jcabot

def test_intent_with_ner_initialization():
    entity: CustomEntity = CustomEntity('city_entity', [CustomEntityEntry('Barcelona', ['BCN']), CustomEntityEntry('Madrid')])
    intent: Intent = Intent('intent_name', ['what is the weather like in mycity', 'forecast for mycity', 'is it sunny?'], [EntityReference('city', 'mycity', entity)])

jcabot

class Entity:
    """An entity to be recognized as part of the matching process"""

    def __init__(self, name: str):
        self.name: str = name


class CustomEntityEntry:
    """Each one of the entries (and its synonyms) a CustomEntity consists of"""

jcabot

@app.post("/bot/{name}/train/")
def bot_train(name: str, configurationdto: ConfigurationDTO):
    if name not in bots.keys():
        raise HTTPException(status_code=422, detail="Bot does not exist")
    bot: Bot = bots[name]
    if len(bot.contexts) == 0:
        raise HTTPException(status_code=422, detail="Bot is empty, nothing to train")
    bot.configuration = configurationdto_to_configuration(configurationdto)
    train(bot)
    return {"status:" : "successful training for " + str(len(bot.contexts)) + " contexts"}

jcabot

def predict(context: NLUContext, sentence: str, configuration: NlpConfiguration) -> numpy.ndarray:

    prediction: numpy.ndarray
    sentences = [preprocess_prediction_sentence(sentence, configuration)]
    sequences = context.tokenizer.texts_to_sequences(sentences)
    if configuration.discard_oov_sentences and all(i==1 for i in sequences[0]):
        # the sentence to predict consists of only out of focabulary tokens so we can automatically assign a zero probability to all classes
        prediction = numpy.zeros(len(context.intents))
    else:
        padded = tf.keras.preprocessing.sequence.pad_sequences(sequences, padding='post', maxlen=configuration.input_max_num_tokens, truncating='post')

jcabot

def train(bot: Bot):
    for context in bot.contexts:
        __train_context(context, bot.configuration)


def __train_context(context: NLUContext, configuration: NlpConfiguration):
    tokenizer = tf.keras.preprocessing.text.Tokenizer(num_words=configuration.num_words, lower=configuration.lower, oov_token=configuration.oov_token)
    total_training_sentences: list[str] = []
    total_labels_training_sentences: list[int] = []
    for intent in context.intents:

jcabot

class Intent:
    def __init__(self, name: str, training_sentences: list[str]):
        self.name: str = name
        self.training_sentences: list[str] = training_sentences

class NLUContext:
    def __init__(self, name: str):
        self.name: str = name
        self.intents: list[Intent] = []

jcabot

    model: tf.keras.models = tf.keras.Sequential([
        tf.keras.layers.Embedding(input_dim=configuration.num_words, output_dim=configuration.embedding_dim, input_length=configuration.input_max_num_tokens),
        tf.keras.layers.GlobalAveragePooling1D(),
        tf.keras.layers.Dense(24, activation='relu'),
        tf.keras.layers.Dense(24, activation='relu'),
        tf.keras.layers.Dense(len(context.intents), activation='sigmoid')  
    ])
    model.compile(loss=tf.keras.losses.SparseCategoricalCrossentropy(), optimizer='adam', metrics=['accuracy'])

jcabot

<div id="xatkit-chat"></div>
<script>
   xatkit.renderXatkitWidget({
      server: "https://yourboturl.com/chat-handler",
      widget: {
         title: "Xatkit Demonstration",
         subtitle: "Add a bot to your Carrd site",
         startMinimized: true,
     },
     storage: { autoClear: true }

jcabot

<link rel="stylesheet" href="https://dev.xatkit.com/static/xatkit.min.css">
<script src="https://dev.xatkit.com/static/xatkit.min.js"></script>

jcabot

@app.route('/ask_QA')
def ask_QA():
    query = request.args.get('question')
    print("Question received: ", query)
    # Connect to Elasticsearch (locally installed)
    document_store = ElasticsearchDocumentStore(host="localhost", username="", password="", index="blogposts")

    retriever = ElasticsearchRetriever(document_store=document_store)
    reader = FARMReader(model_name_or_path="deepset/roberta-base-squad2", use_gpu=True)