The goal of this thesis is to create a dependency treebank for Yorùbá, a language with very little pre-existing machine-readable resources. The treebank follows the Universal Dependencies (UD) annotation standard, certain language-specific guidelines for Yorùbá were specified. Known techniques for porting resources from resource-rich languages were tested, in particular projection of annotation across parallel bilingual data. Manual annotation is not the main focus of this thesis; nevertheless, a small portion of the data was verified manually in order to evaluate the annotation quality. Also, a model was trained on the manual annotation using UDPipe.

This licentiate thesis deals with automatic syntactic analysis, or parsing, of natural languages. A parser constructs the syntactic analysis, which it learns by looking at correctly analyzed sentences, known as training data. The general topic concerns manipulations of the training data in order to improve the parsing accuracy. Several studies using constituency-based theories for natural languages in such automatic and data-driven syntactic parsing have shown that training data, annotated according to a linguistic theory, often needs to be adapted in various ways in order to achieve an adequate, automatic analysis. A linguistically sound constituent structure is not necessarily well-suited for learning and parsing using existing data-driven methods. Modifications to the constituency-based trees in the training data, and corresponding modifications to the parser output, have successfully been applied to increase the parser accuracy. The topic of this thesis is to investigate whether similar modifications in the form of tree transformations to training data, annotated with dependency-based structures, can improve accuracy for data-driven dependency parsers. In order to do this, two types of tree transformations are in focus in this thesis. %This is a topic that so far has been less studied. The first one concerns non-projectivity. The full potential of dependency parsing can only be realized if non-projective constructions are allowed, which pose a problem for projective dependency parsers. On the other hand, non-projective parsers tend, among other things, to be slower. In order to maintain the benefits of projective parsing, a tree transformation technique to recover non-projectivity while using a projective parser is presented here. The second type of transformation concerns linguistic phenomena that are possible but hard for a parser to learn, given a certain choice of dependency analysis. This study has concentrated on two such phenomena, coordination and verb groups, for which tree transformations are applied in order to improve parsing accuracy, in case the original structure does not coincide with a structure that is easy to learn. Empirical evaluations are performed using treebank data from various languages, and using more than one dependency parser. The results show that the benefit of these tree transformations used in preprocessing and postprocessing to a large extent is language, treebank and parser independent.

This licentiate thesis deals with automatic syntactic analysis, or parsing, of natural languages. A parser constructs the syntactic analysis, which it learns by looking at correctly analyzed sentences, known as training data. The general topic concerns manipulations of the training data in order to improve the parsing accuracy. Several studies using constituency-based theories for natural languages in such automatic and data-driven syntactic parsing have shown that training data, annotated according to a linguistic theory, often needs to be adapted in various ways in order to achieve an adequate, automatic analysis. A linguistically sound constituent structure is not necessarily well-suited for learning and parsing using existing data-driven methods. Modifications to the constituency-based trees in the training data, and corresponding modifications to the parser output, have successfully been applied to increase the parser accuracy. The topic of this thesis is to investigate whether similar modifications in the form of tree transformations to training data, annotated with dependency-based structures, can improve accuracy for data-driven dependency parsers. In order to do this, two types of tree transformations are in focus in this thesis. The first one concerns non-projectivity. The full potential of dependency parsing can only be realized if non-projective constructions are allowed, which pose a problem for projective dependency parsers. On the other hand, non-projective parsers tend, among other things, to be slower. In order to maintain the benefits of projective parsing, a tree transformation technique to recover non-projectivity while using a projective parser is presented here. The second type of transformation concerns linguistic phenomena that are possible but hard for a parser to learn, given a certain choice of dependency analysis. This study has concentrated on two such phenomena, coordination and verb groups, for which tree transformations are applied in order to improve parsing accuracy, in case the original structure does not coincide with a structure that is easy to learn. Empirical evaluations are performed using treebank data from various languages, and using more than one dependency parser. The results show that the benefit of these tree transformations used in preprocessing and postprocessing to a large extent is language, treebank and parser independent.

