Determines the maximum edit distance (as an int value) that a suggestion for a spelling error may have. E.g. if set to one suggestions are limited to words within edit distance 1 to the original word.

The character encoding used by the model.

Location from which the model is read. The model file is a simple word-list with one word per line.

Use this language instead of the document language to resolve the model.

Load the chunk tag to UIMA type mapping from this location instead of locating the mapping automatically.

Use the String#intern() method on tags. This is usually a good idea to avoid spamming the heap with thousands of strings representing only a few different tags. Default: true

Use this language instead of the document language to resolve the model.

Load the model from this location instead of locating the model automatically.

Override the default variant used to locate the model.

Log the tag set(s) when a model is loaded. Default: false

Location of the mapping file for chunk tags to UIMA types.

Use this TreeTagger executable instead of trying to locate the executable automatically.

A sequence to flush the internal TreeTagger buffer and to force it to output the rest of the completed analysis. This is typically just a sequence of like 5-10 full stops (".") separated by new line characters. However, some models may require a different flush sequence, e.g. a short sentence in the respective language. For chunker models, mind that the sentence must also be POS tagged, e.g. Nous-PRO:PER\n....

Use the String#intern() method on tags. This is usually a good idea to avoid spaming the heap with thousands of strings representing only a few different tags. Default: true

Use this language instead of the document language to resolve the model.

Load the model from this location instead of locating the model automatically.

Override the default variant used to locate the model.

TT4J setting: Disable some sanity checks, e.g. whether tokens contain line breaks (which is not allowed). Turning this on will increase your performance, but the wrapper may throw exceptions if illegal data is provided.

Log the tag set(s) when a model is loaded. Default: false

DCoRef parameter: Maximum sentence distance between two mentions for resolution (-1: no constraint on the distance)

DCoRef parameter: Do post processing

DCoRef parameter: Scoring the output of the system

DCoRef parameter: Sieve passes - each class is defined in dcoref/sievepasses/.

DCoRef parameter: setting singleton predictor

Location from which the model is read.

Variant of a model the model. Used to address a specific model if here are multiple models for one language.

The maximum n-gram size that should be considered. Default is 3.

The minimum n-gram size that should be considered. Default is 1.

Use this language instead of the document language to resolve the model.

Load the model from this location instead of locating the model automatically.

Override the default variant used to locate the model.

Use this language instead of the document language to resolve the model.

Load the model from this location instead of locating the model automatically.

Override the default variant used to locate the model.

Use this language instead of the document language to resolve the model.

Load the model from this location instead of locating the model automatically.

Try reconstructing proper casing for lemmata. This is useful for German, but e.g. for English creates odd results.

Override the default variant used to locate the model.

Pass part-of-speech information on to Morpha. Since we currently do not know in which format the part-of-speech tags are expected by Morpha, we just pass on the actual pos tag value we get from the token. This may produce worse results than not passing on pos tags at all, so this is disabled by default.

Enable all traditional PTB3 token transforms (like -LRB-, -RRB-).

List of extra token texts (usually single character strings) that should be treated like opening quotes and escaped accordingly before being sent to the parser.

List of extra token texts (usually single character strings) that should be treated like closing quotes and escaped accordingly before being sent to the parser.

Load the part-of-speech tag to UIMA type mapping from this location instead of locating the mapping automatically.

Use this language instead of the document language to resolve the model.

The character encoding used by the model.

Load the model from this location instead of locating the model automatically.

Override the default variant used to locate the model.

Write the tag set(s) to the log when a model is loaded.

Use this language instead of the document language to resolve the model.

Specifies the model encoding.

Load the model from this location instead of locating the model automatically.

Override the default variant used to locate the model.

Write the tag set(s) to the log when a model is loaded.

Write lemma information. Default: true

Write part-of-speech information. Default: true

Location of the mapping file for named entity tags to UIMA types.

Use this language instead of the document language to resolve the model.

Location from which the model is read.

Variant of a model the model. Used to address a specific model if here are multiple models for one language.

Log the tag set(s) when a model is loaded.

