| { |
| "paper_id": "2020", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T03:10:53.781406Z" |
| }, |
| "title": "A Universal Dependencies Conversion Pipeline for a Penn-format Constituency Treebank", |
| "authors": [ |
| { |
| "first": "\u00de\u00f3runn", |
| "middle": [], |
| "last": "Arnard\u00f3ttir", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "University of Iceland", |
| "location": {} |
| }, |
| "email": "" |
| }, |
| { |
| "first": "Hinrik", |
| "middle": [], |
| "last": "Hafsteinsson", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "University of Iceland", |
| "location": {} |
| }, |
| "email": "" |
| }, |
| { |
| "first": "Einar", |
| "middle": [ |
| "Freyr" |
| ], |
| "last": "Sigur\u00f0sson", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "The \u00c1rni Magn\u00fasson Institute for Icelandic Studies Krist\u00edn Bjarnad\u00f3ttir The \u00c1rni Magn\u00fasson Institute for Icelandic Studies", |
| "institution": "", |
| "location": {} |
| }, |
| "email": "" |
| }, |
| { |
| "first": "Anton", |
| "middle": [ |
| "Karl" |
| ], |
| "last": "Ingason", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "University of Iceland", |
| "location": {} |
| }, |
| "email": "" |
| }, |
| { |
| "first": "Hildur", |
| "middle": [], |
| "last": "J\u00f3nsd\u00f3ttir", |
| "suffix": "", |
| "affiliation": {}, |
| "email": "" |
| }, |
| { |
| "first": "Stein\u00fe\u00f3r", |
| "middle": [], |
| "last": "Steingr\u00edmsson", |
| "suffix": "", |
| "affiliation": {}, |
| "email": "" |
| } |
| ], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "The topic of this paper is a rule-based pipeline for converting constituency treebanks based on the Penn Treebank format to Universal Dependencies (UD). We describe an Icelandic constituency treebank, its annotation scheme and the UD scheme. The conversion is discussed, the methods used to deliver a fully automated UD corpus and complications involved. To show its applicability to corpora in different languages, we extend the pipeline and convert a Faroese constituency treebank to a UD corpus. The result is an open-source conversion tool, published under an Apache 2.0 license, applicable to a Penn-style treebank for conversion to a UD corpus, along with the two new UD corpora.", |
| "pdf_parse": { |
| "paper_id": "2020", |
| "_pdf_hash": "", |
| "abstract": [ |
| { |
| "text": "The topic of this paper is a rule-based pipeline for converting constituency treebanks based on the Penn Treebank format to Universal Dependencies (UD). We describe an Icelandic constituency treebank, its annotation scheme and the UD scheme. The conversion is discussed, the methods used to deliver a fully automated UD corpus and complications involved. To show its applicability to corpora in different languages, we extend the pipeline and convert a Faroese constituency treebank to a UD corpus. The result is an open-source conversion tool, published under an Apache 2.0 license, applicable to a Penn-style treebank for conversion to a UD corpus, along with the two new UD corpora.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Abstract", |
| "sec_num": null |
| } |
| ], |
| "body_text": [ |
| { |
| "text": "The Universal Dependencies (UD) project (Nivre et al., 2016) in version 2.6 consists of 163 treebanks in 92 languages and its standardized annotation scheme makes it an appealing option when creating a treebank. Different methods are available to create a UD treebank, ranging from manual parsing to manual correction of automatic parsing. A UD treebank can also be automatically converted from an existing treebank, which uses a different annotation scheme, as in the present work. We describe a conversion pipeline for a constituency treebank, the Icelandic Parsed Historical Corpus (IcePaHC; Wallenberg et al., 2011; . We also apply the pipeline to a smaller corpus, the Faroese Parsed Historical Corpus (FarPaHC; Ingason et al., 2014) , which contains texts in another language but uses the same annotation scheme. 1 With minimal modifications to the pipeline, the conversion is unaffected by linguistic differences in the corpora, which underlines the possibility of using it for other corpora, which use the same scheme.", |
| "cite_spans": [ |
| { |
| "start": 40, |
| "end": 60, |
| "text": "(Nivre et al., 2016)", |
| "ref_id": "BIBREF15" |
| }, |
| { |
| "start": 595, |
| "end": 619, |
| "text": "Wallenberg et al., 2011;", |
| "ref_id": "BIBREF26" |
| }, |
| { |
| "start": 707, |
| "end": 716, |
| "text": "(FarPaHC;", |
| "ref_id": null |
| }, |
| { |
| "start": 717, |
| "end": 738, |
| "text": "Ingason et al., 2014)", |
| "ref_id": "BIBREF7" |
| }, |
| { |
| "start": 819, |
| "end": 820, |
| "text": "1", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "The conversion uses NLTK (Bird et al., 2009) as a base and it is specific to the annotation scheme of IcePaHC, which itself is based on the Penn Parsed Corpora of Historical English (PPCHE; Kroch and Taylor, 2000; Kroch et al., 2004) . Some conversion tools are already available for converting a constituency treebank to the UD scheme, e.g. the PyStanfordDependencies tool 2 and the LTH Constituentto-Dependency Conversion Tool for Penn-style Treebanks (Johansson and Nugues, 2007) . 3 Both tools convert a treebank in the Penn Treebank format to a dependency format. The annotation scheme of the PPCHE is based on the Penn Treebank scheme but the two are not similar enough to be used unaltered with the pre-existing software, so the conversion between the two schemes was deemed to be impractical. Therefore, our conversion tool takes as input an unaltered IcePaHC file and delivers its UD-annotated counterpart. Both IcePaHC and FarPaHC are open-source, published under a CC BY 4.0 license, and we maintain that policy, publishing the conversion tool under an Apache 2.0 license and the resulting UD corpora under a CC BY-SA 4.0 license. The tool is reusable for corpora annotated in the same manner as the constituency treebanks and the source code is available on GitHub. 4 The resulting corpora will be included in the next version of UD, version 2.7, to be released on November 15, 2020.", |
| "cite_spans": [ |
| { |
| "start": 25, |
| "end": 44, |
| "text": "(Bird et al., 2009)", |
| "ref_id": "BIBREF2" |
| }, |
| { |
| "start": 190, |
| "end": 213, |
| "text": "Kroch and Taylor, 2000;", |
| "ref_id": "BIBREF11" |
| }, |
| { |
| "start": 214, |
| "end": 233, |
| "text": "Kroch et al., 2004)", |
| "ref_id": "BIBREF12" |
| }, |
| { |
| "start": 454, |
| "end": 482, |
| "text": "(Johansson and Nugues, 2007)", |
| "ref_id": "BIBREF8" |
| }, |
| { |
| "start": 485, |
| "end": 486, |
| "text": "3", |
| "ref_id": null |
| }, |
| { |
| "start": 1278, |
| "end": 1279, |
| "text": "4", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "The paper is structured as follows. Section 2 describes IcePaHC and its annotation scheme and Section 3 compares PPCHE and UD. Section 4 describes the conversion itself, and its stages, i.e. how the files are prepared for the converter (4.1), how the PoS tags and morphological features are mapped (4.2), and how the relational information of heads and dependency are extracted from the IcePaHC format (4.3). Section 5 discusses the FarPaHC conversion, Section 6 considers further use of the converter and the UD treebanks and Section 7 concludes.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "IcePaHC is a one-million-word, diachronic corpus, which includes texts from the 12 th to 21 st centuries . The trees in the corpus have been manually corrected according to the PPCHE annotation scheme, which uses labeled bracketing in the same way as the Penn Treebank. Some minor adjustments were made to adapt the annotation scheme to Icelandic grammar. The PoS tagset used is based on the one particular to the PPCHE, with minor changes to adapt it to Icelandic grammar. The manual IcePaHC annotation process is built on a number of automatic pre-processing steps involving available Natural Language Processing tools for Icelandic, including PoS tagging and shallow parsing in IceNLP (Loftsson and R\u00f6gnvaldsson, 2007) , and the lemmatizer Lemmald (Ingason et al., 2008) .", |
