Warning: This document is for the development version of Rasa. The latest version is 1.2.3.

Choosing a Pipeline

Choosing an NLU pipeline allows you to customize your model and finetune it on your dataset.

The Short Answer

If you have less than 1000 total training examples, and there is a spaCy model for your language, use the pretrained_embeddings_spacy pipeline:

language: "en"

pipeline: "pretrained_embeddings_spacy"

If you have 1000 or more labelled utterances, use the supervised_embeddings pipeline:

language: "en"

pipeline: "supervised_embeddings"

A Longer Answer

The two most important pipelines are supervised_embeddings and pretrained_embeddings_spacy. The biggest difference between them is that the pretrained_embeddings_spacy pipeline uses pre-trained word vectors from either GloVe or fastText. The supervised embeddings pipeline, on the other hand, doesn’t use any pre-trained word vectors, but instead fits these specifically for your dataset.

The advantage of the pretrained_embeddings_spacy pipeline is that if you have a training example like: “I want to buy apples”, and Rasa is asked to predict the intent for “get pears”, your model already knows that the words “apples” and “pears” are very similar. This is especially useful if you don’t have very much training data.

The advantage of the supervised_embeddings pipeline is that your word vectors will be customised for your domain. For example, in general English, the word “balance” is closely related to “symmetry”, but very different to the word “cash”. In a banking domain, “balance” and “cash” are closely related and you’d like your model to capture that. This pipeline doesn’t use a language-specific model, so it will work with any language that you can tokenize (on whitespace or using a custom tokenizer).

You can read more about this topic here .

Rasa gives you the tools to compare the performance of both of these pipelines on your data directly, see Comparing NLU Pipelines.

You can also use MITIE as a source of word vectors in your pipeline, see MITIE. We do not recommend that you use these; mitie support is likely to be deprecated in a future release.

Note

Intent classification is independent of entity extraction. So sometimes NLU will get the intent right but entities wrong, or the other way around. You need to provide enough data for both intents and entities.

Multiple Intents

If you want to split intents into multiple labels, e.g. for predicting multiple intents or for modeling hierarchical intent structure, you can only do this with the supervised embeddings pipeline. To do this, use these flags:

  • intent_tokenization_flag: If true the algorithm will split the intent labels into tokens and use a bag-of-words representations for them
  • intent_split_symbol: sets the delimiter string to split the intent labels. Default _

Here is a tutorial on how to use multiple intents in Rasa Core and NLU.

Here’s an example configuration:

language: "en"

pipeline:
- name: "CountVectorsFeaturizer"
- name: "EmbeddingIntentClassifier"
  intent_tokenization_flag: true
  intent_split_symbol: "+"

Understanding the Rasa NLU Pipeline

In Rasa NLU, incoming messages are processed by a sequence of components. These components are executed one after another in a so-called processing pipeline. There are components for entity extraction, for intent classification, pre-processing, and others. If you want to add your own component, for example to run a spell-check or to do sentiment analysis, check out Custom NLU Components.

Each component processes the input and creates an output. The output can be used by any component that comes after this component in the pipeline. There are components which only produce information that is used by other components in the pipeline and there are other components that produce Output attributes which will be returned after the processing has finished. For example, for the sentence "I am looking for Chinese food" the output is:

{
    "text": "I am looking for Chinese food",
    "entities": [
        {"start": 8, "end": 15, "value": "chinese", "entity": "cuisine", "extractor": "CRFEntityExtractor", "confidence": 0.864}
    ],
    "intent": {"confidence": 0.6485910906220309, "name": "restaurant_search"},
    "intent_ranking": [
        {"confidence": 0.6485910906220309, "name": "restaurant_search"},
        {"confidence": 0.1416153159565678, "name": "affirm"}
    ]
}

This is created as a combination of the results of the different components in the pre-configured pipeline pretrained_embeddings_spacy. For example, the entities attribute is created by the CRFEntityExtractor component.

Component Lifecycle

Every component can implement several methods from the Component base class; in a pipeline these different methods will be called in a specific order. Lets assume, we added the following pipeline to our config: "pipeline": ["Component A", "Component B", "Last Component"]. The image shows the call order during the training of this pipeline:

../../_images/component_lifecycle.png

Before the first component is created using the create function, a so called context is created (which is nothing more than a python dict). This context is used to pass information between the components. For example, one component can calculate feature vectors for the training data, store that within the context and another component can retrieve these feature vectors from the context and do intent classification.

Initially the context is filled with all configuration values, the arrows in the image show the call order and visualize the path of the passed context. After all components are trained and persisted, the final context dictionary is used to persist the model’s metadata.