Location of the mapping file for named entity tags to UIMA types.

Use this language instead of the document language to resolve the model.

Location from which the model is read.

Variant of a model the model. Used to address a specific model if here are multiple models for one language.

Log the tag set(s) when a model is loaded.

Enable all traditional PTB3 token transforms (like -LRB-, -RRB-).

List of extra token texts (usually single character strings) that should be treated like opening quotes and escaped accordingly before being sent to the parser.

List of extra token texts (usually single character strings) that should be treated like closing quotes and escaped accordingly before being sent to the parser.

Location of the mapping file for constituent tags to UIMA types.

Location of the mapping file for part-of-speech tags to UIMA types.

Set thresholds for accuracy.

Default: false (set thresholds for efficiency)

Output binarized trees.

Default: false

Use the String#intern() method on tags. This is usually a good idea to avoid spaming the heap with thousands of strings representing only a few different tags. Default: true

Retain predicted function labels. Model must have been trained with function labels.

Default: false

Use this language instead of the language set in the CAS to locate the model.

Load the model from this location instead of locating the model automatically.

Override the default variant used to locate the model.

Log the tag set(s) when a model is loaded. Default: false

Sets whether to use or not to use already existing POS tags from another annotator for the parsing process.

Default: false

Output inside scores (only for binarized viterbi trees).

Default: false

Output sub-categories (only for binarized Viterbi trees).

Default: false

Use variational rule score approximation instead of max-rule

Default: false

Compute Viterbi derivation instead of max-rule tree.

Default: false (max-rule)

Sets whether to create or not to create POS tags. The creation of constituent tags must be turned on for this to work.

Default: true

If this parameter is set to true, each sentence is annotated with a PennTree-Annotation, containing the whole parse tree in Penn Treebank style format.

Default: false

Use this language instead of the document language to resolve the model.

Location from which the model is read.

Variant of a model the model. Used to address a specific model if here are multiple models for one language.

Write the tag set(s) to the log when a model is loaded.

Process anyway, even if the model relies on features that are not supported by this component. Default: false

Use this language instead of the document language to resolve the model.

Load the model from this location instead of locating the model automatically.

Override the default variant used to locate the model.

Log the tag set(s) when a model is loaded. Default: false

Load the dependency to UIMA type mapping from this location instead of locating the mapping automatically.

Use this language instead of the document language to resolve the model.

Load the model from this location instead of locating the model automatically.

Override the default variant used to locate the model.

Log the tag set(s) when a model is loaded. Default: false

Load the dependency to UIMA type mapping from this location instead of locating the mapping automatically.

Use this language instead of the document language to resolve the model.

Load the model from this location instead of locating the model automatically.

Override the default variant used to locate the model.

Specifies the order/scope of features. 1 only has features over single edges and 2 has features over pairs of adjacent edges in the tree. The model must have been trained with the respective order set here.

Log the tag set(s) when a model is loaded. Default: false

Location of the mapping file for constituent tags to UIMA types.

Load the part-of-speech tag to UIMA type mapping from this location instead of locating the mapping automatically.

Use the String#intern() method on tags. This is usually a good idea to avoid spaming the heap with thousands of strings representing only a few different tags.

Default: true

Use this language instead of the document language to resolve the model.

Load the model from this location instead of locating the model automatically.

Override the default variant used to locate the model.

Log the tag set(s) when a model is loaded.

Default: false

Sets whether to create or not to create POS tags. The creation of constituent tags must be turned on for this to work.

Default: true

If this parameter is set to true, each sentence is annotated with a PennTree-Annotation, containing the whole parse tree in Penn Treebank style format.

Default: false

Location of the mapping file for constituent tags to UIMA types.

Location of the mapping file for part-of-speech tags to UIMA types.

This parameter can be used to override the standard behavior which uses the Sentence annotation as the basic unit for parsing.

If the parameter is set with the name of an annotation type x, the parser will no longer parse Sentence-annotations, but x-Annotations.

Default: null

Use this language instead of the document language to resolve the model and tag set mapping.