| "cite_spans": [ |
| { |
| "start": 688, |
| "end": 721, |
| "text": "(Loftsson and R\u00f6gnvaldsson, 2007)", |
| "ref_id": "BIBREF13" |
| }, |
| { |
| "start": 743, |
| "end": 773, |
| "text": "Lemmald (Ingason et al., 2008)", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Icelandic Parsed Historical Corpus", |
| "sec_num": "2" |
| }, |
| { |
| "text": "The IcePaHC scheme splits sentences into matrix clauses and marks their phrases. The phrases and their tokens are marked according to the tagset, which consists of 43 tags and their function tags. 5 Moreover, every token's lemma is displayed along with traces and empty phrases. Figure 1 is an original example from IcePaHC where we show a matrix clause (IP-MAT) with a subject (NP-SBJ), a main verb (VB) in the past tense (D), indicative mood (I), i.e. VBDI, an indirect object (NP-OB2), a direct object (NP-OB1) and a prepositional phrase (PP); the sentence is glossed and translated in (1). This rich annotation scheme makes the conversion to a dependency-based scheme possible. The practicality of using a historical (diachronic) corpus for conversion to a descriptive dependency corpus for Icelandic hinges on the fact that Icelandic has changed much less than many other European languages over the last thousand years, syntax included (R\u00f6gnvaldsson and Helgad\u00f3ttir, 2011) . This fact by itself supports the use of IcePaHC in the conversion. However, even though Icelandic has changed less than some other languages, it has in fact gone through various syntactic changes. These include changes from OV to VO word order, the emergence of the first position expletive \u00fea\u00f0 'it, there' and decreased use of certain types of empty arguments (see R\u00f6gnvaldsson, 2005 for an overview).", |
| "cite_spans": [ |
| { |
| "start": 942, |
| "end": 978, |
| "text": "(R\u00f6gnvaldsson and Helgad\u00f3ttir, 2011)", |
| "ref_id": "BIBREF17" |
| }, |
| { |
| "start": 1347, |
| "end": 1365, |
| "text": "R\u00f6gnvaldsson, 2005", |
| "ref_id": "BIBREF20" |
| } |
| ], |
| "ref_spans": [ |
| { |
| "start": 279, |
| "end": 287, |
| "text": "Figure 1", |
| "ref_id": "FIGREF1" |
| } |
| ], |
| "eq_spans": [], |
| "section": "The Icelandic Parsed Historical Corpus", |
| "sec_num": "2" |
| }, |
| { |
| "text": "The PPCHE and UD annotation schemes reflect two different ways of parsing a sentence. While the PPCHE scheme denotes phrases by using brackets, the UD scheme connects each token to a single head (Nivre et al., 2016) . Every sentence in UD has a root and each dependent relationship a label, but neither are marked in the PPCHE scheme. An obvious difference between the two schemes is how information on a sentence's annotation is displayed. Figure 2 displays the dependency-based counterpart to the IcePaHC sentence in Figure 1 . Comparing the two, we see that the PPCHE format focuses on phrases while the UD format, which is delivered in CoNLL-U format, lists all tokens in a sentence and gives information on them. The CoNLL-U format consists of ten fields, or columns, as outlined below. \u2022 The 1st, 2nd, and 3rd columns contain the word index (ID), the word form (FORM), and the lemma (LEMMA), respectively.", |
| "cite_spans": [ |
| { |
| "start": 195, |
| "end": 215, |
| "text": "(Nivre et al., 2016)", |
| "ref_id": "BIBREF15" |
| } |
| ], |
| "ref_spans": [ |
| { |
| "start": 441, |
| "end": 449, |
| "text": "Figure 2", |
| "ref_id": "FIGREF2" |
| }, |
| { |
| "start": 519, |
| "end": 527, |
| "text": "Figure 1", |
| "ref_id": "FIGREF1" |
| } |
| ], |
| "eq_spans": [], |
| "section": "The PPCHE and UD Annotation Schemes", |
| "sec_num": "3" |
| }, |
| { |
| "text": "\u2022 The 4th column contains the Universal PoS tags (UPOS). This is the UD format tag corresponding to the tag in column 5 and it is discussed further in Section 4.2.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The PPCHE and UD Annotation Schemes", |
| "sec_num": "3" |
| }, |
| { |
| "text": "\u2022 The 5th column contains the language-specific PoS tags as used in IcePaHC (XPOS).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The PPCHE and UD Annotation Schemes", |
| "sec_num": "3" |
| }, |
| { |
| "text": "\u2022 The 6th column contains morphological features (FEATS), which give detailed information on a word beyond its UPOS and XPOS tags, for example case, number, tense, voice and person, all depending on word classes. These are discussed further in Section 4.2.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The PPCHE and UD Annotation Schemes", |
| "sec_num": "3" |
| }, |
| { |
| "text": "\u2022 The 7th column (HEAD) specifies which word the current word is dependent on, by using the head's word index. Each sentence has one root, specified with a '0', on which all other words in the sentence are dependent, whether the relation is direct or through other words.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The PPCHE and UD Annotation Schemes", |
| "sec_num": "3" |
| }, |
| { |
| "text": "\u2022 The 8th column contains the code for the universal dependencies relation (DEPREL), which states of what kind the relation between a word and its head is, discussed in Section 4.3.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The PPCHE and UD Annotation Schemes", |
| "sec_num": "3" |
| }, |
| { |
| "text": "\u2022 The 9th column is left blank but it can include information on enhanced dependencies (DEPS).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The PPCHE and UD Annotation Schemes", |
| "sec_num": "3" |
| }, |
| { |
| "text": "\u2022 Finally, there is a miscellaneous column (MISC). In the case of the Icelandic UD corpus, this column is used for displaying the word's revised Icelandic Frequency Dictionary (IFD) (Pind et al., 1991) morphosyntactic PoS tag, given by the ABLTagger (Steingr\u00edmsson et al., 2019) , as discussed in Section 4.2.", |
| "cite_spans": [ |
| { |
| "start": 182, |
| "end": 201, |
| "text": "(Pind et al., 1991)", |
| "ref_id": "BIBREF16" |
| }, |
| { |
| "start": 250, |
| "end": 278, |
| "text": "(Steingr\u00edmsson et al., 2019)", |
| "ref_id": "BIBREF24" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The PPCHE and UD Annotation Schemes", |
| "sec_num": "3" |
| }, |
| { |
| "text": "Our method of converting IcePaHC to a UD corpus mainly consists of three steps. The first step involves text cleanup, i.e. removing information not critical to UD. Next is the actual conversion, extracting information on the words and phrase structure and delivering it according to the CoNLL-U format. Lastly, various post-processing is done on the CoNLL-U output files, in order to meet UD format standards. The converter and its resulting treebank have not been evaluated formally but numerous sentences have been checked manually while developing the converter. Systematic evaluation will be carried out in future work.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Conversion", |
| "sec_num": "4" |
| }, |
| { |
