Iciclecreek.Bot.Builder.Dialogs.Recognizers

Recognizers

This library provides custom recognizers for Bot Framework/Composer.

Installation

To install into your project run the cli:

dotnet add package Iciclecreek.Bot.Builder.Dialogs.Recognizers

In your startup code add:

ComponentRegistration.Add(new RecognizerComponentRegistration());

To add to your schema for usage in Bot Framework Composer from cli:

bf dialog:merge -p yourproj.proj -o your.schema

Library

This library adds recognizers:

• PersonNameEntityRecognizer - Recognizes common @givenName @surname, @fullname entities (like "John Smith"=> GivenName:"john" Surname:"Smith", fullname:"John Smith")
• QuotedTextEntityRecognizer - Recognizes quoted strings as @QuotedText entities
• CsvTextEntityRecognizer - Uses a .CSV file to define tokens to match to entities
• ThresholdRecognizer - Applies a threshold to intents, any intents which have scores which are within the threshold will trigger an OnChooseIntent event.

PersonNameEntityRecognizer

Recognizes common @givenName @surname, @fullname entities (like "John Smith"=> GivenName:"john" Surname:"Smith", fullname:"John Smith"). You can optionally point it to a URL to extend matched given names and surnames.

Sample Code

    var dialog = new AdaptiveDialog()
    {
        Recognizer = new RegexRecognizer()
        {
            Entities = new EntityRecognizerSet()
            {
                new PersonNamEntityRecognizer()
            }
        }
        ...
    }

Sample Json

{
    "$kind": "Microsoft.AdaptiveDialog",
    "recognizer": {
        "$kind":"Microsoft.RegexRecognizer",
        "entities": [
            {
                "$kind":"Iciclecreek.PersonNameEntityRecognizer"
            }
        ]
    }
}

QuotedTextEntityRecognizer

Use the QuotedTextEntityRecognizer to recognize quoted strings as a @QuotedText entity across 64 languages:

• Afrikaans
• Albanian
• Amharic
• Arabic
• Armenian
• Azerbaijani
• Basque
• Belarusian
• Bosnian
• Bulgarian
• Catalan
• Chinese
• Croatian
• Czech
• Danish
• Dutch
• English
• Esperanto
• Estonian
• Filipino
• Finnish
• French
• Galician
• Georgian
• German
• Greek
• Hebrew
• Hindi
• Hungarian
• Icelandic
• Indonesian
• Interlingua
• Irish
• Italian
• Japanese
• Kazakh
• Khmer
• Korean
• Lao
• Latvian
• Lithuanian
• Macedonian
• Maltese
• New Tai Lue
• Norwegian
• Occitan
• Pashto
• Persian
• Polish
• Romanian
• Russian
• Serbian
• Scottish Gaelic
• Slovak
• Slovene
• Sorbian
• Spanish
• Swedish
• Tai Le
• Tamil
• Tibetan
• Tigrinya
• Thai
• Turkish
• Ukrainian
• Uyghur
• Uzbek
• Vietnamese
• Welsh

Sample Code

    var dialog = new AdaptiveDialog()
    {
        Recognizer = new RegexRecognizer()
        {
            Entities = new EntityRecognizerSet()
            {
                new QuotedTextEntityRecognizer()
            }
        }
        ...
    }

Sample Json

{
    "$kind": "Microsoft.AdaptiveDialog",
    "recognizer": {
        "$kind":"Microsoft.RegexRecognizer",
        "entities": [
            {
                "$kind":"Iciclecreek.QuotedTextEntityRecognizer"
            }
        ]
    }
}

CsvEntityRecognizer

Use the CsvEntityRecognizer to easily define custom entities to recognize from a csv file.

The CSV file should be in the format:

TOKEN,TYPE,VALUE1,..,VALUEN

The value columns are property paths which will be used to set modify the entity with the data in the value column.

Example:

TOKEN,TYPE,genus.class,genus.latin
alligator,animal,reptile,alligator mississippiensis
squirrel,animal,mammal,squirrus maximus
...



#### Sample Code
```C#
    var dialog = new AdaptiveDialog()
    {
        Recognizer = new RegexRecognizer()
        {
            Entities = new EntityRecognizerSet()
            {
                new CsvEntityRecognizer()
                {
                    Url = "http://contoso.com/animals.csv"
                }
            }
        }
        ...
    }

Sample Json

{
    "$kind": "Microsoft.AdaptiveDialog",
    "recognizer": {
        "$kind":"Microsoft.RegexRecognizer",
        "entities": [
            {
                "$kind":"Iciclecreek.CsvEntityRecognizer",
                "Url": "http://contoso.com/animals.csv"
            }
        ]
    }
}

ThresholdRecognizer

Applies a threshold to intents, any intents which have scores which are within the threshold will trigger an OnChooseIntent event.

Sample Code

    var dialog = new AdaptiveDialog()
    {
        Recognizer = new ThresholdRecognizer()
        {
            Threshold = 1.0f,
            Recognizer = ...
        }
        ...
    }

Sample Json

{
    "$kind": "Microsoft.AdaptiveDialog",
    "recognizer": {
        "$kind":"Iciclecreek.ThresholdRecognizer",
        "threshold": 0.2,
        "recognizer": {
            ...
        }
    }
}

Sample code for handling OnChooseIntent

    new OnChooseIntent()
    {
        Intents = new List { "foo","bar" },
        Actions = new List()
        {
            new SetProperty()
            {
                Property = "dialog.candidates",
                Value = $"=turn.recognized.candidates"
            },
            new ChoiceInput()
            {
                    Choices = new ChoiceSet()
                    {
                        new Choice("foo"),
                        new Choice("bar"),
                    },
                    Prompt = new ActivityTemplate("Which intent?")
            },
            new EmitEvent()
            {
                EventName = AdaptiveEvents.RecognizedIntent,
                EventValue = "=first(where(dialog.candidates, c, c.intent == turn.lastResult)).result"
            },
            new DeleteProperty()
            {  
                Property = "dialog.candidates"
            }
        }
    }

Sample code for generically handling OnChooseIntent

If you don't have the Intents constraint, then you can programmatically ask for the intents ( or map the intents to human language) like this:

new ChoiceInput()
{
     Choices = "=select(dialog.candidates, result, result.intent)",
     Prompt = new ActivityTemplate("Which intent?")
},