Several tests were necessary to calibrate the queries, which seem to perform poorly when a # is used. A watch of the trends posted on Twitter also led to the addition of the keywords ARN and mRNA, since the corpus analysis will be carried out in French and English.
Each retrieved dataset is first cleaned with Open Refine: column mergers are sometimes necessary because the “text” column is sometimes split into several columns (recording with comma separator). The three corpora with a defined geographical area present fewer quality problems than the general corpus, which targets all directions: big data does not necessarily mean good data.
Academic readings:
- Deng, S., Sinha, A. P., & Zhao, H. (2017). Adapting sentiment lexicons to domain-specific social media texts. Decision Support Systems, 94, 65-76.
- Mowlaei, M. E., Abadeh, M. S., & Keshavarz, H. (2020). Aspect-based sentiment analysis using adaptive aspect-based lexicons. Expert Systems with Applications, 148, 113234.