| "text": "Since the phrase structure of the sentences in IcePaHC is depicted using brackets, the NLTK Catego-rizedBracketParseCorpusReader 6 is used. In order for it to operate, some cleanup needs to be carried out on the corpus files. This mostly consists of removing additional information included in the files, e.g. sentence ID tags and the annotators' notes, along with empty lines and extra brackets. In the annotation scheme, various words that are normally written as one are split into separate tokens, and as the UD format does not follow suit, this process has to be reversed. This includes suffixed articles, e.g. -in in maer-in 'maiden-the'. These are split from the noun in IcePaHC as separate tokens but appear as a part of the noun in the UD scheme. In addition to these changes, a specific script fixes various minor annotation errors in the corpora themselves, discovered while the converter was being developed. The second part of this cleanup stage is carried out after the conversion itself, in which the output CoNLL-U files are modified. For example, we transparently portray the cliticization of pronouns to their corresponding verbs in the CoNLL-U output. The annotation scheme splits these pronouns or pronoun clitics from their corresponding verbs such that, e.g., heyr\u00f0u (imperative 'hear (you)') is split into the imperative form heyr 'hear' and the second person singular pronoun clitic -\u00f0u 'you'. Keeping consistent with UD annotation guidelines (e.g., for Spanish and German), we adopt the annotation scheme shown in Figure 3 , where the word components are shown both combined (ID 1-2) and split (IDs 1 and 2), so that the surface form is apparent but all features and dependency relations of the components are also clear. Included in this last phase of the conversion pipeline is joining together sentences so that a sentence in the final output is defined by a full stop rather than by a matrix clause, to adhere to UD convention.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 1539, |
| "end": 1547, |
| "text": "Figure 3", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Text Cleanup", |
| "sec_num": "4.1" |
| }, |
| { |
| "text": "ID FORM LEMMA UPOS 1-2 Heyr\u00f0u _ _ 1", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Text Cleanup", |
| "sec_num": "4.1" |
| }, |
| { |
| "text": "Heyr heyra VERB 2 \u00fe\u00fa \u00fe\u00fa PRON Figure 3 : An example of a verb-clitic relation marked by index range (further CoNLL-U fields omitted).", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 29, |
| "end": 37, |
| "text": "Figure 3", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Text Cleanup", |
| "sec_num": "4.1" |
| }, |
| { |
| "text": "Information on PoS tags and morphological features is extracted in two ways. To obtain the UD tag, a word's original tag from IcePaHC is used along with handwritten rules, which map each original tag to a corresponding UD tag. The tagset used in the converted corpus consists of 17 tags, displayed in Table 1 . The mapping from an IcePaHC tag to a UD tag is not always unequivocal since the tagging scheme used in the corpus groups some word classes together, which are kept separate in the UD tagging scheme, and vice versa. An example of this variation is the treatment of quantifiers. They are not included in the PoS categories used in UD but they are included in the IcePaHC tagset, and so an appropriate tag has to be chosen. Some quantifiers are classified as adjectives in UD and some as pronouns, but this distinction is not marked in the original corpus and therefore one tag has to be chosen for all. Since the majority of quantifiers are categorized as adjectives, a decision was made to map quantifiers to adjectives. Another problem of a similar kind is that verbs in IcePaHC can have the same tag whether they are a main or an auxiliary verb, but in UD the tags are different. For example, the Icelandic verb hafa 'have' can both act as a main verb, as in (2), and auxiliary verb, as in (3), but it is in both cases tagged as HV in IcePaHC. Therefore, a correct tag has to be chosen based on its context at any given time. Although it is thorough, the IcePaHC PoS tagset does not include enough information to display the complete morphological features for all words in the converted corpus, as various grammatical features are omitted from the annotation. For example, grammatical gender (masculine, feminine and neuter) is not present in the tagset, which affects, e.g., nouns and adjectives. This means that a full, automatic conversion from IcePaHC tags to the UD feature scheme, which correctly describes the grammatical features of Icelandic, is not possible in a simple step.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 301, |
| "end": 308, |
| "text": "Table 1", |
| "ref_id": "TABREF2" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Part-of-Speech Tags and Morphological Features", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "To obtain more information on the tokens, the IcePaHC PoS tags are skipped altogether and the whole text of the treebank is extracted and automatically PoS-tagged using ABLTagger, a state-of-the-art PoS tagger for Icelandic (Steingr\u00edmsson et al., 2019) . It returns IFD-format PoS tags, mentioned in Section 3, from which morphological features are extracted. The output of this step, although not hand-tagged, is considerably more detailed than the IcePaHC tags, thus providing all necessary morphological features for the output UD corpus, as opposed to the incomplete IcePaHC PoS tags. 7 The tags from the ABLTagger are shown in the miscellaneous (MISC) column in the CoNLL-U format in Figure 2 .", |
| "cite_spans": [ |
| { |
| "start": 224, |
| "end": 252, |
| "text": "(Steingr\u00edmsson et al., 2019)", |
| "ref_id": "BIBREF24" |
| }, |
| { |
| "start": 589, |
| "end": 590, |
| "text": "7", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [ |
| { |
| "start": 689, |
| "end": 697, |
| "text": "Figure 2", |
| "ref_id": "FIGREF2" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Part-of-Speech Tags and Morphological Features", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "Extracting relational information from the IcePaHC format involves selecting each phrase's head and its dependents and the type of dependency relation. For this, we build on an experimental project carried out by \u00d6rvar K\u00e1rason. 8 The method of conversion is based on similar work originally done by Magerman (1994) and improved by others (Yamada and Matsumoto, 2003; Johansson and Nugues, 2007) .", |
| "cite_spans": [ |
| { |
| "start": 228, |
| "end": 229, |
| "text": "8", |
| "ref_id": null |
| }, |
| { |
| "start": 338, |
| "end": 366, |
| "text": "(Yamada and Matsumoto, 2003;", |
| "ref_id": "BIBREF27" |
| }, |
| { |
| "start": 367, |
| "end": 394, |
| "text": "Johansson and Nugues, 2007)", |
| "ref_id": "BIBREF8" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Heads and Dependency Relations", |
| "sec_num": "4.3" |
| }, |
| { |
| "text": "Each sentence in the given corpus is read with the help of NLTK 9 and information on tags from the original files is also used. A matrix clause is read bottom-up to determine the head of all its subordinate phrases, skipping traces and empty nodes. This head selection is done with handwritten rules that match a phrase tag with its possible heads, which in turn are ordered according to priority so that a correct head is chosen if more than one are possible. The IcePaHC PoS tags are suitable for this task and as they have been manually corrected, no information on phrase type is lost during conversion between tagsets. By selecting phrasal heads bottom-up, we establish a hierarchy where every word or phrase within a phrase is dependent on the head, culminating in the head of the whole sentence, which, according to UD guidelines, is the sentence's root.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Heads and Dependency Relations", |
| "sec_num": "4.3" |
| }, |
| { |
| "text": "When determining the type of dependency relation between two words, information on a phrasal head is used along with the words' PoS tags. Handwritten rules are used to match a tag with its dependency relation, some of which specify the phrasal head's tag to handle ambiguous cases. Again, the detailed tagset of IcePaHC makes this process possible. The dependency relations used can be seen in Table 2 . All dependency relations used were crucial for the conversion to be precise and, in most cases, phrasal heads in IcePaHC correspond to a dependency relation. An exception to this are the relations 'acl ', 'advcl', 'ccomp' and ' xcomp'. IcePaHC does not make a distinction between 'acl' and 'advcl' on the one hand and 'ccomp' and 'xcomp' on the other, but UD does. These dependency relations are therefore handled specifically and the appropriate relation chosen based on the sentence it appears in.", |
