Lower text and remove punctuation, articles and extra whitespace.

Lower text and remove punctuation, articles and extra whitespace.

Lower text and remove punctuation, articles and extra whitespace.

Write a InputFeature to the TFRecordWriter as a tf.train.Example.

Read examples from RACE json files.

Removes delimiters from a piece of text.

Reset the set of symbols allowed, but keep the same vocabulary size.

Sample an example.

Sample an available position that is next to the given position.

Save examples for the given q_ids as TFRecords.

Gets the most specific dimension size(s) of the given tensor. This is a wrapper around the two ways to get the shape of a tensor: (1) t.get_shape

A dataset with just two columns, x and y. Args: context: a NeuralQueryContext tuple_input: passed to util.tuple_dataset x_type: type of

Create mapping from order 3 tensor to vocabulary for copying mechanism. This will return a SparseTensor of size (batch_size, vocab_size, num_cells)

Group the yield into `QueryAdjustments`.

Creates a new loss. Args: name: String used as the most outer scope name of the loss graph.

Initializes a retriever. Args: batch_size: Batch size for the batched retrieval. log_every: Logging frequency.

Add constants for fully and under-specified city.

Where mask is true, add a large negative value.

Sample n from n with replacement.

Generate decoder output given encoder output and decoder input. Args: encoder_output: Tensor of shape (B, L, D). source_len: Tensor of shap

Returns new states if token_a and token_b are equal.

Converts a function that contains nested arguments.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Creates examples for the training and dev sets.

Decode function. Args: cell: rnn cell helper: a helper instance from tf.contrib.seq2seq. Returns: decoded outputs and leng

Decode function. Args: cell: rnn cell helper: a helper instance from tf.contrib.seq2seq. Returns: decoded outputs and leng

Decodes a record to a TensorFlow example.

Finds a column annotation in a given direction.

Follow a specific relation in the KG. Args: rel_name: a string naming a declared KG relation inverted: +1 to follow the relation norm

helper function to get the multi_reference for a file.

Generate synthetic inputs to test masking.

Generate output logits given decoder output. This effectively combines a Pointer Network (Vinyals et al., 2015) with a standard softmax output la

Returns a list of ActionTypeRange tuples.

Query the environment with a new query. Args: query: Query (reformulation) issued by the agent. original_query: The original query

Returns the frame (intent and slot labels) of the function call.

Get the free NT index.

Returns function name and argument sub-expressions.

Returns function name and argument sub-expressions.

Get a Counter with the ngrams of the given ID list. Args: ids: np.array or a list corresponding to a single sentence n: n-gram size Retu

Build set of special token ids to ignore when computing matching metrics. Args: vocab: An instance of text_utils.Vocab. Returns: remove:

Parses a SPARQL program into a prefix and conjuncts.

Retrieve the appropriate set of documents depending on settings. Args: claim_id: The claim to get documents for scrape_type: The scra

Get target embeddings from either output table or copied from input. Args: encoder_input: Tensor representing encoder output of shape (batch si

Get toy example for testing.

Returns a list of symbols in a vocabularly file.

Returns whether candidate UID is within the configured range.

Returns logits computed using inner product of qry and embeddings.

Returns true if the value is a tensor or can be converted to a tensor.

Parse the part relative to a Function in the SQL query.

Parse the part relative to an Operation in the SQL query.

Formats tensor (potentially ragged) so that it can be printed in Assert.

Replace corresponding non-terminal in node_2_string with node_1_string.