README.md

metadata

dataset_info:
  - config_name: default
    features:
      - name: utterance
        dtype: string
      - name: label
        dtype: int64
    splits:
      - name: train
        num_bytes: 16
        num_examples: 1
    download_size: 1106
    dataset_size: 16
  - config_name: intents
    features:
      - name: id
        dtype: int64
      - name: name
        dtype: string
      - name: tags
        sequence: 'null'
      - name: regexp_full_match
        sequence: string
      - name: regexp_partial_match
        sequence: string
      - name: description
        dtype: 'null'
    splits:
      - name: intents
        num_bytes: 10141
        num_examples: 13
    download_size: 8811
    dataset_size: 10141
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
  - config_name: intents
    data_files:
      - split: intents
        path: intents/intents-*
task_categories:
  - text-classification
language:
  - ru

Dream

This is a text classification dataset. It is intended for machine learning research and experimentation.

This dataset is obtained via formatting another publicly available data to be compatible with our AutoIntent Library.

Usage

It is intended to be used with our AutoIntent Library:

from autointent import Dataset

dream_ru = Dataset.from_datasets("AutoIntent/dream-ru")

Source

This dataset is taken from DeepPavlov Library's repository. It was formatted with our AutoIntent Library:

# define utils
import json
import requests
from autointent import Dataset

def load_json_from_github(github_file: str):
    raw_text = requests.get(github_file).text
    return json.loads(raw_text)

def convert_dream(dream_dict):
    intents = []
    for i, (intent_name, all_phrases) in enumerate(dream_dict["intent_phrases"].items()):
        intent_record = {
            "id": i,
            "name": intent_name,
            "tags": [],
            "regexp_full_match": all_phrases["phrases"],
            "regexp_partial_match": all_phrases.get("reg_phrases", []),
        }
        intents.append(intent_record)
    return Dataset.from_dict({"intents": intents, "train": [{"utterance": "test", "label": 0}]})

# load and format
github_file = "https://raw.githubusercontent.com/deeppavlov/dream/26ff1aaf1c9019c60b74468705bd6d9b7ebc5353/annotators/IntentCatcherTransformers/intent_phrases_RU.json"
dream = load_json_from_github(github_file)
dream_converted = convert_dream(dream)