| "cite_spans": [ |
| { |
| "start": 606, |
| "end": 631, |
| "text": "', 'advcl', 'ccomp' and '", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [ |
| { |
| "start": 394, |
| "end": 401, |
| "text": "Table 2", |
| "ref_id": "TABREF4" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Heads and Dependency Relations", |
| "sec_num": "4.3" |
| }, |
| { |
| "text": "Not only are the dependency relations important, but their direction is too. Some relations can only go from left to right and others from right to left, and this had to be considered when creating the converter since no equivalent rule is present in IcePaHC. In some cases, the original annotation is changed to adhere to UD rules, for example when a phrase includes two auxiliary verbs, one of them being the phrasal head. The other auxiliary verb cannot be dependent on another auxiliary verb and the relation must therefore be changed, so that both auxiliary verbs are dependent on a main verb. ", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Heads and Dependency Relations", |
| "sec_num": "4.3" |
| }, |
| { |
| "text": "The goal of this project is not to produce a one-shot converter for a specific treebank, but a reusable tool, applicable to treebanks in a similar format. Icelandic and Faroese share a number of linguistic properties that make constituency grammar annotation using the same scheme possible. To demonstrate the reusability of the tool, it was adapted to the Faroese Parsed Historical Corpus (FarPaHC) 10 which uses the same annotation scheme as IcePaHC. It consists of three New Testament translations, i.e. one from the 19 th century and two from the 20 th century, 53,000 words in total. The 19 th century text has not been included in the converted UD treebank yet. Its orthography deviates in many ways from modern spelling; even though the text was modernized for FarPaHC, we decided that before we would add it to the converted treebank, more work on fixing inconsistencies would be needed. A few minor additions were needed for the conversion to be possible, which shows the utility of the conversion pipeline. The annotation process for FarPaHC is the same as that for IcePaHC; both of them have been manually corrected according to the PPCHE annotation scheme, with some changes to adapt it to Icelandic and Faroese grammar. The tagset used in FarPaHC is for the most part the same as in IcePaHC, which is possible because of the similarities in the languages' grammars. The main difference in the annotation scheme between the two corpora is that lemmas are not shown in FarPaHC. An example of a parsed sentence in FarPaHC is shown in Figure 4 , which shows a matrix clause (IP-MAT) with a direct object (NP-OB1), an auxiliary verb (HVPI), a subject (NP-SBJ), a main verb in the past participle (VBN) and two prepositional phrases (PP); the sentence is glossed and translated in (4). Since the two treebanks share an annotation scheme and a tagset, FarPaHC was converted using the methods described in Section 4 with only three modifications. First, the script used to fix various annotation errors in IcePaHC was not used as it did not apply to FarPaHC. No such errors were encountered and such a script was therefore unnecessary. Second, since FarPaHC does not include information on lemmas, neither does the resulting UD corpus. Third, morphological features for FarPaHC were extracted in a different manner. The conversion from IcePaHC relies on an Icelandic high-accuracy PoS tagger for additional information on tokens but no such tagger exists for Faroese. Thus, no third-party software was used for feature extraction and all information was retrieved from the PoS tags themselves. Every component in a tag was mapped to a corresponding morphological feature using handwritten rules, which does not provide all possible features for a given token but does provide more details than solely the tag. These rules can then be built upon if the conversion pipeline is extended for another language with a similar tagset.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 1544, |
| "end": 1552, |
| "text": "Figure 4", |
| "ref_id": "FIGREF4" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Extension: Converting a Faroese Treebank", |
| "sec_num": "5" |
| }, |
| { |
| "text": "The conversion results in a new Faroese UD corpus of 40,000 words to be included in the next UD release. Figure 5 shows the sentence from Figure 4 after having been converted to CoNLL-U. As mentioned, lemmas are not shown, with the exception of the punctuation mark, fewer morphological features are shown and no additional tags are shown in the MISC column. In other respects, all grammatical information represented in the converted Icelandic UD corpus is also represented in the Faroese one. Figure 5 : A sentence converted to UD from FarPaHC.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 105, |
| "end": 113, |
| "text": "Figure 5", |
| "ref_id": null |
| }, |
| { |
| "start": 138, |
| "end": 146, |
| "text": "Figure 4", |
| "ref_id": "FIGREF4" |
| }, |
| { |
| "start": 495, |
| "end": 503, |
| "text": "Figure 5", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Extension: Converting a Faroese Treebank", |
| "sec_num": "5" |
| }, |
| { |
| "text": "IcePaHC currently consists of about one million words but 75,000 additional words were annotated in this project by Kristj\u00e1n R\u00fanarsson in accordance with the IcePaHC scheme (R\u00fanarsson and Sigur\u00f0sson, 2020) . This is done to increase the weight of modern Icelandic in the corpus. These additions, which contain genres not found previously in IcePaHC -parliamentary speeches and sports news texts taken from the Icelandic Gigaword Corpus (Steingr\u00edmsson et al., 2018) -will be converted to the UD scheme using the conversion tool described in this paper, making for a close to 1.1 million word corpus. Not only does this enlarge the corpus, but also increases the ratio of modern texts in it.", |
| "cite_spans": [ |
| { |
| "start": 173, |
| "end": 205, |
| "text": "(R\u00fanarsson and Sigur\u00f0sson, 2020)", |
| "ref_id": "BIBREF22" |
| }, |
| { |
| "start": 436, |
| "end": 464, |
| "text": "(Steingr\u00edmsson et al., 2018)", |
| "ref_id": "BIBREF23" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Future Development and Use", |
| "sec_num": "6" |
| }, |
| { |
| "text": "A possible improvement to the Faroese UD treebank is adding information on lemmas. Since FarPaHC does not include lemmas, a Faroese lemmatizer would have to be used. A few lemmatizers are available for this task (Yildiz and Tantug, 2019; Kanerva et al., 2018; Rosa and Marecek, 2018) , which would have to be evaluated further in regards to their feasibility.", |
| "cite_spans": [ |
| { |
| "start": 212, |
| "end": 237, |
| "text": "(Yildiz and Tantug, 2019;", |
| "ref_id": "BIBREF28" |
| }, |
| { |
| "start": 238, |
| "end": 259, |
| "text": "Kanerva et al., 2018;", |
| "ref_id": "BIBREF10" |
| }, |
| { |
| "start": 260, |
| "end": 283, |
| "text": "Rosa and Marecek, 2018)", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Future Development and Use", |
| "sec_num": "6" |
| }, |
| { |
