Computer and information sciences #Information Retrieval (cs.IR) #Natural Language Processing Techniques #Sentiment Analysis and Opinion Mining #Topic Modeling
paper · pdf · doi:10.48550/arxiv.1707.01780
openalex publication_date 2017/07/06 · openalex created_date 2022/08/19 · openalex updated_date 2026/07/28
Text preprocessing is often the first step in the pipeline of a Natural\nLanguage Processing (NLP) system, with potential impact in its final\nperformance. Despite its importance, text preprocessing has not received much\nattention in the deep learning literature. In this paper we investigate the\nimpact of simple text preprocessing decisions (particularly tokenizing,\nlemmatizing, lowercasing and multiword grouping) on the performance of a\nstandard neural text classifier. We perform an extensive evaluation on standard\nbenchmarks from text categorization and sentiment analysis. While our\nexperiments show that a simple tokenization of input text is generally\nadequate, they also highlight significant degrees of variability across\npreprocessing techniques. This reveals the importance of paying attention to\nthis usually-overlooked step in the pipeline, particularly when comparing\ndifferent models. Finally, our evaluation provides insights into the best\npreprocessing practices for training word embeddings.\n