Lemmatization

With increasingly higher numbers of non-English language web searchers the problems of efficient handling of non-English Web documents and user queries are becoming major issues for search engines. The main aim of this review paper is to make researchers aware of the existing problems in monolingual non-English Web retrieval by providing an overview of open issues. A significant number of

La production de corpus d'occitan médiéval et prémoderne: problèmes et perspectives de travail Ce dépôt contient les fichiers et modèles décrits dans: Jean-Baptiste Camps et Gilles Guilhem Couffignal, « La production de corpus d'occitan médiéval et prémoderne : problèmes et perspectives de travail », <em>Actes du XIIe Congrès International de l'Association Internationale d'Études Occitanes, Albi 2017</em>, à paraître. Il contient: <code>htr_data</code>: les données du Chansonnier I utilisées pour l'entraînement de modèles d'HTR; <code>models_CLSTM</code>: le meilleur modèle obtenu avec OCRopy/CLSTM; <code>models_pandora</code>: le modèle de lemmatisation obtenu avec Pandora; <code>tsv</code>: les données d'entraînement et de test pour la lemmatisation, au format .tsv; <code>xml</code>: les données de <em>Montferrand</em> au format XML. ainsi que le texte de l'article. ...

En este artículo presentamos una serie de técnicas de Procesamiento de Lenguaje Natural aplicadas a la normalización de términos en Recuperación de Información Textual. El objetivo de dichas técnicas es el tratamiento de los fenómenos de variación lingüística morfológica y léxica. En concreto explorará la utilización de la lematización, su empleo combinado con el stemming y la expansión de consultas mediante umbrales de sinonimia.

This research focuses on the implementation of Gramatika, a grammar checker designed for the Filipino language given its available resources and linguistic tools. The checker uses hybrid n-grams generated from n-grams of words, part-of-speech tags, and lemmas of grammatically-correct texts. It covers a variety of error types including those unique in Filipino: wrong word form, and incorrectly merged / unmerged words. The grammar checker performed 64% accuracy on producing the correct suggestions on erroneous phrases and 85% on error-free texts when using Part-of-Speech (POS) tags from a Hybrid POS tagger (HPOST) for Filipino. Recommendations to improve Gramatika is to implement linguistic tools such as constituency parser, incorrect affix detection system, and a spell checker for the Filipino language.

This project sets out to discover and develop techniques for the lemmatisation of a historical corpus of the Cornish language in order that a lemmatised dictionary macrostructure can be generated from the corpus. The system should be capable of uniquely identifying every lexical item that is attested in the corpus. A survey of published and unpublished Cornish dictionaries, glossaries and lexicographical notes was carried out. A corpus was compiled incorporating specially prepared new critical editions. An investigation into the history of Cornish lemmatisation was undertaken. A systemic description of Cornish inflection was written. Three methods of corpus lemmatisation were trialed. Findings were as follows. Lexicographical history shapes current Cornish lexicographical practice. Lexicon based tokenisation has advantages over character based tokenisation. System networks provide the means to generate base forms from attested word types. Grammatical difference is the most reliable way of disambiguating homographs. A lemma that contains three fields, the canonical form, the part-of-speech and a semantic field label, provides of a unique code for every lexeme attested in the corpus. Programs which involve human interaction during the lemmatisation process allow bootstrapping of the lemmatisation database. Computerised morphological processing may be used at least to partially create the lemmatisation database. Disambiguation of at least some of the most common homographs may be automated by the use of computer programs.

