Escapes special markdown characters in a string.

Compile data and return an iterator over samples.

Reads a tab separated value file.

Replace nt in rhs_a with rhs_b, re-indexing non-terminals if needed.

Returns a parsed and preprocessed `tf.data.Dataset`.

Compute loss between logits and labels. Args: logits: batch_size, num_candidate labels: batch_size, num_candidate Returns:

Build the full text transformer.

Converts a list of string tokens to a list of vocab identifiers.

convert data format for queries and filter queries without answers.

Decodes the given serialized TF-example.

Disambiguates redirects.

Alias for weighted_by which makes more sense for 'hard' relationships. Args: rel_name: a string naming a declared KG relation related

Gets the list of labels for this data set.

Get the proto for the wikipedia page by url. Args: wikipedia_url: Url to lookup Returns: Returns proto of page if it exists, oth

Return all the defined symbols in a list.

Alias for weighted_by_sum which makes more sense for 'hard' sets. Args: other: A NeuralQueryExpression Returns: A NeuralQueryExp

Lower text and remove punctuation, articles and extra whitespace.

Pads a 1-D Tensor with zeros to the target length.

Load queries per second data from output produced by scaling_eval.py. Args: filename: string name of file squarex: if true, square the x co

Simple method to query the KG, mainly for debugging. Finds things related to the named entity_name by the named relation. When as_object is

Removes double square bracket elements from the text. This gets rid of things that otherwise look like entities, e.g.: remove_element(text, 'Fi

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

Check if two types can be used with an operation. Args: type_name1: A string name for a type type_name2: A string name for a type

Convert (year, month, day) to integer representation. Args: date: Tuple of (year, month, day). Returns: an int of year * 1e4 + month * 1

Get absolute path.

Gets the value of the specified function argument.

Gets the number of tokens in a string.

Return string for RHS of node with non-terminals anonymized.

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

Returns list of lines from the text file. Performs lowercasing and white-space tokenization on each line before returning. Slightly different tha

Returns list of lines from the text file. Performs lowercasing and white-space tokenization on each line before returning. Args: text_file

Compute application scores.

Loads a data file into a list of `InputBatch`s.

Loads a data file into a list of `InputBatch`s.

Filter distinct objects with filter function.

Gathers the beam slices indexed by beam_indices into new beam array. Args: nested: pytree of arrays or scalars (the latter ignored). beam_i

Return aliases for given wikidata id.

Returns a text representation of the candidate at the given index.

Get entries based on full key.

Return value that will be used to initialize a relation matrix. Args: rel_name: string name of relation Returns: A numpy matrix

Return a sketch for set xs (all zeros if xs not specified). This function takes a list of elements xs, take their hash values, and set 1.0 to

Get the Tensor that represents a relation. Args: rel_name: string naming a declared relation Returns: tf.SparseTensor Raise

Creates an `input_fn` closure to be passed to TPUEstimator.

Test if the relation has been previously declared. Args: rel_name: a string naming a declared relation. Returns: a boolean

Embed query into start and end vectors for dense retrieval for a hop.

Lower text and remove punctuation, articles and extra whitespace.

Creates a sequence of output / decoder actions from a raw SQL query. Args: sql: The SQL query to convert. example: The NLToSQLExample objec

Compute evaluation metrics. Args: logits: batch_size, num_candidate labels: batch_size, num_candidate prefix: prefix for evalua

Provide value that will be used to initialize a relation matrix. Args: rel_name: string name of relation m: a matrix that can used as