Using Dependency Relations for Intent Recognition

After extracting the entities, we want to find out what sort of intent the user carries—to book a flight, to purchase a meal on their already booked flight, cancel their flight, and so on. If you look at the intents list again, you will see that every intent includes a verb (to book) and an object that the verb acts on (flight, hotel, meal).

In this lesson, we'll extract transitive verbs and their direct objects from utterances. We'll begin our intent recognition section by extracting the transitive verb and the direct object of the verb. Then, we'll explore how to understand a user's intent by recognizing synonyms of verbs and nouns. Finally, we'll see how to determine a user's intent with semantic similarity methods. Before we move on to extracting transitive verbs and their direct objects, let's first quickly go over the concepts of transitive verbs and direct/indirect objects.

Linguistic primer

In this section, we'll explore some linguistic concepts related to sentence structure, including verbs and verb-object relations. A verb is a very important component of the sentence as it indicates the action in the sentence. The object of the sentence is the thing/person that is affected by the action of the verb. Hence, there's a natural connection between the sentence verb and objects. The concept of transitivity captures verb-object relations. A transitive verb is a verb that needs an object to act upon. Let's see some examples: