Model descriptions

更新时间:
复制 MD 格式

Conversational text classification offers several models. Use this guide to select the model that best fits your scenario.

  • Conversational Classification - High-Performance Edition: This model is based on a self-developed Hierarchical Attention Network (HAN) architecture that uses a dual-layer BiLSTM with attention. It is fast, requires minimal computing resources, and is ideal for high-performance scenarios.

  • Conversational Classification - High-Accuracy Edition (StructBERT): This model is based on the StructBERT classification architecture from the AliceMind model family. It provides high accuracy and is ideal for scenarios that require high accuracy but have low performance requirements.

  • Detailed comparison

  • Model

    Performance

    Results

    Supported task types

    Notes

    Conversational Classification - High-Performance Edition

    Faster

    Multitasking, multiclass classification, and multi-label

    Conversational Classification - High-Accuracy Edition (StructBERT)

    Good

    Multitasking, multiclass classification, multi-label, and multi-level

    Recommended