Response selector is not working

Rasa Open Source
1

Response selector is not working properly in Rasa 1.8 (TED Policy , DIET classifier) means not giving correct answer.

2

Hi @ravikrcs

What data did you test on?

3

Normal FAQ kinda of… Basically it’s not picking intents properly I am surprised why FAQ kinda question answer is not working properly.

e.g, User : who are you? Bot : I’m great! Thanks for asking.

However in my response.md

bot_intro

  • chitchat/bot_intro
    • Hi I’m Lxie! How can i help you

what doing

  • chitchat/ask_HowDoing
    • I’m great! Thanks for asking.

chitchat.md

intent:chitchat/bot_intro

  • who are you
  • who
  • who r u
  • about yourself
  • tell me about yourself
  • who is this
  • who am i talking to
  • what’s your name
  • who am i speaking with
  • what’s your name bot
  • whats ur name
  • what are you called?
  • who is talking?
  • who speaks?
  • who talks?
  • who are you and what do you want from me
  • who are you?
  • who is it?
  • who are u?
  • I want to know who you are
  • who are y
  • yes but who are you in your heart
  • are you a bot?
  • who is this?
  • hi! what’s your name?
  • who r you
  • you?
  • what’s it name
  • who r u?
  • tell me ur name
  • tell me your name
  • about ur self
  • who is this
  • who are you
  • who am i talking to
  • what’s your name
  • who am i speaking with
  • who r u
  • what’s your name bot
  • whats ur name
  • what are you called?
  • who are you and what do you want from me
  • who are you?
  • who is it?
  • who are u?
  • are you a bot?
  • who is this?
  • hi! what’s your name?
  • am i talking to bot?
  • give me your introduction.
  • introduce yourself.
  • give me your intro.
  • knock knock!!who is it?
  • tell me about yourself

intent:chitchat/ask_HowDoing

  • Ahoy matey how are you?
  • are you alright
  • are you having a good day
  • Are you ok?
  • are you okay
  • Do you feel good?
  • how are things going
  • how are things with you?
  • How are things?
  • how are you
  • how are you doing
  • how are you doing this morning
  • how are you feeling
  • how are you today
  • How are you?
  • how do you do?
  • how have you been
  • how is it going
  • how is your day
  • how is your day going
  • how is your evening
  • How was your day?
  • How’s it going
  • How’s it hanging?
  • how’s life
  • how’s life been treating you?
  • How’s life treating you friend?
  • how’s your day going
  • How’ve you been?
  • I’m fine and you
  • is everything all right
  • Is everything ok?
  • is everything okay
  • what about your day
  • What’s going on?
  • what’s good
  • What’s new?
  • What’s up man
  • What’s up
  • how are you?
  • Do you have a great day?
  • how are you doing today my sweet friend
  • how are you doing?
  • how are u
  • How are you?
4

@ravikrcs Can you please post the complete pipeline configuration that you use?

1 Like
5

Configuration for Rasa NLU.

language: en

pipeline:

  • name: WhitespaceTokenizer

  • name: RegexFeaturizer

  • name: LexicalSyntacticFeaturizer

  • name: CountVectorsFeaturizer

  • name: CountVectorsFeaturizer

    analyzer: “char_wb”

    min_ngram: 1

    max_ngram: 4

  • name: DIETClassifier

    epochs: 100

  • name: EntitySynonymMapper

  • name: ResponseSelector

    epochs: 100

Configuration for Rasa Core.

Policies

policies:

  • name: TEDPolicy

    max_history: 8

    epochs: 100

    batch_size:

    • 32
    • 64

  • max_history: 6

    name: AugmentedMemoizationPolicy

  • name: FallbackPolicy

    nlu_threshold: 0.8

    core_threshold: 0.4

    fallback_action_name: ‘action_default_fallback’

  • name: MappingPolicy

6

@ravikrcs We shipped an important bug fix in ResponseSelector in rasa==1.9.4 . Can you try with that? Also, can you try removing this from your pipeline and then train?

name: CountVectorsFeaturizer

analyzer: “char_wb”

min_ngram: 1

max_ngram: 4

Another alternative is try using ConveRTTokenizer and ConveRTFeaturizer since your training data is in English. See Choosing a Pipeline

7

Thanks @dakshvar22, i will try😊

8

Hey did you try and did it work ?