def task(source, dest, cred_file, percent):
# load Rasa NLU training data
r = MarkdownReader()
with open(source, "r") as fin:
nlu = fin.read()
nlu_train = r.reads(nlu)
translate_client = translate.Client.from_service_account_json(cred_file)
def trans(text):
trans_text = translate_client.translate(text,
source_language="en",
target_language="zh-TW")
logger.info(u'origin: {}, translated: {}'.format(
example.text, trans_text['translatedText']))
return trans_text['translatedText']
nlu_train.training_examples = random_select_samples(
nlu_train.training_examples, percent)
for example in nlu_train.training_examples:
example.text = trans(example.text)
if example.get("entities"):
for entity in example.get("entities"):
entity["value"] = trans(entity['value'])
# Generate Rasa NLU translated training data
w = MarkdownWriter()
w.dump(dest, nlu_train)
def test_dump_nlu_with_responses():
md = """## intent:greet
- hey
- howdy
- hey there
- hello
- hi
- good morning
- good evening
- dear sir
## intent:chitchat/ask_name
- What's your name?
- What can I call you?
## intent:chitchat/ask_weather
- How's the weather?
- Is it too hot outside?
"""
r = MarkdownReader()
nlu_data = r.reads(md)
dumped = nlu_data.nlu_as_markdown()
assert dumped == md
def _write_nlu_yaml(training_data_path: Path, output_path: Path,
source_path: Path) -> None:
reader = MarkdownReader()
writer = RasaYAMLWriter()
training_data = reader.read(training_data_path)
writer.dump(output_path, training_data)
print_success(f"Converted NLU file: '{source_path}' >> '{output_path}'.")
def test_markdown_entity_regex():
r = MarkdownReader()
md = """
## intent:restaurant_search
- i'm looking for a place to eat
- i'm looking for a place in the [north](loc-direction) of town
- show me [chines](cuisine:chinese) restaurants
- show me [chines](22_ab-34*3.A:43er*+?df) restaurants
"""
result = r.reads(md)
assert len(result.training_examples) == 4
first = result.training_examples[0]
assert first.data == {"intent": "restaurant_search"}
assert first.text == "i'm looking for a place to eat"
second = result.training_examples[1]
assert second.data == {
"intent":
"restaurant_search",
"entities": [{
"start": 31,
"end": 36,
"value": "north",
"entity": "loc-direction"
}],
}
assert second.text == "i'm looking for a place in the north of town"
third = result.training_examples[2]
assert third.data == {
"intent":
"restaurant_search",
"entities": [{
"start": 8,
"end": 14,
"value": "chinese",
"entity": "cuisine"
}],
}
assert third.text == "show me chines restaurants"
fourth = result.training_examples[3]
assert fourth.data == {
"intent":
"restaurant_search",
"entities": [{
"start": 8,
"end": 14,
"value": "43er*+?df",
"entity": "22_ab-34*3.A"
}],
}
assert fourth.text == "show me chines restaurants"
def read_inputs_md(input_path):
reader = MarkdownReader()
reader.read(input_path, language='de', fformat='MARKDOWN')
texts = []
cats = []
for message in reader.training_examples:
texts.append(message.text)
cats.append(message.get('intent'))
return texts, cats
def test_check_check_correct_entity_annotations(text: Text, warnings: int):
reader = MarkdownReader()
tokenizer = WhitespaceTokenizer()
training_data = reader.reads(text)
tokenizer.train(training_data)
with pytest.warns(UserWarning) as record:
EntityExtractor.check_correct_entity_annotations(training_data)
assert len(record) == warnings
assert all([excerpt in record[0].message.args[0]]
for excerpt in ["Misaligned entity annotation in sentence"])
def test_markdown_entity_regex(example: Text, expected_num_entities: int):
r = MarkdownReader()
md = f"""
## intent:test-intent
- {example}
"""
result = r.reads(md)
assert len(result.training_examples) == 1
