[TEXTOM Manual] Sentiment Analysis and Document Classification Feature Upgrade (ELECTRA, GPT Models)

 ✅Sentiment Analysis and Document Classification Feature Upgrade (ELECTRA, GPT Models)

TEXTOM's sentiment analysis (document classification) feature has been upgraded! With this upgrade, new models have been integrated, enabling more effective analysis. Here's a brief summary of the key points! *Currently running in beta version.

In the previous document classification system, a Bayesian classifier was used. However, many users faced difficulties due to the inconvenience of manually uploading positive and negative training data. To address these challenges and deliver more accurate analysis results, we have introduced a new model to simplify the process and enhance analysis precision.

The ELECTRA model is a sentiment analysis model based on deep learning, utilizing lightweight pre-training and fine-tuning techniques to analyze the sentiment of text. After fine-tuning for each language, it predicts the probability of belonging to positive or negative sentiment categories for classification. While the previous Bayesian classifier required a minimum of 100 sentiment training data points, this model can perform sentiment analysis with as few as 20 training data points. The simplified labeling process for training data is expected to enhance both accessibility and convenience for users.

① Download the raw data for sentiment analysis from the [Sentiment Analysis] - Document Classification page.

② Create training data by modifying the Excel file (insert sentiment labeling data in column B only).

③ Upload the training data file.

The GPT model is designed to deeply understand text context and analyze sentiment through extensive pre-training. As a large-scale language model, GPT demonstrates exceptional performance in various natural language processing (NLP) tasks, including sentiment analysis, text generation, and Q&A. It provides highly accurate results with ease, requiring no separate training data preparation.

If you're curious about the upgraded and more accurate analysis results, try the enhanced sentiment analysis and document classification features of TEXTOM today!