Controls when the factored parser considers a sentence to be too complex and falls back to the PCFG parser.

Default: 200000

Maximum number of tokens in a sentence. Longer sentences are not parsed. This is to avoid out of memory exceptions.

Default: 130

Sets the kind of dependencies being created.

Default: DependenciesMode#COLLAPSED TREE

Location from which the model is read.

Variant of a model the model. Used to address a specific model if here are multiple models for one language.

Write the tag set(s) to the log when a model is loaded.

Enable all traditional PTB3 token transforms (like -LRB-, -RRB-).

List of extra token texts (usually single character strings) that should be treated like opening quotes and escaped accordingly before being sent to the parser.

List of extra token texts (usually single character strings) that should be treated like closing quotes and escaped accordingly before being sent to the parser.

Sets whether to use or not to use already existing POS tags from another annotator for the parsing process.

Default: true

Sets whether to create or not to create constituent tags. This is required for POS-tagging and lemmatization.

Default: true

Sets whether to create or not to create dependency annotations.

Default: true

Sets whether to create or not to create POS tags. The creation of constituent tags must be turned on for this to work.

Default: false

If this parameter is set to true, each sentence is annotated with a PennTree-Annotation, containing the whole parse tree in Penn Treebank style format.

Default: false

Location of the mapping file for part-of-speech tags to UIMA types.

Use this language instead of the document language to resolve the model and tag set mapping.

Location from which the model is read.

Variant of a model the model. Used to address a specific model if here are multiple models for one language.

Load the part-of-speech tag to UIMA type mapping from this location instead of locating the mapping automatically.

Load the dictionary from this location instead of locating the dictionary automatically.

Override the default variant used to locate the dictionary.

Use the String#intern() method on tags. This is usually a good idea to avoid spaming the heap with thousands of strings representing only a few different tags.

Use this language instead of the document language to resolve the model.

Load the model from this location instead of locating the pos-tagging model automatically.

Override the default variant used to locate the pos-tagging model.

Log the tag set(s) when a model is loaded.

Load the part-of-speech tag to UIMA type mapping from this location instead of locating the mapping automatically.

Use the String#intern() method on tags. This is usually a good idea to avoid spaming the heap with thousands of strings representing only a few different tags. Default: true

Use this language instead of the document language to resolve the model.

Load the lexicon from this location instead of locating it automatically.

Override the default variant used to locate the model.

Log the tag set(s) when a model is loaded. Default: false

Load the ruleset from this location instead of locating it automatically.

Load the part-of-speech tag to UIMA type mapping from this location instead of locating the mapping automatically.

Use the String#intern() method on tags. This is usually a good idea to avoid spaming the heap with thousands of strings representing only a few different tags. Default: true

Use this language instead of the document language to resolve the model.

Load the model from this location instead of locating the model automatically.

Override the default variant used to locate the model.

Log the tag set(s) when a model is loaded. Default: false

Use this language instead of the document language to resolve the model.

Load the model from this location instead of locating the model automatically.

Override the default variant used to locate the model.

Load the part-of-speech tag to UIMA type mapping from this location instead of locating the mapping automatically.

Use this language instead of the document language to resolve the model.

Load the model from this location instead of locating the model automatically.

Override the default variant used to locate the model.

Log the tag set(s) when a model is loaded. Default: false

The language.

Strict zoning causes the segmentation to be applied only within the boundaries of a zone annotation. This works only if a single zone type is specified (the zone annotations should NOT overlap) or if no zone type is specified - in which case the whole document is taken as a zone. If strict zoning is turned off, multiple zone types can be specified. A list of all zone boundaries (start and end) is created and segmentation happens between them.

Create Sentence annotations.

Create Token annotations.

A list of type names used for zoning.

Load the part-of-speech tag to UIMA type mapping from this location instead of locating the mapping automatically.

Use the String#intern() method on tags. This is usually a good idea to avoid spaming the heap with thousands of strings representing only a few different tags. Default: true

Use this language instead of the document language to resolve the model.

Load the model from this location instead of locating the model automatically.