actual_example = result.training_examples[0]
assert actual_example.data["intent"] == "test-intent"
assert len(actual_example.data.get("entities", [])) == expected_num_entities
def test_markdown_order():
r = MarkdownReader()
md = """## intent:z
- i'm looking for a place to eat
- i'm looking for a place in the [north](loc-direction) of town
## intent:a
- intent a
- also very important
"""
training_data = r.reads(md)
assert training_data.nlu_as_markdown() == md
def test_deprecation_warning_logged():
r = MarkdownReader()
md = """
## intent:test-intent
- I want to go to [LA](city:Los Angeles)
"""
with pytest.warns(
FutureWarning,
match=
r"You are using the deprecated training data format to declare "
r"synonyms.*",
):
r.reads(md)
def test_markdown_unespace_tokens():
r = MarkdownReader()
md = """## intent:test-intent
- Hi \\t Can you help me?\\n I want to go to [Alexandria]{"entity": "city"}
"""
expected_num_entities = 1
training_data = r.reads(md)
assert len(training_data.training_examples) == 1
actual_example = training_data.training_examples[0]
assert actual_example.data["intent"] == "test-intent"
assert len(actual_example.data.get("entities",
[])) == expected_num_entities
def _convert_to_yaml(args: argparse.Namespace, is_nlu: bool) -> None:
output = Path(args.out)
if not os.path.exists(output):
print_error_and_exit(
f"The output path '{output}' doesn't exist. Please make sure to specify "
f"an existing directory and try again."
)
training_data = Path(args.data)
if not os.path.exists(training_data):
print_error_and_exit(
f"The training data path {training_data} doesn't exist "
f"and will be skipped."
)
num_of_files_converted = 0
for file in os.listdir(training_data):
source_path = training_data / file
output_path = output / f"{source_path.stem}{CONVERTED_FILE_SUFFIX}"
if MarkdownReader.is_markdown_nlu_file(source_path):
if not is_nlu:
continue
_write_nlu_yaml(source_path, output_path, source_path)
num_of_files_converted += 1
elif not is_nlu and MarkdownStoryReader.is_markdown_story_file(source_path):
_write_core_yaml(source_path, output_path, source_path)
num_of_files_converted += 1
else:
print_warning(f"Skipped file: '{source_path}'.")
print_info(f"Converted {num_of_files_converted} file(s), saved in '{output}'.")
def _reader_factory(fformat: Text) -> Optional["TrainingDataReader"]:
"""Generates the appropriate reader class based on the file format."""
from rasa.nlu.training_data.formats import (
MarkdownReader,
WitReader,
LuisReader,
RasaReader,
DialogflowReader,
NLGMarkdownReader,
)
reader = None
if fformat == LUIS:
reader = LuisReader()
elif fformat == WIT:
reader = WitReader()
elif fformat in DIALOGFLOW_RELEVANT:
reader = DialogflowReader()
elif fformat == RASA:
reader = RasaReader()
elif fformat == MARKDOWN:
reader = MarkdownReader()
elif fformat == MARKDOWN_NLG:
reader = NLGMarkdownReader()
return reader
def test_markdown_entity_regex():
r = MarkdownReader()
md = """
## intent:restaurant_search
- i'm looking for a place to eat
- i'm looking for a place in the [north](loc-direction) of town
- show me [chines](cuisine:chinese) restaurants
- show me [chines](22_ab-34*3.A:43er*+?df) restaurants
"""
result = r.reads(md)
assert len(result.training_examples) == 4
first = result.training_examples[0]
assert first.data == {"intent": "restaurant_search"}
assert first.text == "i'm looking for a place to eat"
second = result.training_examples[1]
assert second.data == {'intent': 'restaurant_search',
'entities': [
{'start': 31,
'end': 36,
'value': 'north',
'entity': 'loc-direction'}
]}
assert second.text == "i'm looking for a place in the north of town"
third = result.training_examples[2]
assert third.data == {'intent': 'restaurant_search',
