Comparative Analysis of Text Vectorization Methods

The paper considers methods of vectorization textual properties natural language in the context task intellectual text analysis. most common statistical analysis feature extraction and that taking into account are analyzed. work describes above types embeddings their variations implementations. Their comparative was performed, which showed relationship between type method showing best metrics. topology neural network, is basis for solving problem obtaining metrics, described, implemented. carried out using relative time theory algorithms classification metrics: accuracy, f1-score, precision, recall. metrics taken from results building a network model described framing methods. As result, analyzing tonality text, based on n-grams character sequences turned to be best.