The “entity” object explained

After parsing, the entity is returned as a dictionary. There are two fields that show information about how the pipeline impacted the entities returned: the extractor field of an entity tells you which entity extractor found this particular entity, and the processors field contains the name of components that altered this specific entity.

The use of synonyms can also cause the value field not match the text exactly. Instead it will return the trained synonym.

{
  "text": "show me chinese restaurants",
  "intent": "restaurant_search",
  "entities": [
    {
      "start": 8,
      "end": 15,
      "value": "chinese",
      "entity": "cuisine",
      "extractor": "CRFEntityExtractor",
      "confidence": 0.854,
      "processors": []
    }
  ]
}

Note

The confidence will be set by the CRF entity extractor (CRFEntityExtractor component). The duckling entity extractor will always return 1. The SpacyEntityExtractor extractor does not provide this information and returns null.

Pre-configured Pipelines

A template is just a shortcut for a full list of components. For example, these two configurations are equivalent:

language: "en"

pipeline: "pretrained_embeddings_spacy"
language: "en"

pipeline:
- name: "SpacyNLP"
- name: "SpacyTokenizer"
- name: "SpacyFeaturizer"
- name: "RegexFeaturizer"
- name: "CRFEntityExtractor"
- name: "EntitySynonymMapper"
- name: "SklearnIntentClassifier"

Below is a list of all the pre-configured pipeline templates with customization information.

supervised_embeddings

To train a Rasa model in your preferred language, define the supervised_embeddings pipeline as your pipeline in your config.yml or other configuration file:

language: "en"

pipeline: "supervised_embeddings"

The supervised_embeddings pipeline supports any language that can be tokenized. By default it uses whitespace for tokenization. You can customize the setup of this pipeline by adding or changing components. Here are the default components that make up the supervised_embeddings pipeline:

language: "en"

pipeline:
- name: "WhitespaceTokenizer"
- name: "RegexFeaturizer"
- name: "CRFEntityExtractor"
- name: "EntitySynonymMapper"
- name: "CountVectorsFeaturizer"
- name: "CountVectorsFeaturizer"
  analyzer: "char_wb"
  min_ngram: 1
  max_ngram: 4
- name: "EmbeddingIntentClassifier"

So for example, if your chosen language is not whitespace-tokenized (words are not separated by spaces), you can replace the WhitespaceTokenizer with your own tokenizer. We support a number of different tokenizers, or you can create your own.

The pipeline uses two instances of CountVectorsFeaturizer. The first one featurizes text based on words. The second one featurizes text based on character n-grams, preserving word boundaries. We empirically found the second featurizer to be more powerful, but we decided to keep the first featurizer as well to make featurization more robust.

pretrained_embeddings_spacy

To use the pretrained_embeddings_spacy template:

language: "en"

pipeline: "pretrained_embeddings_spacy"

See Pre-trained Word Vectors for more information about loading spacy language models. To use the components and configure them separately:

language: "en"

pipeline:
- name: "SpacyNLP"
- name: "SpacyTokenizer"
- name: "SpacyFeaturizer"
- name: "RegexFeaturizer"
- name: "CRFEntityExtractor"
- name: "EntitySynonymMapper"
- name: "SklearnIntentClassifier"

MITIE

To use the MITIE pipeline, you will have to train word vectors from a corpus. Instructions can be found here. This will give you the file path to pass to the model parameter.

language: "en"

pipeline:
- name: "MitieNLP"
  model: "data/total_word_feature_extractor.dat"
- name: "MitieTokenizer"
- name: "MitieEntityExtractor"
- name: "EntitySynonymMapper"
- name: "RegexFeaturizer"
- name: "MitieFeaturizer"
- name: "SklearnIntentClassifier"

Another version of this pipeline uses MITIE’s featurizer and also its multi-class classifier. Training can be quite slow, so this is not recommended for large datasets.

language: "en"

pipeline:
- name: "MitieNLP"
  model: "data/total_word_feature_extractor.dat"
- name: "MitieTokenizer"
- name: "MitieEntityExtractor"
- name: "EntitySynonymMapper"
- name: "RegexFeaturizer"
- name: "MitieIntentClassifier"

Custom pipelines

You don’t have to use a template, you can also run a fully custom pipeline by listing the names of the components you want to use:

pipeline:
- name: "SpacyNLP"
- name: "CRFEntityExtractor"
- name: "EntitySynonymMapper"

This creates a pipeline that only does entity recognition, but no intent classification. So Rasa NLU will not predict any intents. You can find the details of each component in Components.

If you want to use custom components in your pipeline, see Custom NLU Components.