自然语言理解任务(Natural Language Understanding),也被称为SLU(Spoken LU)或者(LU),或语义解码(Semantic Decoding)。
NLU一般分为两个子任务:意图识别(Intent Detection)、槽识别(Slot Filling)。
意图识别是一个分类任务,所以总体思路和一般的文本分类(Text Classification)或者短文本分类(Short TC)的解决思路基本一致。
槽识别是一个识别任务,类似抽取任务,所以一般方法和命名实体识别(Named Entity Recognization)任务基本一致。
主流的发展方向有以下几个:
也叫Belief Tracking
Neural Belief Tracker: Data-Driven Dialogue State Tracking [Nikola Mrksić et al., 2016]
Fully Statistical Neural Belief Tracking [Nikola Mrksić et al., 2018]
In slot-based spoken Dialogue systems, tracking the entities in context can be cast as slot carryover task - only the relevant slots from the dialogue context are carried over to the current turn.
Improving Long Distance Slot Carryover in Spoken Dialogue Systems [Chen et al., 2019]
Training Neural Response Selection for Task-Oriented Dialogue Systems [Henderson et al., 2019]
对于简单的FAQ问答,也可以被称为 Question Answering Pairing 或者 Neural Response Selection 问题
One-Shot Learning for Text-to-SQL Generation [Lee et al., 2019]
FAQ Retrieval using Query-Question Similarity and BERT-Based Query-Answer Relevance [Wataru Sakata et al., 2019]
A Context-aware Natural Language Generator for Dialogue Systems [Duˇsek et al., 2016]