| "text": "Compatibility between the current project and the IcePaHC and FarPaHC corpora means that various possibilities exist for future interlinking of projects that make use of the two resources. Firstly, by using the conversion tool for both an Icelandic and a Faroese treebank, we have shown that cross-lingual use is possible. Secondly, with the possibility of user customization, e.g. specifying language-specific rules and parameters, the converter may be applied -with necessary modifications -to various other constituency treebanks to produce new UD corpora for a wide range of languages. These include corpora of Historical Greek (Beck, 2011) , Historical Portuguese (Galves, 2018) and Middle Low German (Booth et al., 2020) , in addition to PPCHE. Finally, the conversion tool can also be used to convert output from automatic parsers which use the IcePaHC annotation scheme, e.g. IceNeuralParsingPipeline (Arnard\u00f3ttir and Ingason, 2020) , to UD format. Since the original corpus would be automatically created, the UD format output would have to be manually corrected before adding it to the Icelandic UD corpus.", |
| "cite_spans": [ |
| { |
| "start": 632, |
| "end": 644, |
| "text": "(Beck, 2011)", |
| "ref_id": "BIBREF1" |
| }, |
| { |
| "start": 669, |
| "end": 683, |
| "text": "(Galves, 2018)", |
| "ref_id": "BIBREF4" |
| }, |
| { |
| "start": 706, |
| "end": 726, |
| "text": "(Booth et al., 2020)", |
| "ref_id": "BIBREF3" |
| }, |
| { |
| "start": 909, |
| "end": 940, |
| "text": "(Arnard\u00f3ttir and Ingason, 2020)", |
| "ref_id": "BIBREF0" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Future Development and Use", |
| "sec_num": "6" |
| }, |
| { |
| "text": "In this paper, we have described a rule-based conversion pipeline applicable to the Icelandic Parsed Historical Corpus and the Faroese Parsed Historical Corpus, which results in corpora annotated according to Universal Dependencies. The pipeline is used for converting the two treebanks, delivering a 1 million word Icelandic UD corpus and a 40,000 word Faroese UD corpus. The Icelandic corpus will consequently be among the largest of its type.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusion", |
| "sec_num": "7" |
| }, |
| { |
| "text": "We have discussed the benefits of a UD-based corpus for Icelandic and Faroese and why IcePaHC is a good candidate for building the conversion on. The process of converting sentences was described and the methodology behind it, extracting information needed for a UD corpus by using both information included in the IcePaHC annotation scheme and a state-of-the-art PoS tagger for Icelandic. We also discussed some challenges faced because of the different annotation scheme of the original treebank and UD and how these were dealt with. We demonstrated the conversion pipeline's applicability to different languages by converting a Faroese constituency treebank to a UD corpus using the pipeline. Lastly, we discussed further possibilities in using the converter and the UD corpora with available language resources along with the 75,000 word addition to IcePaHC, which will be converted using the pipeline and added to the Icelandic UD corpus.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusion", |
| "sec_num": "7" |
| }, |
| { |
| "text": "The benefits of an Icelandic Universal Dependencies corpus are significant, as the UD annotation scheme offers a standard parsing framework across different languages. The information that is annotated in IcePaHC is a superset of what is required by the UD scheme and this makes conversion between the two formats a feasible approach to creating a UD treebank for Icelandic. Furthermore, the open-source policy of IcePaHC as well as the global UD project is important and therefore maintained in this project. The conversion tool is released under an Apache 2.0 license on GitHub 11 and the two UD corpora are included in the next release of UD under a CC BY-SA 4.0 license.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusion", |
| "sec_num": "7" |
| }, |
| { |
| "text": "A description of the tagset can be found at https://linguist.is/icelandic_treebank/Tagset", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "https://www.nltk.org/_modules/nltk/corpus/reader/bracket_parse.html", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "This approach has various issues in itself, both in execution, as it relies on external software, and in output accuracy, as the PoS tagger used is considered state-of-the-art for modern Icelandic, but presumably performs worse as the input texts deviate more from the standard language and orthography. This will eventually be hand-checked and evaluated.8 https://github.com/OKarason/venzl 9 https://www.nltk.org/_modules/nltk/corpus/reader/bracket_parse.html", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "https://github.com/einarfs/farpahc", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "https://github.com/thorunna/UDConverter", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| } |
| ], |
| "back_matter": [ |
| { |
| "text": "This project was funded by the Strategic Research and Development Programme for Language Technology, grant no. 180020-5301. The conversion of the Faroese corpus was funded by the Icelandic Student Innovation Fund, grant no. 206457-0091. Thanks are due to \u00d6rvar K\u00e1rason, whose work was used as a basis for the conversion. We also thank Kristj\u00e1n R\u00fanarsson for helpful discussions. Finally, we would like to thank two anonymous reviewers for their helpful comments.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Acknowledgements", |
| "sec_num": "8" |
| } |
| ], |
| "bib_entries": { |
| "BIBREF0": { |
| "ref_id": "b0", |
| "title": "A neural parsing pipeline for Icelandic using the Berkeley neural parser", |
| "authors": [ |
| { |
| "first": "\u00de\u00f3runn", |
| "middle": [], |
| "last": "Arnard\u00f3ttir", |
| "suffix": "" |
| }, |
| { |
| "first": "Anton", |
| "middle": [ |
| "Karl" |
| ], |
| "last": "Ingason", |
| "suffix": "" |
| } |
| ], |
| "year": 2020, |
| "venue": "Proceedings of CLARIN 2020", |
| "volume": "", |
| "issue": "", |
| "pages": "48--51", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "\u00de\u00f3runn Arnard\u00f3ttir and Anton Karl Ingason. 2020. A neural parsing pipeline for Icelandic using the Berkeley neural parser. In Costanza Navarretta and Maria Eskevich, editors, Proceedings of CLARIN 2020, pages 48-51.", |
| "links": null |
| }, |
| "BIBREF1": { |
| "ref_id": "b1", |
| "title": "Penn Parsed Corpora of Historical Greek (PPCHiG)", |
| "authors": [ |
| { |
| "first": "Jana", |
| "middle": [ |
| "E" |
| ], |
| "last": "Beck", |
| "suffix": "" |
| } |
| ], |
| "year": 2011, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Jana E. Beck. 2011. Penn Parsed Corpora of Historical Greek (PPCHiG). http://ling.upenn.edu/~janabeck/greek- corpora.html.", |
| "links": null |
| }, |
| "BIBREF2": { |
| "ref_id": "b2", |
| "title": "Natural Language Processing with Python", |
| "authors": [ |
| { |
| "first": "Steven", |
| "middle": [], |
| "last": "Bird", |
| "suffix": "" |
| }, |
| { |
| "first": "Ewan", |
| "middle": [], |
| "last": "Klein", |
| "suffix": "" |
| }, |
| { |
| "first": "Edward", |
| "middle": [], |
| "last": "Loper", |
| "suffix": "" |
| } |
| ], |
| "year": 2009, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Steven Bird, Ewan Klein, and Edward Loper. 2009. Natural Language Processing with Python. O'Reilly Media, Sebastopol, CA.", |
| "links": null |
| }, |
| "BIBREF3": { |
| "ref_id": "b3", |
| "title": "A Penn-style treebank of Middle Low German", |
| "authors": [ |
| { |
| "first": "Hannah", |
| "middle": [], |
| "last": "Booth", |
| "suffix": "" |
| }, |
| { |
| "first": "Anne", |
| "middle": [], |
| "last": "Breitbarth", |
| "suffix": "" |
| }, |
| { |
| "first": "Aaron", |
| "middle": [], |
| "last": "Ecay", |
| "suffix": "" |
| }, |
| { |
| "first": "Melissa", |
| "middle": [], |
| "last": "Farasyn", |
| "suffix": "" |
| } |
| ], |
| "year": 2020, |
| "venue": "Proceedings of The 12th Language Resources and Evaluation Conference (LREC 2020)", |
| "volume": "", |
| "issue": "", |
| "pages": "766--775", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Hannah Booth, Anne Breitbarth, Aaron Ecay, and Melissa Farasyn. 2020. A Penn-style treebank of Middle Low German. In Proceedings of The 12th Language Resources and Evaluation Conference (LREC 2020), pages 766-775, Marseille, France. European Language Resources Association.", |
