Abstract

Language understanding is an important aspect of supporting spoken queries on a device. Intensive ML models are unsuitable on resource-constrained devices and in such cases, pattern matching grammar can be utilized. However, pattern matching can have low recall and may be limited in understanding user queries. This disclosure describes the use of normalization, optional/stopwords, synonym clusters, and argument expansion to generate more expansive grammar with limited training data and without the use of machine learning. The use of such a grammar can improve recall while maintaining high precision. The techniques are suitable for any context where grammar patterns are used and where user queries are served by resource-constrained devices.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.

Share

COinS