A Lightweight Sentiment Analysis Framework for a Micro-Intelligent Terminal.

Area of Science:

• Natural Language Processing
• Machine Learning
• Artificial Intelligence

Background:

• Sentiment analysis is crucial for intelligent terminals to understand customer opinions.
• Current state-of-the-art models often require significant computational resources, hindering deployment on micro-intelligent terminals.
• There is a need for efficient and lightweight sentiment analysis models.

Purpose of the Study:

• To propose a lightweight and efficient framework for sentiment analysis suitable for devices with limited computing power.
• To enhance the accuracy and speed of sentiment classification through a novel embedding approach.

Main Methods:

• Developed a hybrid multi-grained embedding framework named MC-GGRU.
• Incorporated a global attention structure within a gated recurrent unit (GRU) to learn contextual representations from word tokens.
• Utilized a multi-grained feature layer to enrich sentence representations with implicit character-level semantics.

Main Results:

• MC-GGRU achieves high inference performance with a shallow network structure.
• The proposed method demonstrates state-of-the-art (SOTA) performance in sentiment classification across five public datasets.
• The framework offers a favorable trade-off between accuracy and inference speed.

Conclusions:

• MC-GGRU provides an effective solution for deploying advanced sentiment analysis on resource-constrained intelligent terminals.
• The hybrid multi-grained embedding approach successfully enhances model efficiency and performance.
• This work contributes to making sophisticated NLP tasks more accessible on edge devices.