| "links": null |
| }, |
| "BIBREF4": { |
| "ref_id": "b4", |
| "title": "The Tycho Brahe Corpus of Historical Portuguese: Methodology and results. Linguistic Variation", |
| "authors": [ |
| { |
| "first": "Charlotte", |
| "middle": [], |
| "last": "Galves", |
| "suffix": "" |
| } |
| ], |
| "year": 2018, |
| "venue": "", |
| "volume": "18", |
| "issue": "", |
| "pages": "49--73", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Charlotte Galves. 2018. The Tycho Brahe Corpus of Historical Portuguese: Methodology and results. Linguistic Variation, 18(1):49-73.", |
| "links": null |
| }, |
| "BIBREF5": { |
| "ref_id": "b5", |
| "title": "A Mixed Method Lemmatization Algorithm Using a Hierarchy of Linguistic Identities (HOLI)", |
| "authors": [ |
| { |
| "first": "Anton", |
| "middle": [], |
| "last": "Karl Ingason", |
| "suffix": "" |
| }, |
| { |
| "first": "Sigr\u00fan", |
| "middle": [], |
| "last": "Helgad\u00f3ttir", |
| "suffix": "" |
| } |
| ], |
| "year": 2008, |
| "venue": "Proceedings of Sixth International Conference on Natural Language Processing", |
| "volume": "", |
| "issue": "", |
| "pages": "205--216", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Anton Karl Ingason, Sigr\u00fan Helgad\u00f3ttir, Hrafn Loftsson, and Eir\u00edkur R\u00f6gnvaldsson. 2008. A Mixed Method Lemmatization Algorithm Using a Hierarchy of Linguistic Identities (HOLI). In Proceedings of Sixth Interna- tional Conference on Natural Language Processing, GoTAL 2008, pages 205-216, Gothenburg, Sweden.", |
| "links": null |
| }, |
| "BIBREF6": { |
| "ref_id": "b6", |
| "title": "Faroese Parsed Historical Corpus (FarPaHC)", |
| "authors": [ |
| { |
| "first": "Anton", |
| "middle": [], |
| "last": "Karl Ingason", |
| "suffix": "" |
| }, |
| { |
| "first": "Eir\u00edkur", |
| "middle": [], |
| "last": "R\u00f6gnvaldsson", |
| "suffix": "" |
| }, |
| { |
| "first": "Joel", |
| "middle": [ |
| "C" |
| ], |
| "last": "Einar Freyr Sigur\u00f0sson", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Wallenberg", |
| "suffix": "" |
| } |
| ], |
| "year": 2012, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Anton Karl Ingason, Eir\u00edkur R\u00f6gnvaldsson, Einar Freyr Sigur\u00f0sson, and Joel C. Wallenberg. 2012. Faroese Parsed Historical Corpus (FarPaHC). Version 0.1.", |
| "links": null |
| }, |
| "BIBREF7": { |
| "ref_id": "b7", |
| "title": "Rapid deployment of phrase structure parsing for related languages: A case study of Insular Scandinavian", |
| "authors": [ |
| { |
| "first": "Anton", |
| "middle": [], |
| "last": "Karl Ingason", |
| "suffix": "" |
| }, |
| { |
| "first": "Hrafn", |
| "middle": [], |
| "last": "Loftsson", |
| "suffix": "" |
| }, |
| { |
| "first": "Eir\u00edkur", |
| "middle": [], |
| "last": "R\u00f6gnvaldsson", |
| "suffix": "" |
| } |
| ], |
| "year": 2014, |
| "venue": "Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC 2014)", |
| "volume": "", |
| "issue": "", |
| "pages": "91--95", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Anton Karl Ingason, Hrafn Loftsson, Eir\u00edkur R\u00f6gnvaldsson, Einar Freyr Sigur\u00f0sson, and Joel Wallenberg. 2014. Rapid deployment of phrase structure parsing for related languages: A case study of Insular Scandinavian. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC 2014), pages 91-95, Reykjav\u00edk, Iceland. European Language Resources Association.", |
| "links": null |
| }, |
| "BIBREF8": { |
| "ref_id": "b8", |
| "title": "Extended constituent-to-dependency conversion for English", |
| "authors": [ |
| { |
| "first": "Richard", |
| "middle": [], |
| "last": "Johansson", |
| "suffix": "" |
| }, |
| { |
| "first": "Pierre", |
| "middle": [], |
| "last": "Nugues", |
| "suffix": "" |
| } |
| ], |
| "year": 2007, |
| "venue": "Proceedings of the 16th Nordic Conference of Computational Linguistics (NODALIDA 2007)", |
| "volume": "", |
| "issue": "", |
| "pages": "105--112", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Richard Johansson and Pierre Nugues. 2007. Extended constituent-to-dependency conversion for English. In Proceedings of the 16th Nordic Conference of Computational Linguistics (NODALIDA 2007), pages 105-112, Tartu, Estonia. University of Tartu, Estonia.", |
| "links": null |
| }, |
| "BIBREF9": { |
| "ref_id": "b9", |
| "title": "Creating a parallel Icelandic dependency treebank from raw text to Universal Dependencies", |
| "authors": [ |
| { |
| "first": "Hildur", |
| "middle": [], |
| "last": "J\u00f3nsd\u00f3ttir", |
| "suffix": "" |
| }, |
| { |
| "first": "Anton", |
| "middle": [ |
| "Karl" |
| ], |
| "last": "Ingason", |
| "suffix": "" |
| } |
| ], |
| "year": 2020, |
| "venue": "Proceedings of The 12th Language Resources and Evaluation Conference (LREC 2020)", |
| "volume": "", |
| "issue": "", |
| "pages": "2924--2931", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Hildur J\u00f3nsd\u00f3ttir and Anton Karl Ingason. 2020. Creating a parallel Icelandic dependency treebank from raw text to Universal Dependencies. In Proceedings of The 12th Language Resources and Evaluation Conference (LREC 2020), pages 2924-2931, Marseille, France. European Language Resources Association.", |
| "links": null |
| }, |
| "BIBREF10": { |
| "ref_id": "b10", |
| "title": "Turku neural parser pipeline: An end-to-end system for the CoNLL 2018 shared task", |
| "authors": [ |
| { |
| "first": "Jenna", |
| "middle": [], |
| "last": "Kanerva", |
| "suffix": "" |
| }, |
| { |
| "first": "Filip", |
| "middle": [], |
| "last": "Ginter", |
| "suffix": "" |
| }, |
| { |
| "first": "Niko", |
| "middle": [], |
| "last": "Miekka", |
| "suffix": "" |
| }, |
| { |
| "first": "Akseli", |
| "middle": [], |
| "last": "Leino", |
| "suffix": "" |
| }, |
| { |
| "first": "Tapio", |
| "middle": [], |
| "last": "Salakoski", |
| "suffix": "" |
| } |
| ], |
| "year": 2018, |
| "venue": "Proceedings of the CoNLL 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies", |
| "volume": "", |
| "issue": "", |
| "pages": "133--142", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Jenna Kanerva, Filip Ginter, Niko Miekka, Akseli Leino, and Tapio Salakoski. 2018. Turku neural parser pipeline: An end-to-end system for the CoNLL 2018 shared task. In Proceedings of the CoNLL 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies, pages 133-142, Brussels, Belgium. Association for Computational Linguistics.", |
| "links": null |
| }, |
| "BIBREF11": { |
| "ref_id": "b11", |
| "title": "Penn-Helsinki Parsed Corpus of Middle English. CD-ROM", |
| "authors": [ |
| { |
| "first": "Anthony", |
| "middle": [ |
| "S" |
| ], |
| "last": "Kroch", |
| "suffix": "" |
| }, |
| { |
| "first": "Ann", |
| "middle": [], |
| "last": "Taylor", |
| "suffix": "" |
| } |
| ], |
| "year": 2000, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Anthony S. Kroch and Ann Taylor. 2000. Penn-Helsinki Parsed Corpus of Middle English. CD-ROM. Second edition. Size: 1.3 million words.", |
| "links": null |
| }, |
| "BIBREF12": { |
| "ref_id": "b12", |
| "title": "Penn-Helsinki Parsed Corpus of Early Modern English. CD-ROM", |
| "authors": [ |
| { |
| "first": "Anthony", |
| "middle": [ |
| "S" |
| ], |
| "last": "Kroch", |
| "suffix": "" |
| }, |
| { |
| "first": "Beatrice", |
| "middle": [], |
| "last": "Santorini", |
| "suffix": "" |
| }, |
| { |
| "first": "Lauren", |
| "middle": [], |
| "last": "Delfs", |
| "suffix": "" |
| } |
| ], |
| "year": 2004, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Anthony S. Kroch, Beatrice Santorini, and Lauren Delfs. 2004. Penn-Helsinki Parsed Corpus of Early Modern English. CD-ROM. First Edition. Size: 1.8 million words.", |
| "links": null |
| }, |
| "BIBREF13": { |
| "ref_id": "b13", |
| "title": "IceNLP: A Natural Language Processing Toolkit for Icelandic", |
| "authors": [ |
| { |
| "first": "Hrafn", |
| "middle": [], |
| "last": "Loftsson", |
| "suffix": "" |
