Intents

Where as an Utterance is what a user might say, an Intent is what they mean.

In the context of a digital assistant we use Intents to help us reply to the users question. Behind the scenes the platform will take the utterance and pass it through some Artificial Intelligence that uses Natural Language Processing to recognise the Intent from the users Utterance. To do this it needs training.

The training is a collection of utterances that provide a representative sample of ways a user might mean the same thing. The NLP uses this information, together with the understanding of the language to understand how words are used together. By doing this it can then establish how close similar words are.

For example, let's consider the following utterances, which represent the Bin Collection intent.

• When are my bins being collected?

• What day is my bin collection day?

• Is my green bin being collected today?

• Remind me what day my recycling is being collected please.

Based on these examples, the NLP will be able to establish that bin or bins is frequently used when conveying the Bin Collection intent. Furthermore, because of the ways these words are used in the (English) language, it can also deduce that rubbish and recycling are equivalent words because they are often used in similar contexts, and in similar frequencies. With that in mind, this training means that if someone were to then type I can't remember what day my rubbish is being collected, because of the similarity of the words used, it'll still be identified as the Bin Collection intent.

We often refer to the process of identifying the users Intent as Classification. As part of the process the users utterance will be classified against all available intents and given a score between 0 and 1. The intent that has the highest score ( also known as Confidence ) will be the Intent that this utterance will be classified as.

Once a users Intent has been identified this can then be used to determine what response to provide to the user.