Override the default variant used to locate the model.

Log the tag set(s) when a model is loaded. Default: false

Location of the mapping file for part-of-speech tags to UIMA types.

Use the String#intern() method on tags. This is usually a good idea to avoid spaming the heap with thousands of strings representing only a few different tags. Default: false

Use this language instead of the document language to resolve the model and tag set mapping.

Sentences with more tokens than the specified max amount will be ignored if this parameter is set to a value larger than zero. The default value zero will allow all sentences to be POS tagged.

Location from which the model is read.

Variant of a model the model. Used to address a specific model if here are multiple models for one language.

Log the tag set(s) when a model is loaded. Default: false

Enable all traditional PTB3 token transforms (like -LRB-, -RRB-).

List of extra token texts (usually single character strings) that should be treated like opening quotes and escaped accordingly before being sent to the parser.

List of extra token texts (usually single character strings) that should be treated like closing quotes and escaped accordingly before being sent to the parser.

Load the part-of-speech tag to UIMA type mapping from this location instead of locating the mapping automatically.

Use this TreeTagger executable instead of trying to locate the executable automatically.

Use the String#intern() method on tags. This is usually a good idea to avoid spaming the heap with thousands of strings representing only a few different tags. Default: true

Use this language instead of the document language to resolve the model.

The character encoding used by the model.

Load the model from this location instead of locating the model automatically.

Override the default variant used to locate the model.

TT4J setting: Disable some sanity checks, e.g. whether tokens contain line breaks (which is not allowed). Turning this on will increase your performance, but the wrapper may throw exceptions if illegal data is provided.

Log the tag set(s) when a model is loaded. Default: false

Write lemma information. Default: true

Write part-of-speech information. Default: true

A set of annotation types that should be filtered.

Any annotation in filterAnnotations shorter than this value will be removed.

Any annotation in filterTypes shorter than this value will be removed.

The language.

Per default the Java BreakIterator does not split off contractions like John's into two tokens. When this parameter is enabled, a non-default token split is generated when an apostrophe (') is encountered.

Strict zoning causes the segmentation to be applied only within the boundaries of a zone annotation. This works only if a single zone type is specified (the zone annotations should NOT overlap) or if no zone type is specified - in which case the whole document is taken as a zone. If strict zoning is turned off, multiple zone types can be specified. A list of all zone boundaries (start and end) is created and segmentation happens between them.

Create Sentence annotations.

Create Token annotations.

A list of type names used for zoning.

Wether to remove the original token. Default: true

Use this language instead of the document language to resolve the model.

Load the model from this location instead of locating the model automatically.

Override the default variant used to locate the model.

Strict zoning causes the segmentation to be applied only within the boundaries of a zone annotation. This works only if a single zone type is specified (the zone annotations should NOT overlap) or if no zone type is specified - in which case the whole document is taken as a zone. If strict zoning is turned off, multiple zone types can be specified. A list of all zone boundaries (start and end) is created and segmentation happens between them.

Create Sentence annotations.

Create Token annotations.

A list of type names used for zoning.

The language.

Strict zoning causes the segmentation to be applied only within the boundaries of a zone annotation. This works only if a single zone type is specified (the zone annotations should NOT overlap) or if no zone type is specified - in which case the whole document is taken as a zone. If strict zoning is turned off, multiple zone types can be specified. A list of all zone boundaries (start and end) is created and segmentation happens between them.

Create Paragraph annotations.

Create Sentence annotations.

Create Token annotations.

A list of type names used for zoning.

The language.

Strict zoning causes the segmentation to be applied only within the boundaries of a zone annotation. This works only if a single zone type is specified (the zone annotations should NOT overlap) or if no zone type is specified - in which case the whole document is taken as a zone. If strict zoning is turned off, multiple zone types can be specified. A list of all zone boundaries (start and end) is created and segmentation happens between them.

Create Sentence annotations.

Create Token annotations.

A list of type names used for zoning.

The language.