Aplikasi pemeriksa ejaan (spelling checker) merupakan sebuah tool yang dapat mendeteksi kesalahan penulisan ejaan pada suatu kata atau teks. Aplikasi pemeriksa ejaan untuk bahasa Indonesia pada umumnya memeriksa dengan cara membandingkan satu persatu kata yang diperiksa pada suatu kamus kata yang berisi semua kata pada bahasa Indonesia. Selain itu, aplikasi pemeriksa ejaan yang umum digunakan hanya dapat menangani kesalahan tipografi (Misalnya ‘bagaimna’, seharunya ‘bagaimana’) dan belum dapat menangani kesalahan akibat pemisahan atau penggabungan kata yang tidak tepat (Misalnya ‘bagai mana’, seharusnya ‘bagaimana’). Pada tugas akhir ini dibangun sebuah aplikasi pemeriksa ejaan untuk tulisan ilmiah berbahasa indonesia dengan memanfaatkan penggunaan analisis morfologi dan aturan penulisan pemisahan atau penggabungan kata dalam proses pemeriksaan ejaan. Analisis morfologi dibuat dengan menghilangkan imbuhan yang memungkinkan pada suatu kata dengan menggunakan algoritme lemmatization. Modifikasi dan desain yang dibuat untuk proses lemmatization adalah dengan menambahkan aturan penulisan kata ganti orang (ku- dan kau-) dan menyimpan semua aturan penghilangan imbuhan pada suatu database. Aplikasi pemeriksa ejaan yang telah dibuat dalam tugas akhir ini masih belum sempurna dari segi keakuratan pemeriksaan. Hasil pemeriksaan masih mengalami ketidakakuratan untuk kasus kata berulang, kasus pengecualian penulisan imbuhan, dan kasus kata gabung dan bentuk terikat yang memiliki imbuhan.

The paper presents the system used in the EvaLatin shared task to POS tag and lemmatize Latin. It consists of two components. A gradient boosting machine (LightGBM) is used for POS tagging, mainly fed with pre-computed word embeddings of a window of seven contiguous tokens—the token at hand plus the three preceding and following ones—per target feature value. Word embeddings are trained on the texts of the Perseus Digital Library, Patrologia Latina, and Biblioteca Digitale di Testi Tardo Antichi, which together comprise a high number of texts of different genres from the Classical Age to Late Antiquity. Word forms plus the outputted POS labels are used to feed a Seq2Seq algorithm implemented in Keras to predict lemmas. The final shared-task accuracies measured for Classical Latin texts are in line with state-of-the-art POS taggers (∼96%) and lemmatizers (∼95%).

Traditionally, Zulu adjectives have been lemmatized under their stems only. In this research article, an in-depth analysis is undertaken to make a case for the lemmatization of all frequent adjectival forms with their adjective concords rather. It is shown that the supposed explosion in size of the dictionary may be contained within a corpus-driven Sinclairian framework. The advantages of such a word-like treatment far outnumber the generalizations that have hitherto characterized the lexicographic treatment of adjectives in Zulu. The study is supported by ample dictionary extracts from a Zulu–English dictionary project aimed at junior users. Comparisons with existing dictionaries and textbook data are also made.

The authors of this article firmly believe in the advantages of utilising a corpus for lemma-sign list creation. However, one should not overreact and assume that alternative methods for the creation of a dictionary’s macrostructure have no virtues, or that alternative methods are in principle per definition marred by inconsistencies. What is called for is a perspective on corpus-based activities versus intuition-based compilations by lexicographers. Therefore, while the supremacy of a corpus remains undisputed in compiling a lemma-sign list, this article also intends to show that a well-planned combination of a variety of existing lists that were assembled manually, results in a lemma-sign list with a remarkable internal consistency. Hence, the aim of this article is twofold. Besides a brief illustration of typical macrostructural inconsistencies, the main focus will be on a series of consistencies encountered in the compilation of lemma-sign lists for different sub-dictionaries at various stages of the Sepedi Dictionary Project (SeDiPro). Some attention will also be devoted to the so-called Miraculous Consistency Ratio ‘(x 1.25)4 = x 2.4’ – being a sequence of four 25% increases which result from a collation of five manually compiled Sepedi lemma-sign lists.

SUMMARY

This study examines L2 French learner corpora for adjective complexity through the data in The Newcastle Corpus (Myles & Mitchell, 2016). Data suggest that over the span of one year student usage patterns become more native-like by all metrics analyzed, providing pedagogical implications for corpus-informed pedagogy and data-driven learning.

Keywords: Corpus Linguistics, French, Adjectives, Syntax, Pedagogy, Genre, Learner Corpora, Oral Corpora

ABSTRACT
This study examines L2 French learner corpora for usage patterns in adjective complexity through the data in The Newcastle Corpus, French Language Learner Oral Corpora (FLLOC) (Myles & Mitchell, 2016).

