Skip to content

token_indexer

allennlp.data.token_indexers.token_indexer

[SOURCE]


IndexedTokenList

IndexedTokenList = Dict[str, List[Any]]

TokenIndexer

class TokenIndexer(Registrable):
 | def __init__(self, token_min_padding_length: int = 0) -> None

A TokenIndexer determines how string tokens get represented as arrays of indices in a model. This class both converts strings into numerical values, with the help of a Vocabulary, and it produces actual arrays.

Tokens can be represented as single IDs (e.g., the word "cat" gets represented by the number 34), or as lists of character IDs (e.g., "cat" gets represented by the numbers [23, 10, 18]), or in some other way that you can come up with (e.g., if you have some structured input you want to represent in a special way in your data arrays, you can do that here).

Parameters

  • token_min_padding_length : int, optional (default = 0)
    The minimum padding length required for the TokenIndexer. For example, the minimum padding length of SingleIdTokenIndexer is the largest size of filter when using CnnEncoder. Note that if you set this for one TokenIndexer, you likely have to set it for all TokenIndexer for the same field, otherwise you'll get mismatched tensor sizes.

default_implementation

class TokenIndexer(Registrable):
 | ...
 | default_implementation = "single_id"

has_warned_for_as_padded_tensor

class TokenIndexer(Registrable):
 | ...
 | has_warned_for_as_padded_tensor = False

count_vocab_items

class TokenIndexer(Registrable):
 | ...
 | def count_vocab_items(
 |     self,
 |     token: Token,
 |     counter: Dict[str, Dict[str, int]]
 | )

The Vocabulary needs to assign indices to whatever strings we see in the training data (possibly doing some frequency filtering and using an OOV, or out of vocabulary, token). This method takes a token and a dictionary of counts and increments counts for whatever vocabulary items are present in the token. If this is a single token ID representation, the vocabulary item is likely the token itself. If this is a token characters representation, the vocabulary items are all of the characters in the token.

tokens_to_indices

class TokenIndexer(Registrable):
 | ...
 | def tokens_to_indices(
 |     self,
 |     tokens: List[Token],
 |     vocabulary: Vocabulary
 | ) -> IndexedTokenList

Takes a list of tokens and converts them to an IndexedTokenList. This could be just an ID for each token from the vocabulary. Or it could split each token into characters and return one ID per character. Or (for instance, in the case of byte-pair encoding) there might not be a clean mapping from individual tokens to indices, and the IndexedTokenList could be a complex data structure.

indices_to_tokens

class TokenIndexer(Registrable):
 | ...
 | def indices_to_tokens(
 |     self,
 |     indexed_tokens: IndexedTokenList,
 |     vocabulary: Vocabulary
 | ) -> List[Token]

Inverse operations of tokens_to_indices. Takes an IndexedTokenList and converts it back into a list of tokens.

get_empty_token_list

class TokenIndexer(Registrable):
 | ...
 | def get_empty_token_list(self) -> IndexedTokenList

Returns an already indexed version of an empty token list. This is typically just an empty list for whatever keys are used in the indexer.

get_padding_lengths

class TokenIndexer(Registrable):
 | ...
 | def get_padding_lengths(
 |     self,
 |     indexed_tokens: IndexedTokenList
 | ) -> Dict[str, int]

This method returns a padding dictionary for the given indexed_tokens specifying all lengths that need padding. If all you have is a list of single ID tokens, this is just the length of the list, and that's what the default implementation will give you. If you have something more complicated, like a list of character ids for token, you'll need to override this.

as_padded_tensor_dict

class TokenIndexer(Registrable):
 | ...
 | def as_padded_tensor_dict(
 |     self,
 |     tokens: IndexedTokenList,
 |     padding_lengths: Dict[str, int]
 | ) -> Dict[str, torch.Tensor]

This method pads a list of tokens given the input padding lengths (which could actually truncate things, depending on settings) and returns that padded list of input tokens as a Dict[str, torch.Tensor]. This is a dictionary because there should be one key per argument that the TokenEmbedder corresponding to this class expects in its forward() method (where the argument name in the TokenEmbedder needs to make the key in this dictionary).

The base class implements the case when all you want to do is create a padded LongTensor for every list in the tokens dictionary. If your TokenIndexer needs more complex logic than that, you need to override this method.