Strict zoning causes the segmentation to be applied only within the boundaries of a zone annotation. This works only if a single zone type is specified (the zone annotations should NOT overlap) or if no zone type is specified - in which case the whole document is taken as a zone. If strict zoning is turned off, multiple zone types can be specified. A list of all zone boundaries (start and end) is created and segmentation happens between them.

Create Sentence annotations.

Create Token annotations.

A list of type names used for zoning.

Use this language instead of the document language to resolve the model.

Override the default variant used to locate the model.

Load the segmentation model from this location instead of locating the model automatically.

Strict zoning causes the segmentation to be applied only within the boundaries of a zone annotation. This works only if a single zone type is specified (the zone annotations should NOT overlap) or if no zone type is specified - in which case the whole document is taken as a zone. If strict zoning is turned off, multiple zone types can be specified. A list of all zone boundaries (start and end) is created and segmentation happens between them.

Load the tokenization model from this location instead of locating the model automatically.

Create Sentence annotations.

Create Token annotations.

A list of type names used for zoning.

A regular expression used to detect paragraph splits. Default: #DOUBLE_LINE_BREAKS_PATTERN (split on two consecutive line breaks)

Wether to remove the original token. Default: true

A list of regular expressions, prefixed with #INCLUDE_PREFIX or #EXCLUDE_PREFIX. If neither of the prefixes is used, #EXCLUDE_PREFIX is assumed.

The language.

Define the sentence boundary. Default: \n (assume one sentence per line).

Strict zoning causes the segmentation to be applied only within the boundaries of a zone annotation. This works only if a single zone type is specified (the zone annotations should NOT overlap) or if no zone type is specified - in which case the whole document is taken as a zone. If strict zoning is turned off, multiple zone types can be specified. A list of all zone boundaries (start and end) is created and segmentation happens between them.

Defines the pattern that is used as token end boundary. Default: [\s\n]+ (matching whitespace and linebreaks.

When setting custom patterns, take into account that the final token is often terminated by a linebreak rather than the boundary character. Therefore, the newline typically has to be added to the group of matching characters, e.g. "tokenized-text" is correctly tokenized with the pattern [-\n].

Create Sentence annotations.

Create Token annotations.

A list of type names used for zoning.

Whether to generate empty sentences.

This is a Set of String that are matched with .equals() which are allowed to be tacked onto the end of a sentence after a sentence boundary token, for example ")".

The set of regex for sentence boundary tokens that should be discarded.

The set of boundary tokens. If null, use default.

Whether to treat all input as one sentence.

The language.

Strategy for treating newlines as paragraph breaks.

A regular expression for element names containing a sentence region. Only tokens in such elements will be included in sentences. The start and end tags themselves are not included in the sentence.

Strict zoning causes the segmentation to be applied only within the boundaries of a zone annotation. This works only if a single zone type is specified (the zone annotations should NOT overlap) or if no zone type is specified - in which case the whole document is taken as a zone. If strict zoning is turned off, multiple zone types can be specified. A list of all zone boundaries (start and end) is created and segmentation happens between them.

The set of regex for sentence boundary tokens that should be discarded.

Create Sentence annotations.

Create Token annotations.

These are elements like "p" or "sent", which will be wrapped into regex for approximate XML matching. They will be deleted in the output, and will always trigger a sentence boundary.

A list of type names used for zoning.

Override the tagset mapping.

Annotation type for which tokens should be merged.

A constraint on the annotations that should be considered in form of a JXPath statement. Example: set #PARAM_ANNOTATION_TYPE to a NamedEntity type and set the #PARAM_CONSTRAINT to ".[value = 'LOCATION']" to merge only tokens that are part of a location named entity.

Use this language instead of the document language to resolve the model and tag set mapping.

Configure what should happen to the lemma of the merged tokens. It is possible to JOIN the lemmata to a single lemma (space separated), to REMOVE the lemma or LEAVE the lemma of the first token as-is.

Set a new POS tag for the new merged token. This is the mapped type. If this is specified, tag set mapping will not be performed. This parameter has no effect unless PARAM_POS_VALUE is also set.