French adjective variation in prenominal and postnominal placement has been notoriously difficult to explain through prescriptive rules (Sleeman et al., 2014) and variation is oft attributed to semantic shifts (Alexiadou et al., 2007; Laenzlinger, 2005), usage contexts (Delbecque, 1990) and user preference (Thuilier, 2013). Recently, large-scale corpus research of native speaker production suggests that variation is more linked to genre and speaker preference than has been previous understood (Thuilier, 2013). Empirical data indicate that French adjectives in the noun phrase (NP) are significantly impacted in nominal placement by adjective lemma frequency and syllable length (Delbecque, 1990; Thuilier, 2013; Wilmet, 1980).

These native speaker metrics provide a framework to consider the progression of L2 French learning through corpus data (Filipovic & Hawkins, 2013; Shirato & Stapleton, 2007; Wulff & Gries, 2015). The comparison of data to native speaker norms is further facilitated by data of the Newcastle Corpus, which provides native and learner speaker productions for elicited language tasks (Myles & Mitchell, 2016). Thus, learner progress is analyzed through the complexity of adjectives in the NP as explored through lemma frequency, lemma syllable length, and nominal placement variation.

Data suggest that over the span of a year student usage patterns become more native-like by all metrics. More advanced students use greater placement variation and lemmas of longer syllable length than their counterparts. Additionally, lemma frequency and nominal placement are also more closely aligned with native speaker usage patterns. These data display the lexical and syntactic understanding acquired by L2 French students and have pedagogical implications for corpus-informed pedagogy and data-driven learning.

(1) Introduction
(2) Corpora and the compilation of the lemma-sign list
(3) Corpora and the battle against inconsistencies
(4) Corpora as an aid for conjunctively written languages
(5) Corpora as the key to writing better dictionary articles
(6) Corpora as an aid to sense distinction for the writing of better definitions (monolingual dictionaries) or for the selection of suitable translation equivalents (bilingual dictionaries)
(7) Corpora as an aid to finding typical collocations, idioms and proverbs
(8) Corpora as an aid in pinpointing clusters and choosing better examples of usage
(9) Integration of different corpus query tools in order to enhance the quality of data presentation in the dictionary
(10) Conclusion
References

This  paper  deals  with  the  impact  of  complex  morphological  structures  on  essential  aspects  of lexicology. On the basis of data from the Kartvelian (South-Caucasian) language family consisting of Georgian and its sister-languages, it discusses questions of how to lemmatize nominal and verbal word forms so that they can be arranged in a consistent and appropriate way in dictionaries and databases.  The  most  prominent  approaches  that  have  been  developed  since  the  first  attempt  to provide  a  dictionary  of  Georgian  (Paolini  &  Irbachi  1629)  are  analysed  with  a  view  to  their applicability, usability, and conclusiveness.

The aim of this article is (a) to reflect on the contributions made by P.S. Groenewald to the field of lexicography in South Africa, focusing on the importance of determining the relative frequency of individual words in Sesotho sa Leboa, (b) to indicate how this initial basic need for knowing the frequency of use of words gradually grew and motivated the creation of text corpora, first for Sesotho sa Leboa, later for all official South African languages, (c) to illustrate how frequency counts and keywords in context can be used to improve dictionary compilation at the macrostructural and microstructural level, and (d) to utilise frequency counts for a novel cutting edge dimension, namely to craft a multidimensional Ruler for the compilation of ‘major Sesotho sa Leboa dictionaries’ – such dictionaries being the Groenewald dream.

In Zulu, there are three kinds of quantitatives: inclusive, exclusive and numeral. For the lemmatization of these, even existing traditional dictionaries felt the need to move away from a pure 'stem' approach towards a 'word' approach. In a new Zulu–English dictionary project, this is not only confirmed, but is taken one step further with particular attention to the microstructure.

This short article presents the result of accuracy tests for currently available Ancient Greek lemmatizers and recently published lemmatized corpora. We ran a blinded experiment in which three highly proficient readers of Ancient Greek evaluated the output of the CLTK lemmatizer, of the CLTK backoff lemmatizer, and of GLEM, together with the lemmatizations offered by the Diorisis corpus and the Lemmatized Ancient Greek Texts repository. The texts chosen for this experiment are Homer, Iliad 1.1–279 and Lysias 7. The results suggest that lemmatization methods using large lexica as well as part-of-speech tagging—such as those employed by the Diorisis corpus and the CLTK backoff lemmatizer—are more reliable than methods that rely more heavily on machine learning and use smaller lexica.

