Adapt Tutorial

Search…
Adapt Tutorial
A step by step tutorial in using Adapt Intent Parser, using the `single_intent_parser.py` file as a starting point.
Prerequisites
To complete this tutorial, you should have some basic knowledge of the Python programming language, and be comfortable executing commands on a Linux command line.
If you are looking to use Adapt in a Mycroft Skill, please see Skill Development > Intents​
Step 1 - Sample Intent
This is the sample Intent around which the tutorial is based.
1
A sample intent that uses a fixed vocabulary to extract entities for an intent
2
​
3
try with the following:
4
PYTHONPATH=. python examples/single_intent_parser.py "what's the weather like in tokyo"
Copied!
Step 2 - Import Libraries
First, we need to import json for serializing the Adapt Intent Parser output, and sys for reading in command line arguments.
1
import json
2
import sys
Copied!
Next, we import the IntentBuilder and `IntentDeterminationEngine.
1
from adapt.intent import IntentBuilder
2
from adapt.engine import IntentDeterminationEngine
Copied!
Next, we instantiate an IntentDeterminationEngine object.
1
engine = IntentDeterminationEngine()
Copied!
Next, we delcare a collection of weather Keywords, in JSON syntax. These Keywords act as hints to the Adapt Intent Parser about which intent context is being referenced by an Utterance.
1
weather_keyword = [
2
    "weather"
3
]
Copied!
Register each Keyword with the engine.
1
for wk in weather_keyword:
2
    engine.register_entity(wk, "WeatherKeyword")
Copied!
Next, we declare a collection of weather types. These act as a query parameter on a Weather Intent.
For example, in the sentence:
Will it rain in Seattle tomorrow?
the collection of weather types can then be used to determine whether that weather type is occurring in Seattle.
Next, each weather type is registered with the engine.
1
for wt in weather_types:
2
    engine.register_entity(wt, "WeatherType")
Copied!
Next, a collection of locations is declared. These also act as a query parameter on a Weather Intent, and can be used in combination with the weather type collection.
1
locations = [
2
    "Seattle",
3
    "San Francisco",
4
    "Tokyo"
5
]
Copied!
Next, each location is registered with the engine.
1
for loc in locations:
2
    engine.register_entity(loc, "Location")
Copied!
Next, we construct an intent parser. The intent parser is named WeatherIntent and requires both a WeatherKeyword and Location, and can optionally include a WeatherType.
1
weather_intent = IntentBuilder("WeatherIntent")
2
    .require("WeatherKeyword")
3
    .optionally("WeatherType")
4
    .require("Location")
5
    .build()
Copied!
Next, we register the intent parser with the engine.
1
engine.register_intent_parser(weather_intent)
Copied!
We then declare an entry point for the script. @TODO - need to explain here what an entry point is.
1
if __name__ == "__main__":
Copied!
Next, pass the command line arguments to this script as an Utterance into engine.determine_intent(). This function returns a generator, and we then use the generator to iterate through the results.
1
for intent in engine.determine_intent(' '.join(sys.argv[1:])):
Copied!
If the confidence is >0, this is a valid Intent.
1
if intent.get('confidence') > 0:
Copied!
Next, serialize the Intent and print it to stdout.
1
print(json.dumps(intent, indent=4))
Copied!
Of course, you don't just have to output the Intent to stdout - you can use it to build all sorts of tools.
Adapt Examples
Next - Mycroft Technologies
Lingua Franca
Last modified 1yr ago
Copy link
Contents
Prerequisites
Step 1 - Sample Intent
Step 2 - Import Libraries