| }, |
| { |
| "first": "Eir\u00edkur", |
| "middle": [], |
| "last": "R\u00f6gnvaldsson", |
| "suffix": "" |
| } |
| ], |
| "year": 2007, |
| "venue": "Proceedings of Interspeech -Speech and language technology for less-resourced languages", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Hrafn Loftsson and Eir\u00edkur R\u00f6gnvaldsson. 2007. IceNLP: A Natural Language Processing Toolkit for Icelandic. In Proceedings of Interspeech -Speech and language technology for less-resourced languages, Interspeech 2007, Antwerp, Belgium.", |
| "links": null |
| }, |
| "BIBREF14": { |
| "ref_id": "b14", |
| "title": "Natural Language Parsing As Statistical Pattern Recognition", |
| "authors": [ |
| { |
| "first": "David", |
| "middle": [], |
| "last": "Mitchell Magerman", |
| "suffix": "" |
| } |
| ], |
| "year": 1994, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "David Mitchell Magerman. 1994. Natural Language Parsing As Statistical Pattern Recognition. Ph.D. thesis, Stanford University.", |
| "links": null |
| }, |
| "BIBREF15": { |
| "ref_id": "b15", |
| "title": "Universal Dependencies v1: A multilingual treebank collection", |
| "authors": [ |
| { |
| "first": "Joakim", |
| "middle": [], |
| "last": "Nivre", |
| "suffix": "" |
| }, |
| { |
| "first": "Marie-Catherine", |
| "middle": [], |
| "last": "De Marneffe", |
| "suffix": "" |
| }, |
| { |
| "first": "Filip", |
| "middle": [], |
| "last": "Ginter", |
| "suffix": "" |
| }, |
| { |
| "first": "Yoav", |
| "middle": [], |
| "last": "Goldberg", |
| "suffix": "" |
| }, |
| { |
| "first": "Jan", |
| "middle": [], |
| "last": "Hajic", |
| "suffix": "" |
| }, |
| { |
| "first": "Christopher", |
| "middle": [ |
| "D" |
| ], |
| "last": "Manning", |
| "suffix": "" |
| }, |
| { |
| "first": "Ryan", |
| "middle": [], |
| "last": "Mcdonald", |
| "suffix": "" |
| }, |
| { |
| "first": "Slav", |
| "middle": [], |
| "last": "Petrov", |
| "suffix": "" |
| }, |
| { |
| "first": "Sampo", |
| "middle": [], |
| "last": "Pyysalo", |
| "suffix": "" |
| }, |
| { |
| "first": "Natalia", |
| "middle": [], |
| "last": "Silveira", |
| "suffix": "" |
| }, |
| { |
| "first": "Reut", |
| "middle": [], |
| "last": "Tsarfaty", |
| "suffix": "" |
| }, |
| { |
| "first": "Daniel", |
| "middle": [], |
| "last": "Zeman", |
| "suffix": "" |
| } |
| ], |
| "year": 2016, |
| "venue": "Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC 2016)", |
| "volume": "", |
| "issue": "", |
| "pages": "1659--1666", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Joakim Nivre, Marie-Catherine de Marneffe, Filip Ginter, Yoav Goldberg, Jan Hajic, Christopher D. Manning, Ryan McDonald, Slav Petrov, Sampo Pyysalo, Natalia Silveira, Reut Tsarfaty, and Daniel Zeman. 2016. Universal Dependencies v1: A multilingual treebank collection. In Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC 2016), pages 1659-1666, Paris, France. European Language Resources Association.", |
| "links": null |
| }, |
| "BIBREF16": { |
| "ref_id": "b16", |
| "title": "\u00cdslensk or\u00f0t\u00ed\u00f0nib\u00f3k [The Icelandic Frequency Dictionary", |
| "authors": [ |
| { |
| "first": "J\u00f6rgen", |
| "middle": [], |
| "last": "Pind", |
| "suffix": "" |
| }, |
| { |
| "first": "Fri\u00f0rik", |
| "middle": [], |
| "last": "Magn\u00fasson", |
| "suffix": "" |
| }, |
| { |
| "first": "Stef\u00e1n", |
| "middle": [], |
| "last": "Briem", |
| "suffix": "" |
| } |
| ], |
| "year": 1991, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "J\u00f6rgen Pind, Fri\u00f0rik Magn\u00fasson, and Stef\u00e1n Briem. 1991. \u00cdslensk or\u00f0t\u00ed\u00f0nib\u00f3k [The Icelandic Frequency Dictio- nary]. The Institute of Lexicography, University of Iceland, Reykjav\u00edk.", |
| "links": null |
| }, |
| "BIBREF17": { |
| "ref_id": "b17", |
| "title": "Morphological tagging of Old Icelandic texts and its use in studying syntactic variation and change", |
| "authors": [ |
| { |
| "first": "Eir\u00edkur", |
| "middle": [], |
| "last": "R\u00f6gnvaldsson", |
| "suffix": "" |
| }, |
| { |
| "first": "Sigr\u00fan", |
| "middle": [], |
| "last": "Helgad\u00f3ttir", |
| "suffix": "" |
| } |
| ], |
| "year": 2011, |
| "venue": "Language Technology for Cultural Heritage: Selected Papers from the LaTeCH Workshop Series", |
| "volume": "", |
| "issue": "", |
| "pages": "63--76", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Eir\u00edkur R\u00f6gnvaldsson and Sigr\u00fan Helgad\u00f3ttir. 2011. Morphological tagging of Old Icelandic texts and its use in studying syntactic variation and change. In Caroline Sporleder, Antal van den Bosch, and Kalliopi Zervanou, editors, Language Technology for Cultural Heritage: Selected Papers from the LaTeCH Workshop Series, pages 63-76, Berlin. Springer.", |
| "links": null |
| }, |
| "BIBREF18": { |
| "ref_id": "b18", |
| "title": "Creating a dualpurpose treebank", |
| "authors": [ |
| { |
| "first": "Eir\u00edkur", |
| "middle": [], |
| "last": "R\u00f6gnvaldsson", |
| "suffix": "" |
| } |
| ], |
| "year": 2011, |
| "venue": "Einar Freyr Sigur\u00f0sson, and Joel Wallenberg", |
| "volume": "26", |
| "issue": "", |
| "pages": "139--150", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Eir\u00edkur R\u00f6gnvaldsson, Anton Karl Ingason, Einar Freyr Sigur\u00f0sson, and Joel Wallenberg. 2011. Creating a dual- purpose treebank. JLCL, 26(2):139-150.", |
| "links": null |
| }, |
| "BIBREF19": { |
| "ref_id": "b19", |
| "title": "The Icelandic Parsed Historical Corpus (IcePaHC)", |
| "authors": [ |
| { |
| "first": "Eir\u00edkur", |
| "middle": [], |
| "last": "R\u00f6gnvaldsson", |
| "suffix": "" |
| }, |
| { |
| "first": "Anton", |
| "middle": [ |
| "Karl" |
| ], |
| "last": "Ingason", |
| "suffix": "" |
| }, |
| { |
| "first": "Einar", |
| "middle": [], |
| "last": "Freyr Sigur\u00f0sson", |
| "suffix": "" |
| }, |
| { |
| "first": "Joel", |
| "middle": [], |
| "last": "Wallenberg", |
| "suffix": "" |
| } |
| ], |
| "year": 2012, |
| "venue": "Proceedings of the Eight International Conference on Language Resources and Evaluation (LREC 2012)", |
| "volume": "", |
| "issue": "", |
| "pages": "1977--1984", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Eir\u00edkur R\u00f6gnvaldsson, Anton Karl Ingason, Einar Freyr Sigur\u00f0sson, and Joel Wallenberg. 2012. The Icelandic Parsed Historical Corpus (IcePaHC). In Proceedings of the Eight International Conference on Language Re- sources and Evaluation (LREC 2012), pages 1977-1984, Istanbul, Turkey. European Language Resource Asso- ciation.", |
| "links": null |
| }, |
| "BIBREF20": { |
| "ref_id": "b20", |
| "title": "Setningafrae\u00f0ilegar breytingar \u00ed \u00edslensku", |
| "authors": [ |
| { |
| "first": "", |
| "middle": [], |
| "last": "Eir\u00edkur R\u00f6gnvaldsson", |
| "suffix": "" |
| } |
| ], |
| "year": 2005, |
| "venue": "Setningar. Handb\u00f3k um setningafrae\u00f0i. \u00cdslensk tunga III", |
| "volume": "", |
| "issue": "", |
| "pages": "602--635", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Eir\u00edkur R\u00f6gnvaldsson. 2005. Setningafrae\u00f0ilegar breytingar \u00ed \u00edslensku. In H\u00f6skuldur Thr\u00e1insson, editor, Setningar. Handb\u00f3k um setningafrae\u00f0i. \u00cdslensk tunga III, pages 602-635. Almenna b\u00f3kaf\u00e9lagi\u00f0, Reykjav\u00edk.", |
| "links": null |
| }, |
| "BIBREF21": { |
| "ref_id": "b21", |
| "title": "CUNI x-ling: Parsing under-resourced languages in CoNLL 2018 UD shared task", |
| "authors": [ |
| { |
| "first": "Rudolf", |
| "middle": [], |
| "last": "Rosa", |
| "suffix": "" |
| }, |
| { |
| "first": "David", |
| "middle": [], |
| "last": "Mare\u010dek", |
| "suffix": "" |
| } |
| ], |
| "year": 2018, |
| "venue": "Proceedings of the CoNLL 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies", |
| "volume": "", |
| "issue": "", |