This paper deals with the impact of complex morphological structures on essential aspects of lexicology. On the basis of data from the Kartvelian (South-Caucasian) language family consisting of Georgian and its sister-languages, it discusses questions of how to lemmatize nominal and verbal word forms so that they can be arranged in a consistent and appropriate way in dictionaries and databases. The most prominent approaches that have been developed since the first attempt to provide a dictionary of Georgian (Paolini & Irbachi 1629) are analysed with a view to their applicability, usability, and conclusiveness.

This paper explores the problem of developing NLP tools for morphologically rich and orthographi-cally inconsistent classical languages. It is a case study of building a lemmatizer for Old Irish using only a dictionary and an unlabeled corpus as sources of data. At the current stage, the lemmatizer shows 76.31% average recall score on a corpus of ca. 100,000 tokens and is able to predict lemmas for out-of-vocabulary words. However, as it is the work in progress, the lemmatizer lacks some functionality such as disambiguation. There is no gold standard to measure accuracy yet either.

Cet aide-mémoire résume l'essentiel du travail effectué lors de la journée CeLiSo (EA7332) « Créer soi-même un corpus étiqueté d'un million de mots en anglais, français ou allemand » à l'Université Paris IV-Sorbonne le 25/03/2016. Il s'adresse en priorité aux doctorants et enseignants-chercheurs qui ont participé à cette journée, ainsi qu'à toutes les autres personnes (a priori plutôt des novices) qui souhaitent apprendre à construire leur propre corpus (en anglais " DIY corpus ").

[FR - English below]
À l’heure où la quantité de données disponibles, plus ou moins librement, s’accroît de manière importante, grâce aux corpus, éditions ou bibliothèques numériques, le développement d’outils de fouille de données ou de méthodes d’apprentissage profond permet au chercheur de se constituer un corpus d’étude adapté à ses recherches, d’enrichir ses données et des les exploiter. Des outils ouverts de reconnaissance optique des caractères peuvent être adaptés à un imprimé ancien, un incunable, voire un manuscrit, avec des résultats exploitables, autorisant la constitution rapide de corpus textuels. L’alternance de phases d’entraînement et de correction permet de faire progresser la qualité des résultats, en accumulant rapidement des données textuelles brutes. Celles-ci peuvent ensuite être structurées, par exemple en xml/tei, et enrichies. L’enrichissement par des annotations graphiques ou linguistiques connaît également des automatisations. Ces procédés, connus des linguistes et fonctionnels pour les langues modernes, posent des difficultés pour des langues comme l’occitan médiéval, dues en partie à l’absence de corpus lemmatisés conséquents. Des pistes pour la création d’outils adaptés à la grande variabilité graphique des états anciens de langue, seront présentées, ainsi que des expérimentations pour la lemmatisation de l’occitan médiéval et prémoderne. Ces techniques ouvrent la porte à de nombreuses exploitations. L’augmentation, tant souhaitée, de la quantité de textes et données de qualité disponibles, permet le progrès des méthodes de philologie numérique, si tant est que chacun prenne la peine de rendre ses données librement disponibles en ligne et réutilisables. Par l’exposition de différentes solutions techniques et de quelques micro-analyses à titre d’exemple, cette communication entend montrer une partie de ce que la philologie numérique peut offrir au chercheur en domaine occitan, tout en rappelant les enjeux éthiques sur lesquels reposent de telles pratiques.

[EN]
At a time when the quantity of - more or less freely - available data is increasing significantly, thanks to digital corpora, editions or libraries, the development of data mining tools or deep learning methods allows researchers to build a corpus of study tailored for their research, to enrich their data and to exploit them. Open optical character recognition (OCR) tools can be adapted to old prints, incunabula or even manuscripts, with usable results, allowing the rapid creation of textual corpora. The alternation of training and correction phases makes it possible to improve the quality of the results by rapidly accumulating raw text data. These can then be structured, for example in XML/TEI, and enriched. The enrichment of the texts with graphic or linguistic annotations can also be automated. These processes, known to linguists and functional for modern languages, present difficulties for languages such as Medieval Occitan, due in part to the absence of big enough lemmatized corpora. Suggestions for the creation of tools adapted to the considerable spelling variation of ancient languages will be presented, as well as experiments for the lemmatization of Medieval and Premodern Occitan. These techniques open the way for many exploitations. The much desired increase in the amount of available quality texts and data makes it possible to improve digital philology methods, if everyone takes the trouble to make their data freely available online and reusable. By exposing different technical solutions and some micro-analyses as examples, this paper aims to show part of what digital philology can offer to researchers in the Occitan domain, while recalling the ethical issues on which such practices are based.

