API Documentation

Getting Started

To parse text to UCCA passages, download a model file from the latest release, extract it, and use the following code template:

from tupa.parse import Parser
from ucca.convert import from_text
parser = Parser("models/ucca_bilstm")
for passage in parser.parse(from_text(...))):
    ...

Each passage instantiates the ucca.core.Passage class.

tupa.parse Module

Functions

average_f1(scores[, eval_type])
filter_passages_for_bert(passages, args)
from_text_format(*args, **kwargs)
generate_and_len(it)
get_eval_type(scores)
get_output_converter(out_format[, default])
glob(pathname, *[, recursive])	Return a list of paths matching a pathname pattern.
main()
main_generator()
percents_str(part, total[, infix, fraction])
print_scores(scores, filename[, prefix, …])
read_passages(args, files)
set_traceback_listener([sig])
single_to_iter(it)
to_lower_case(passages)
train_test(train_passages, dev_passages, …)	Train and test parser on given passage :param train_passages: passage to train on :param dev_passages: passages to evaluate on every iteration :param test_passages: passages to test on after training :param args: extra argument :param model_suffix: string to append to model filename before file extension :return: generator of Scores objects: dev scores for each training iteration (if given dev), and finally test scores

Classes

AbstractParser(config, models[, training, …])
BatchParser(*args, **kwargs)	Parser for a single training iteration or single pass over dev/test passages
ClassifierProperty	An enumeration.
Enum	Generic enumeration.
Iterations(args)
Model(filename[, config])
Oracle(passage)	Oracle to produce gold transition parses given UCCA passages To be used for creating training data for a transition-based UCCA parser :param passage gold passage to get the correct edges from
ParseMode	An enumeration.
Parser([model_files, config, beam])	Main class to implement transition-based UCCA parser
ParserException
PassageParser(passage, *args, **kwargs)	Parser for a single passage, has a state and optionally an oracle
State(passage)	The parser’s state, responsible for applying actions and creating the final Passage :param passage: a Passage object to get the tokens from, and everything else if training
defaultdict	defaultdict(default_factory[, …]) –> dict with default factory
partial	partial(func, *args, **keywords) - new function with partial application of the given arguments and keywords.

Class Inheritance Diagram

tupa.action Module

Classes

Action(action_type[, tag, orig_edge, …])
Actions([actions, size])
Labels(size)

Class Inheritance Diagram

tupa.config Module

Functions

add_param_arguments([ap, arg_default])
deepcopy(x[, memo, _nil])	Deep copy operation on arbitrary Python objects.
load_enum(filename)

Classes

CategoricalParameter(values, string)
Hyperparams(parent[, shared])
HyperparamsInitializer([name])
Iterations(args)

Class Inheritance Diagram

tupa.labels Module

Classes

Labels(size)

Class Inheritance Diagram

tupa.model Module

Functions

load_json(filename)	Load dictionary from JSON file :param filename: file to read from
remove_backup(*filenames)
save_json(filename, d)	Save dictionary to JSON file :param filename: file to write to :param d: dictionary to save

Classes

Actions([actions, size])
AutoIncrementDict([size, keys, d, unknown])	DefaultOrderedDict that returns an auto-incrementing index for new keys
Classifier(config, labels[, input_params])	Interface for classifier used by the parser.
ClassifierProperty	An enumeration.
Enum	Generic enumeration.
FeatureParameters(suffix, dim, size[, …])
Model(filename[, config])
OrderedDict	Dictionary that remembers insertion order
ParameterDefinition(args, name, attr_to_arg)
UnknownDict([d])	DefaultOrderedDict that has a single default value for missing keys

Class Inheritance Diagram

tupa.model_util Module

Functions

glob(pathname, *[, recursive])	Return a list of paths matching a pathname pattern.
jsonify(o)
load_dict(filename)	Load dictionary from Pickle file :param filename: file to read from
load_enum(filename)
load_json(filename)	Load dictionary from JSON file :param filename: file to read from
remove_backup(*filenames)
remove_existing(*filenames)
save_dict(filename, d)	Save dictionary to Pickle file :param filename: file to write to :param d: dictionary to save
save_json(filename, d)	Save dictionary to JSON file :param filename: file to write to :param d: dictionary to save

Classes

AutoIncrementDict([size, keys, d, unknown])	DefaultOrderedDict that returns an auto-incrementing index for new keys
Counter(**kwds)	Dict subclass for counting hashable items.
DefaultOrderedDict([default_factory, size])
DropoutDict([d, dropout, size, keys, min_count])	UnknownDict that sometimes returns the unknown value even for existing keys
IdentityVocab()
KeyBasedDefaultDict
Labels(size)
Lexeme(index, text)
OrderedDict	Dictionary that remembers insertion order
Strings(vocab)
UnknownDict([d])	DefaultOrderedDict that has a single default value for missing keys
Vocab(tuples)
defaultdict	defaultdict(default_factory[, …]) –> dict with default factory
itemgetter	itemgetter(item, …) –> itemgetter object

Class Inheritance Diagram

tupa.oracle Module

Functions

is_implicit_node(node)
is_remote_edge(edge)
is_terminal_edge(edge)

Classes

Actions([actions, size])
InvalidActionError(*args[, is_type])
Oracle(passage)	Oracle to produce gold transition parses given UCCA passages To be used for creating training data for a transition-based UCCA parser :param passage gold passage to get the correct edges from

Class Inheritance Diagram