| "pages": "187--196", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Rudolf Rosa and David Mare\u010dek. 2018. CUNI x-ling: Parsing under-resourced languages in CoNLL 2018 UD shared task. In Proceedings of the CoNLL 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies, pages 187-196, Brussels, Belgium. Association for Computational Linguistics.", |
| "links": null |
| }, |
| "BIBREF22": { |
| "ref_id": "b22", |
| "title": "Parsing Icelandic Al\u00feingi transcripts: Parliamentary speeches as a genre", |
| "authors": [ |
| { |
| "first": "Kristj\u00e1n", |
| "middle": [], |
| "last": "R\u00fanarsson", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Einar Freyr Sigur\u00f0sson", |
| "suffix": "" |
| } |
| ], |
| "year": 2020, |
| "venue": "Proceedings of the Second ParlaCLARIN Workshop", |
| "volume": "", |
| "issue": "", |
| "pages": "44--50", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Kristj\u00e1n R\u00fanarsson and Einar Freyr Sigur\u00f0sson. 2020. Parsing Icelandic Al\u00feingi transcripts: Parliamentary speeches as a genre. In Proceedings of the Second ParlaCLARIN Workshop, pages 44-50, Marseille, France. European Language Resources Association.", |
| "links": null |
| }, |
| "BIBREF23": { |
| "ref_id": "b23", |
| "title": "Risam\u00e1lheild: A Very Large Icelandic Text Corpus", |
| "authors": [ |
| { |
| "first": "Stein\u00fe\u00f3r", |
| "middle": [], |
| "last": "Steingr\u00edmsson", |
| "suffix": "" |
| }, |
| { |
| "first": "Sigr\u00fan", |
| "middle": [], |
| "last": "Helgad\u00f3ttir", |
| "suffix": "" |
| }, |
| { |
| "first": "Eir\u00edkur", |
| "middle": [], |
| "last": "R\u00f6gnvaldsson", |
| "suffix": "" |
| }, |
| { |
| "first": "Starka\u00f0ur", |
| "middle": [], |
| "last": "Barkarson", |
| "suffix": "" |
| }, |
| { |
| "first": "J\u00f3n", |
| "middle": [], |
| "last": "Gu\u00f0nason", |
| "suffix": "" |
| } |
| ], |
| "year": 2018, |
| "venue": "Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018)", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Stein\u00fe\u00f3r Steingr\u00edmsson, Sigr\u00fan Helgad\u00f3ttir, Eir\u00edkur R\u00f6gnvaldsson, Starka\u00f0ur Barkarson, and J\u00f3n Gu\u00f0nason. 2018. Risam\u00e1lheild: A Very Large Icelandic Text Corpus. In Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018), Miyazaki, Japan. European Language Resources Asso- ciation.", |
| "links": null |
| }, |
| "BIBREF24": { |
| "ref_id": "b24", |
| "title": "Augmenting a BiLSTM tagger with a morphological lexicon and a lexical category identification step", |
| "authors": [ |
| { |
| "first": "Stein\u00fe\u00f3r", |
| "middle": [], |
| "last": "Steingr\u00edmsson", |
| "suffix": "" |
| }, |
| { |
| "first": "\u00d6rvar", |
| "middle": [], |
| "last": "K\u00e1rason", |
| "suffix": "" |
| }, |
| { |
| "first": "Hrafn", |
| "middle": [], |
| "last": "Loftsson", |
| "suffix": "" |
| } |
| ], |
| "year": 2019, |
| "venue": "Proceedings of RANLP 2019", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Stein\u00fe\u00f3r Steingr\u00edmsson, \u00d6rvar K\u00e1rason, and Hrafn Loftsson. 2019. Augmenting a BiLSTM tagger with a mor- phological lexicon and a lexical category identification step. In Proceedings of RANLP 2019, Varna, Bulgaria.", |
| "links": null |
| }, |
| "BIBREF25": { |
| "ref_id": "b25", |
| "title": "Multi-source synthetic treebank creation for improved cross-lingual dependency parsing", |
| "authors": [ |
| { |
| "first": "Francis", |
| "middle": [ |
| "M" |
| ], |
| "last": "Tyers", |
| "suffix": "" |
| }, |
| { |
| "first": "Mariya", |
| "middle": [], |
| "last": "Sheyanova", |
| "suffix": "" |
| }, |
| { |
| "first": "Alexandra", |
| "middle": [], |
| "last": "Martynova", |
| "suffix": "" |
| }, |
| { |
| "first": "Pavel", |
| "middle": [], |
| "last": "Stepachev", |
| "suffix": "" |
| }, |
| { |
| "first": "Konstantin", |
| "middle": [], |
| "last": "Vinogradovsky", |
| "suffix": "" |
| } |
| ], |
| "year": 2018, |
| "venue": "Proceedings of the Second Workshop on Universal Dependencies", |
| "volume": "", |
| "issue": "", |
| "pages": "144--150", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Francis M. Tyers, Mariya Sheyanova, Alexandra Martynova, Pavel Stepachev, and Konstantin Vinogradovsky. 2018. Multi-source synthetic treebank creation for improved cross-lingual dependency parsing. In Proceedings of the Second Workshop on Universal Dependencies (UDW 2018), pages 144-150.", |
| "links": null |
| }, |
| "BIBREF26": { |
| "ref_id": "b26", |
| "title": "Anton Karl Ingason, Einar Freyr Sigur\u00f0sson, and Eir\u00edkur R\u00f6gnvaldsson", |
| "authors": [ |
| { |
| "first": "Joel", |
| "middle": [ |
| "C" |
| ], |
| "last": "Wallenberg", |
| "suffix": "" |
| } |
| ], |
| "year": 2011, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Joel C. Wallenberg, Anton Karl Ingason, Einar Freyr Sigur\u00f0sson, and Eir\u00edkur R\u00f6gnvaldsson. 2011. Icelandic Parsed Historical Corpus (IcePaHC). Version 0.9. http://www.linguist.is/icelandic_treebank.", |
| "links": null |
| }, |
| "BIBREF27": { |
| "ref_id": "b27", |
| "title": "Statistical Dependency Analysis with Support Vector Machines", |
| "authors": [ |
| { |
| "first": "Hiroyasu", |
| "middle": [], |
| "last": "Yamada", |
| "suffix": "" |
| }, |
| { |
| "first": "Yuji", |
| "middle": [], |
| "last": "Matsumoto", |
| "suffix": "" |
| } |
| ], |
| "year": 2003, |
| "venue": "The 8th International Workshop of Parsing Technologies (IWPT2003)", |
| "volume": "", |
| "issue": "", |
| "pages": "195--206", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Hiroyasu Yamada and Yuji Matsumoto. 2003. Statistical Dependency Analysis with Support Vector Machines. In The 8th International Workshop of Parsing Technologies (IWPT2003), pages 195-206, Nancy, France.", |
| "links": null |
| }, |
| "BIBREF28": { |
| "ref_id": "b28", |
| "title": "Morpheus: A neural network for jointly learning contextual lemmatization and morphological tagging", |
| "authors": [ |
| { |
| "first": "Eray", |
| "middle": [], |
| "last": "Yildiz", |
| "suffix": "" |
| }, |
| { |
| "first": "A. C\u00fcneyd", |
| "middle": [], |
| "last": "Tantug", |
| "suffix": "" |
| } |
| ], |
| "year": 2019, |
| "venue": "Proceedings of the 16th Workshop on Computational Research in Phonetics, Phonology, and Morphology", |
| "volume": "", |
| "issue": "", |
| "pages": "25--34", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Eray Yildiz and A. C\u00fcneyd Tantug. 2019. Morpheus: A neural network for jointly learning contextual lemmatiza- tion and morphological tagging. In Proceedings of the 16th Workshop on Computational Research in Phonetics, Phonology, and Morphology, pages 25-34, Florence, Italy, August. Association for Computational Linguistics.", |
| "links": null |
| } |
| }, |
| "ref_entries": { |
| "FIGREF0": { |
| "type_str": "figure", |
| "num": null, |
| "uris": null, |
| "text": "of the house paid him for carrying the water at certain times.' ( (IP-MAT (NP-SBJ (NS-N H\u00fasbaendur$-h\u00fasb\u00f3ndi) (D-N $nir-hinn)) (VBDI borgu\u00f0u-borga) (NP-OB2 (PRO-D honum-hann)) (NP-OB1 (N-A vatnsbur\u00f0$-vatnsbur\u00f0ur) (D-A $inn-hinn)) (PP (P \u00e1-\u00e1) (NP (ADJ-D vissum-viss) (NS-D t\u00edmum-t\u00edmi))) (. .-.)) (ID 1883.VOGGUR.NAR-FIC,.37))" |
| }, |
| "FIGREF1": { |
| "type_str": "figure", |
| "num": null, |
| "uris": null, |
| "text": "An example of the IcePaHC format." |
| }, |
| "FIGREF2": { |
| "type_str": "figure", |
| "num": null, |
| "uris": null, |
| "text": "An example of the output CoNLL-U format, taken from the Icelandic UD corpus." |
| }, |
| "FIGREF4": { |
| "type_str": "figure", |
| "num": null, |
| "uris": null, |
| "text": "-.)) (ID 1936.NTJOHN.REL-BIB,.1310)) An example of the FarPaHC format." |
| }, |
| "TABREF2": { |
| "html": null, |
| "content": "<table/>", |
| "text": "The UD tags used in the converted corpus.", |
| "type_str": "table", |
| "num": null |
| }, |
| "TABREF4": { |
| "html": null, |
| "content": "<table/>", |
| "text": "Dependency relations used in the converted corpus.", |
| "type_str": "table", |
| "num": null |
| } |
| } |
| } |
| } |