'entities': [
{'start': 8,
'end': 14,
'value': 'chinese',
'entity': 'cuisine'}]}
assert third.text == "show me chines restaurants"
fourth = result.training_examples[3]
assert fourth.data == {'intent': 'restaurant_search',
'entities': [
{'start': 8,
'end': 14,
'value': '43er*+?df',
'entity': '22_ab-34*3.A'}]}
assert fourth.text == "show me chines restaurants"
def test_markdown_entity_regex(
example: Text,
expected_entities: Optional[List[Dict[Text, Any]]],
expected_text: Text,
):
r = MarkdownReader()
md = f"""
## intent:test-intent
- {example}
"""
result = r.reads(md)
assert len(result.training_examples) == 1
actual_example = result.training_examples[0]
assert actual_example.data["intent"] == "test-intent"
assert actual_example.data.get("entities") == expected_entities
assert actual_example.text == expected_text
def md_format_message(text, intent, entities):
from rasa.nlu.training_data.formats import MarkdownWriter, MarkdownReader
message_from_md = MarkdownReader()._parse_training_example(text)
deserialised_entities = deserialise_entities(entities)
return MarkdownWriter()._generate_message_md(
{
"text": message_from_md.text,
"intent": intent,
"entities": deserialised_entities,
}
)
def md_format_message(text, intent, entities) -> Text:
from rasa.nlu.training_data.formats import MarkdownReader
from rasa.nlu.training_data.formats.readerwriter import TrainingDataWriter
message_from_md = MarkdownReader().parse_training_example(text)
deserialised_entities = deserialise_entities(entities)
return TrainingDataWriter.generate_message(
{
"text": message_from_md.text,
"intent": intent,
"entities": deserialised_entities,
}
)
async def replace_placeholders(self, example: Message, faker_: Faker,
matches: List[Tuple[Any, ...]],
count: int) -> AsyncIterator[Message]:
original_text = await self.rebuild_original_text(example)
for _ in range(count):
text = await self.replace_placeholders_in_text(
example.text, faker_, matches)
original_text = await self.replace_placeholders_in_text(
original_text, faker_, matches)
entities = MarkdownReader._find_entities_in_training_example(
original_text)
new_message = Message.build(text, example.get("intent"), entities)
yield new_message
async def _correct_entities(latest_message: Dict[Text, Any],
endpoint: EndpointConfig,
sender_id: Text) -> List[Dict[Text, Any]]:
"""Validate the entities of a user message.
Returns the corrected entities"""
from rasa.nlu.training_data.formats import MarkdownReader
parse_original = latest_message.get("parse_data", {})
entity_str = _as_md_message(parse_original)
question = questionary.text(
"Please mark the entities using [value](type) notation",
default=entity_str)
annotation = await _ask_questions(question, sender_id, endpoint)
# noinspection PyProtectedMember
parse_annotated = MarkdownReader()._parse_training_example(annotation)
corrected_entities = _merge_annotated_and_original_entities(
parse_annotated, parse_original)
return corrected_entities
from rasa.nlu.training_data.formats import MarkdownReader
import xlsxwriter
workbook = xlsxwriter.Workbook('filename.xlsx')
worksheet = workbook.add_worksheet()
worksheet.write('A1', 'question')
worksheet.write('B1', 'label')
worksheet.write('C1', 'answer')
training_data = ()
row = 1
col = 0
doc = "PATH\\TO\\nlu.md"
reader = MarkdownReader()
reader.read(doc, language='de', fformat='MARKDOWN')
for message in reader.training_examples:
training_data = training_data + ([message.text, message.get('intent')], )
for question, label in (training_data):
worksheet.write_string(row, col, question)
worksheet.write_string(row, col + 1, label)
worksheet.write_string(row, col + 2, '')
row += 1
workbook.close()