Lemmatization is a process of finding the base morphological form (lemma) of a word. It is an important step in many natural language processing, information retrieval, and information extraction tasks, among others. We present an open-source language-independent lemmatizer based on the Random Forest classification model. This model is a supervised machine-learning algorithm with decision trees that are constructed corresponding to the grammatical features of the language. This lemmatizer does not require any manual work for hard-coding of the rules, and at the same time, it is simple and interpretable. We compare the performance of our lemmatizer with that of the UDPipe lemmatizer on twenty-two out of twenty-five languages we work on for which UDPipe has models. Our lemmatization method shows good performance in different languages from various language groups, and it is easily extensible to other languages. The source code of our lemmatizer is publicly available.

Sentiment classification of texts written in Serbian is still an under-researched topic. One of the open issues is how the different forms of morphological normalization affect the performances of different sentiment classifiers and which normalization procedure is optimal for this task. In this paper we assess and compare the impact of lemmatizers and stemmers for Serbian on classifiers trained and evaluated on the Serbian Movie Review Dataset.

(1) Introduction
(2) Lemma-sign lists in dictionaries for the elementary level
(3) Compiling the lemma-sign list of the Junior Dictionary
(4) Comparison of the compiled lemma-sign list with the manually excerpted vocabulary
(5) The final lemma-sign list for the Junior Dictionary and onwards
(6) References

REVIEW. Fragments, (Studies in Iconology, 14), Leuven-Walpole, 2018 (Fragments is co-edited by Stephanie Heremans and is a limited edition and celebratory publication signed by the author). This 400 pages book, consists of 110 lemmata, 111 color images, and indices nominum & subjects)

The aim of this article is to analyze traditional approaches to the lemmatization of nouns on the macrostructural level in African languages against the background of the user- perspective, the physical limitations on volume, the consideration of currently available dictionaries and the utilization of a corpus. Five basic lemmatization strategies are given due attention, namely lemmatizing nouns under stems, according to morpho-lexical fields, under both singular and plural forms, solely under singular forms, or finally on first or third letters. It is shown that any modern strategy that aims at avoiding the shortcomings and pitfalls of these five basic types whilst at the same time exploiting their virtues, will have to give full weight to target users' desires. A means to incorporate those desires is illustrated through a small dictionary project to which the concept of ‘simultaneous feedback’ is applied. In conclusion four rules of thumb and ten general guidelines are presented.

We present a new mixed method lemmatizer for Icelandic, Lemmald, which achieves good performance by relying on IceTagger [1] for tagging and The Icelandic Frequency Dictionary [2] corpus for training. We combine the advantages of data-driven machine learning with linguistic insights to maximize performance. To achieve this, we make use of a novel approach: Hierarchy of Linguistic Identities (HOLI), which involves organizing features and feature structures for the machine learning based on linguistic knowledge. Accuracy of the lemmatization is further improved using an add-on which connects to the Database of Modern Icelandic Inflections [3]. Given correct tagging, our system lemmatizes Icelandic text with an accuracy of 99.55%. We believe our method can be fruitfully adapted to other morphologically rich languages.

We describe and compare two tools for processing Middle Russian texts. Both tools provide lemmatization, part-of-speech and morphological annotation. One (“RNC”) was developed for annotating texts in the Russian National Corpus and is rule-based. The other one (“TOROT”) is being used for annotating the eponymous corpus and is statistical. We apply the two analyzers to the same Middle Russian text and then compare their outputs with high-quality manual annotation. Since the analyzers use different annotation schemes and spelling principles, we have to harmonize their outputs before we can compare them. The comparison shows that TOROT performs considerably better than RNC (lemmatization 69.8% vs. 47.3%, part of speech 89.5% vs. 54.2%, morphology 81.5% vs. 16.7%). If, however, we limit the evaluation set only to those tokens for which the analyzers provide a guess and in addition consider the RNC response correct if one of the multiple guesses it provides is correct, the numbers become comparable (88.5% vs. 91.9%, 93.9% vs. 95.2%, 81.5% vs. 86.8%). We develop a simple procedure which boosts TOROT lemmatization accuracy by 8.7% by using RNC lemma guesses when TOROT fails to provide one and matching them against the existing TOROT lemma database. We conclude that a statistical analyzer (trained on a large material) can deal with non-standardised historical texts better than a rule-based one. Still, it is possible to make the analyzers collaborate, boosting the performance of the superior one.

