- Usually bootstrapped from GUS-style rule-based systems in a semi-supervised learning manner.
- A rule-based system is first built for the domain, and a test set is carefully
labeled -> As new user utterances come in, they are paired with the labeling provided
by the rule-based system to create training tuples -> A classifier can then be trained
on these tuples, using the test set to test the performance of the classifier against
the rule-based system (some heuristics can be used to eliminate errorful training
tuples, with the goal of increasing precision).
- The resulting classifier can often outperform the original rule-based system


Saint Olaf College

- Slot-filling, and the simpler tasks of domain and intent classification, are
special cases of the task of supervised semantic parsing,  in which we have a training set that associates each sentence with the correct set of slots, domain, and intent.
- A simple method is to train a sequence model to map from input word representation to slot fillers, domain, and intent.


Learn Before

Related