Set a new POS value for the new merged token. This is the actual tag set value and is subject to tagset mapping. For example when merging tokens for named entities, the new POS value may be set to "NNP" (English/Penn Treebank Tagset).

List of prefixes to remove.

List of suffixes to remove.

The language.

Strict zoning causes the segmentation to be applied only within the boundaries of a zone annotation. This works only if a single zone type is specified (the zone annotations should NOT overlap) or if no zone type is specified - in which case the whole document is taken as a zone. If strict zoning is turned off, multiple zone types can be specified. A list of all zone boundaries (start and end) is created and segmentation happens between them.

Create Sentence annotations.

Create Token annotations.

A list of type names used for zoning.

Normally the arguments point only to the head words of arguments in the dependency tree. With this option enabled, they are expanded to the text covered by the minimal and maximal token offsets of all descendants (or self) of the head word.

Warning: this parameter should be used with caution! For one, if the descentants of a head word cover a non-continuous region of the text, this information is lost. The arguments will appear to be spanning a continuous region. For another, the arguments may overlap with each other. E.g. if a sentence contains a relative clause with a verb, the subject of the main clause may be recognized as a dependent of the verb and may cause the whole main clause to be recorded in the argument.

Use this language instead of the document language to resolve the model.

Variant of a model the model. Used to address a specific model if here are multiple models for one language.

Location from which the predicate identifier model is read.

Write the tag set(s) to the log when a model is loaded.

Location from which the roleset classification model is read.

Location from which the semantic role labeling model is read.

Use this language instead of the document language to resolve the model.

Load the model from this location instead of locating the model automatically.

Override the default variant used to locate the model.

Specifies the operator for a filtering condition.

It is only used if PARAM_FILTER_FEATUREPATH is set.

Specifies the value for a filtering condition.

It is only used if PARAM_FILTER_FEATUREPATH is set.

Specifies a feature path that is used in the filter. If this is set, you also have to specify PARAM_FILTER_CONDITION_OPERATOR and PARAM_FILTER_CONDITION_VALUE.

Use this language instead of the document language to resolve the model.

Per default the stemmer runs in case-sensitive mode. If this parameter is enabled, tokens are lower-cased before being passed to the stemmer.

Specify a path that is used for annotation. Format is de.type.name/feature/path. All type objects will be annotated with a IndexTermAnnotation. The value of the IndexTerm is specified by the feature path.

The sum of alphas over all topics. Default: 1.0.

Another recommended value is 50 / T (number of topics).

Beta for a single dimension of the Dirichlet prior. Default: 0.01.

The number of iterations before hyperparameter optimization begins. Default: 100

The interval in which to display the estimated topics. Default: 50.

The number of top words to display during estimation. Default: 7.

Ignore tokens (or lemmas, respectively) that are shorter than the given value. Default: 3.

If specific, the text contained in the given segmentation type annotations are fed as separate units to the topic model estimator e.g. de.tudarmstadt.ukp.dkpro.core.api.segmentation.type.sentence. Text that is not within such annotations is ignored.

By default, the full document text is used as a document.

The number of iterations during model estimation. Default: 1000.

The number of threads to use during model estimation. Default: 1.

The number of topics to estimate for the topic model.

Interval for optimizing Dirichlet hyperparameters. Default: 50

Set random seed. If set to -1 (default), uses random generator.

Define how often to save a serialized model during estimation. Default: 0 (only save when estimation is done).

The target model file location.

The annotation type to use for the topic model. Default: de.tudarmstadt.ukp.dkpro.core.api.segmentation.type.Token.

If set, uses lemmas instead of original text as features.

Use a symmatric alpha value during model estimation? Default: false.

The number of iterations before hyperparameter optimization begins. Default: 1

Maximum number of topics to assign. If not set (or <= 0), the number of topics in the model divided by 10 is set.

Ignore tokens (or lemmas, respectively) that are shorter than the given value. Default: 3.

Minimum topic proportion for the document-topic assignment.

The number of iterations during inference. Default: 10.

The annotation type to use as tokens. Default: Token