(1) Defining a dictionary’s macrostructure
(2) On the need for rulers, part 1
(3) Part-of-Speech rulers (‘POS Rulers’)
(4) On the need for rulers, part 2
(5) Multidimensional lexicographic rulers
(6) Characterising POS Rulers and multidimensional lexicographic rulers
(7) Integrating rulers with dictionary compilation software
(8) Selected bibliography

In this article a four-step methodology is proposed for the creation of the lemma-sign list of a Nguni-language reference work. The theoretical principles are illustrated throughout with a full-scale case study revolving around isiNdebele. For the suggested approach raw corpus data is utilised, and only standard, straightforward and widely-available software tools are required to process the data. Apart from the inherent value of having an entire macrostructure at one's disposal right from the start of a dictionary project, it is shown how such a list can also be used for both predictions and measurements on lemma-sign, page and time levels. As such, drawing up the macrostructure of a dictionary automatically leads to a 'ruler' with which the entire lexicographic process can successfully be monitored. Specifically for isiNdebele, suggestions are made for the way ahead.

An open issue in the sentiment classification of texts written in Serbian is the effect of different forms of morphological normalization and the usefulness of leveraging large amounts of unlabeled texts. In this paper, we assess the impact of lemmatizers and stemmers for Serbian on classifiers trained and evaluated on the Serbian Movie Review Dataset. We also consider the effectiveness of using word embeddings, generated from a large unlabeled corpus, as classification features.

Resumen En este art culo se describe el sistema ERIAL, llevado a cabo en el marco del proyecto del mismo nombre, para Recuperaci on de Informaci on. Tras una primera descripci on externa del proyecto (Secci on 1), se presenta el entorno modular LEIRA desarrollado con tal n (Secci on 2) y, a continuaci on (Secci on 3), se describen en detalle tanto los recursos ling u sticos integrados en los m odulos computacionalmente diferenciados en el entorno en cuesti on como los m odulos en s mismos. En la Secci on 4 ...

In this article, we consider the problem of supervised morphological analysis using an approach that differs from industry spread analogs. The article describes a new method of lemmatization based on the algorithms of machine learning, in particular, on the algorithms of regression analysis, trained on the open grammatical dictionary of Russian language. Comparison of obtained results was performed with existing alternative applications that are used nowadays for addressing lemmatization problems in NLP problems for Russian language. The proposed method shows some potential for further development as it has comparable quality but uses relatively simple machine learning algorithm and at the same time is not rule based involving no manual work. The source code for our lemmatizer is publicly available.

The research group L.A.S.L.A. (Laboratoire d’Analyse Statistique des Langues Anciennes, University of Liege, Belgium) began in 1961 a project of lemmatization and morphosyntactic tagging of Latin texts. This project continues with new texts lemmatized each year (see http://web.philo.ulg.ac.be/lasla/ ). The resulting files, which contain approximatively 2,500,000 words, whose lemmatization and tagging have been verified by a philologist, have recently been made available to interested scholars. In the early 2000s, Collatinus was developed by Yves Ouvrard for teaching. Its goal was to generate a complete lexical aid, with a short translation and the morphological analyses of the forms, for any text that can be given to the students (see https://outils.biblissima.fr/fr/collatinus/ ). Although these two projects look very different, they met a few years ago in the conception of a new tool to speed up the lemmatization process of Latin texts at L.A.S.L.A. This tool is based on a concurre...

c○Springer-Verlag Abstract. In this our first participation in CLEF, we have applied Natural Language Processing techniques for single word and multiword term conflation. We have tested several approaches at different levels of text processing in our experiments: firstly, we have lemmatized the text to avoid inflectional variation; secondly, we have expanded the queries through synonyms according to a fixed threshold of similarity; thirdly, we have employed morphological families to deal with derivational variation; and fourthly, we have tested a mixed approach based on the employment of such families and syntactic dependencies to deal with the syntactic content of the document. 1