Search papers, labs, and topics across Lattice.
This paper evaluates Naive Bayes, BiLSTM, and BERT models for document categorization, focusing on the trade-off between accuracy and computational efficiency in real-world automation pipelines. The study found that BERT achieved the highest accuracy (99%+) but with high computational cost, while BiLSTM offered a good balance (98.56% accuracy) with moderate training costs. A demonstrative system was built to validate the practical applicability of these models for automated routing of technical requests.
For document categorization, BiLSTMs strike the best balance between accuracy and computational cost, outperforming Naive Bayes and rivaling BERT's accuracy at a fraction of the training time.
This article presents an evaluation of several machine learning methods applied to automated text classification, alongside the design of a demonstrative system for unbalanced document categorization and distribution. The study focuses on balancing classification accuracy with computational efficiency, a key consideration when integrating AI into real world automation pipelines. Three models of varying complexity were examined: a Naive Bayes classifier, a bidirectional LSTM network, and a fine tuned transformer based BERT model. The experiments reveal substantial differences in performance. BERT achieved the highest accuracy, consistently exceeding 99\%, but required significantly longer training times and greater computational resources. The BiLSTM model provided a strong compromise, reaching approximately 98.56\% accuracy while maintaining moderate training costs and offering robust contextual understanding. Naive Bayes proved to be the fastest to train, on the order of milliseconds, yet delivered the lowest accuracy, averaging around 94.5\%. Class imbalance influenced all methods, particularly in the recognition of minority categories. A fully functional demonstrative system was implemented to validate practical applicability, enabling automated routing of technical requests with throughput unattainable through manual processing. The study concludes that BiLSTM offers the most balanced solution for the examined scenario, while also outlining opportunities for future improvements and further exploration of transformer architectures.
