一种用于智能语音助手的对话短文本语句匹配方法

本发明涉及一种用于智能语音助手的对话短文本语句匹配方法，属于人工智能技术领域。该方法包括：S1：对智能语音助手对话系统中的文本数据进行向量化，使用堆叠CNN和并行CNN提取文本的短语特征；S2：使用堆叠的BiLSTM提取上下文特征，再经过新的句内自注意力机制，提取文本内部的关键特征；S3：使用句间注意力机制提取交互特征，并通过压缩函数将多个交互特效进行聚合压缩，得到文本的匹配特征；S4：将匹配特征输入MLP中，预测出文本的标签并进行后处理。本发明可以有效地对智能语音助手的对话文本进行语句匹配，预测两个文本是否属于同一语义。 The invention relates to a dialogue...

Full description

Saved in:
Bibliographic Details
Format Patent
LanguageChinese
Published 30.08.2024
Subjects
CALCULATING
COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
COMPUTING
COUNTING
ELECTRIC DIGITAL DATA PROCESSING
PHYSICS
Online AccessGet full text

Cover

More Information
Summary:本发明涉及一种用于智能语音助手的对话短文本语句匹配方法，属于人工智能技术领域。该方法包括：S1：对智能语音助手对话系统中的文本数据进行向量化，使用堆叠CNN和并行CNN提取文本的短语特征；S2：使用堆叠的BiLSTM提取上下文特征，再经过新的句内自注意力机制，提取文本内部的关键特征；S3：使用句间注意力机制提取交互特征，并通过压缩函数将多个交互特效进行聚合压缩，得到文本的匹配特征；S4：将匹配特征输入MLP中，预测出文本的标签并进行后处理。本发明可以有效地对智能语音助手的对话文本进行语句匹配，预测两个文本是否属于同一语义。 The invention relates to a dialogue short text statement matching method for an intelligent voice assistant, and belongs to the technical field of artificial intelligence. The method comprises the following steps: S1, vectorizing text data in an intelligent voice assistant dialogue system, and extracting phrase features of a text by using a stacked CNN and a parallel CNN; s2, using stacked BiLSTM to extract context features, and extracting key features in a text through a new in-sentence self-attention mechanism; s3, extracting interaction features by using an inter-sentence attention mechanism, and aggregating and compressing the plurality of interaction special effects through a compression function to obtain matching features of the text; and S4,
Bibliography:Application Number: CN202111422626