Skip to content

Latest commit

 

History

History
150 lines (121 loc) · 4.93 KB

README.md

File metadata and controls

150 lines (121 loc) · 4.93 KB
                   _           _  _                  
                  | |         | |(_)                 
 _   _ _____  ____| |__  _____| | _  ___ _____  ____ 
| | | | ___ |/ ___)  _ \(____ | || |/___) ___ |/ ___)
 \ V /| ____| |   | |_) ) ___ | || |___ | ____| |    
  \_/ |_____)_|   |____/\_____|\_)_(___/|_____)_|

Verbaliser

The Verbaliser produces Roboy's utterances. It diversifies the interactions with Roboy by randomizing the output given a specific intent.

Using the Verbaliser

Question-Answer Lists

YAML files are used to define the actual utterances. In other words: they store everything Roboy can vocalise. To diversify his remarks the Verbaliser randomises similar outputs.

The class QAPhrases retrieves the values from a YAML file and parses the containing phrases. Here is a template for such a YAML file:

type: qa                                    # remark types: question answering (qa), phrases
name: "INTENT"                              # intent or topic
Q:                                          # possible questions
- "Question phrasing 1"
- "Question phrasing 2"
- "Question phrasing 3"
A:                                          # answers 
  SUCCESS:
  - "Possible answer on success 1"
  - "Possible answer on success 2"
  FAILURE:
  - "Possible answer on failure"
FUP:                                        # follow up questions (for interlocutors that are already known to Roboy)
  Q:
  - "Possible follow up question"
  A:
  - "Possible follow up answer"

See more examples here.

Example for Answering the Question: What happened to the Dinosaurs?

  • Creating the YAML file:

Fill in all the possible answers.

type: qa
name: "DINO"
A:
  SUCCESS:
  - "I am sure it was a mind-boggingly huge meteorite!"
  - "They smoked too much ash!"
  - "A vulcano had flatulences."
  - "The chicken were stronger."
  FAILURE:
  - "I have no idea what you just said."
  - "Sorry, I am only interested in dinosaurs."
  • Adding this file to the Verbaliser:

In this case the file is going to be located in a folder of important facts. However, the single file can similarly be added by itself. The folder is in the same path as the python file adding it.

from ravestate_verbaliser import verbaliser
from os.path import realpath, dirname, join

verbaliser.add_folder(join(dirname(realpath(__file__)), "important_facts_folder"))
  • Using the Verbaliser for fancy outputs:

This outputs an answer to the question. To understand how to analyse the context of the question have a look at the Natural Language Processing README

if input_had_something_to_do_with_dinos and was_a_question:
    ctx["rawio:out"] = verbaliser.get_random_successful_answer("DINO")
else:
    ctx["rawio:out"] = verbaliser.get_random_failure_answer("DINO")
  • Possible conversation flow:
Interlocutor: "What happend to the Dinosaurs?"
Roboy: "The chicken were stronger."

Example for Extracting Phrase Lists

The Verbaliser can also be used to get all the imported phrases for a specific intent as a list.

  • Creating the phrases.yml:
type: phrases
name: "dino"
opts:
- "Dinos can not scratch their backs."
- "Once upon a time these mind-bogglingly huge creatures wandered the earth."
- "The longest Dinosaur was the Argentiosaurus."
---
type: phrases
name: "chicken"
opts:
- " Chickens are not completely flightless."
- " There are more chickens out there than programmers."
- " If I were a chicken for one day I would say: 'Puk Puk Pukaaak'.
  • Adding the file to the Verbaliser:

The YAML file is assumed to be located in the important_phrases folder. The folder is again in the same path as this python script:

from ravestate_verbaliser import verbaliser
from os.path import realpath, dirname, join

verbaliser.add_file(join(dirname(realpath(__file__)), "important_phrases", "phrases.yml"))
  • Using the Verbaliser to get a list of phrases:

Given a specific intent the Verbaliser can be used to return a list of phrases.

import ravestate_verbaliser
dino_list = ravestate_verbaliser.verbaliser.get_phrase_list('dino')

The verbaliser:intent Property

The verbaliser:react_to_intent state produces a random phrase output for a given intent. All the possible intents are specified in YAML files that are located in the ravestate_phrases_basic_en folder.

The state reads the verbaliser:intent property and outputs one random phrase in the list with that specific intent. It can therefor be triggered as follows:

Let's assume that phrases.yml is now located in avestate_phrases_basic_en.

@state(cond=s("triggered_by_some_signal"), write="verbaliser:intent")
def say_some_nice_chicken_suff(ctx: ContextWrapper):
    ctx["verbaliser:intent"] = "chicken"