Sentiment Analysis on a long text

Generic
1

FINBert (and many sentiment analysis programs) seem to have a limiation of 512 characters.  I am interested in the strategies people use to analyze text longer than 512 characters.

2

Hello Tim, 



There are several strategies that people use to analyze text longer than 512 characters in sentiment analysis, including:

  • Chunking: dividing the text into smaller, manageable chunks and analyzing each chunk separately.
  • Sampling: selecting a representative subset of the text for analysis, rather than analyzing the entire text.
  • Summarization: creating a summary of the text using text summarization techniques and analyzing the sentiment of the summary.
  • Pre-processing: removing irrelevant or redundant information from the text, such as stop words or non-relevant entities, before analyzing the sentiment.
Alternatively, you can explore the use of other transformer-based models, such as Longformer or Reformer, which are designed to handle longer sequences and could be fine-tuned for financial sentiment analysis tasks.

I hope this helps!