Sentiment Analysis of Love Letters: Comparative Evaluation of TextBlob, Vader, Flair, and Hugging Face Transformer.

This paper explores the computer identification and quantification of ideas and emotions represented in text, with a particular emphasis on the difficulties that current sentiment analysis methods encounter when used in complex and private contexts such as love letters. After randomly selecting 500 phrases from a corpus of 300 love letters, four well-known Python modules (TextBlob, Vader, Flair, and Hugging Face Transformer) were used to assess the polarity and strength of sentiments.

In order to compensate for the deficiency of labeled data, human experts evaluated the precision and caliber of sentiment annotations. Two separate blind rounds with four judges each were used to compute inter-rater agreements, and a randomly selected