<div>Structured prediction makes coherent decisions as structured objects to present the interrelations of these predicted variables. They have been widely used in many areas, such as bioinformatics, computer vision, speech recognition, and natural language processing. Machine Learning with reduced supervision aims to leverage the laborious and error-prone annotation effects and benefit the low-resource languages. In this dissertation we study structured prediction with reduced supervision for two sets of problems, sequence labeling and dependency parsing, both of which are representatives of structured prediction problems in NLP. We investigate three different approaches.</div><div> </div><div>The first approach is learning with modular architecture by task decomposition. By decomposing the labels into location sub-label and type sub-label, we designed neural modules to tackle these sub-labels respectively, with an additional module to infuse the information. The experiments on the benchmark datasets show the modular architecture outperforms existing models and can make use of partially labeled data together with fully labeled data to improve on the performance of using fully labeled data alone.</div><div><br></div><div>The second approach builds the neural CRF autoencoder (NCRFAE) model that combines a discriminative component and a generative component for semi-supervised sequence labeling. The model has a unified structure of shared parameters, using different loss functions for labeled and unlabeled data. We developed a variant of the EM algorithm for optimizing the model with tractable inference. The experiments on several languages in the POS tagging task show the model outperforms existing systems in both supervised and semi-supervised setup.</div><div><br></div><div>The third approach builds two models for semi-supervised dependency parsing, namely local autoencoding parser (LAP) and global autoencoding parser (GAP). LAP assumes the chain-structured sentence has a latent representation and uses this representation to construct the dependency tree, while GAP treats the dependency tree itself as a latent variable. Both models have unified structures for sentence with and without annotated parse tree. The experiments on several languages show both parsers can use unlabeled sentences to improve on the performance with labeled sentences alone, and LAP is faster while GAP outperforms existing models.</div>

Automatic syntactic analysis is essential for extracting useful information from large-scale learner data for linguistic research and natural language processing (NLP). Currently, researchers use standard POS taggers and parsers developed on native language to analyze learner language. Investigation of how such systems perform on learner data is needed to develop strategies for minimizing the cross-domain effects. Furthermore, POS taggers and parsers are developed for generic NLP purposes and may not be useful for identifying specific syntactic constructs such as subcategorization frames (SCFs). SCFs have attracted much research attention as they provide unique insight into the interplay between lexical and structural information. An automatic SCF identification system adapted for learner language is needed to facilitate research on L2 SCFs. In this thesis, we first provide a comprehensive evaluation of standard POS taggers and parsers on learner and native English. We show that the common practice of constructing a gold standard by manually correcting the output of a system can introduce bias to the evaluation, and we suggest a method to control for the bias. We also quantitatively evaluate the impact of fine-grained learner errors on POS tagging and parsing, identifying the most influential learner errors. Furthermore, we show that the performance of probabilistic POS taggers and parsers on native English can predict their performance on learner English. Secondly, we develop an SCF identification system for learner English. We train a machine learning model on both native and learner English data. The system can label individual verb occurrences in learner data for a set of 49 distinct SCFs. Our evaluation shows that the system reaches an accuracy of 84\% F1 score. We then demonstrate that the level of accuracy is adequate for linguistic research. We design the first multidimensional SCF diversity metrics and investigate how SCF diversity changes with L2 proficiency on a large learner corpus. Our results show that as L2 proficiency develops, learners tend to use more diverse SCF types with greater taxonomic distance; more advanced learners also use different SCF types more evenly and locate the verb tokens of the same SCF type further away from each other. Furthermore, we demonstrate that the proposed SCF diversity metrics contribute a unique perspective to the prediction of L2 proficiency beyond existing syntactic complexity metrics.

We present a novel approach to training data-driven dependency parsers on incomplete annotations. Our parsers are simple modifications of two well-known dependency parsers, the transition-based Malt parser and the graph-based MST parser. While previous work on parsing with incomplete data has typically couched the task in frameworks of unsupervised or semi-supervised machine learning, we essentially treat it as a supervised problem. In particular, we propose what we call agnostic parsers which hide all fragmentation in the training data from their supervised components. We present experimental results with training data that was obtained by means of annotation projection. Annotation projection is a resource-lean technique which allows us to transfer annotations from one language to another within a parallel corpus. However, the output tends to be noisy and incomplete due to cross-lingual non-parallelism and error-prone word alignments. This makes the projected annotations a suitable test bed for our fragment parsers. Our results show that (i) dependency parsers trained on large amounts of projected annotations achieve higher accuracy than the direct projections, and that (ii) our agnostic fragment parsers perform roughly on a par with the original parsers which are trained only on strictly filtered, complete trees. Finally, (iii) when our fragment parsers are trained on artificially fragmented but otherwise gold standard dependencies, the performance loss is moderate even with up to 50% of all edges removed. / Wir präsentieren eine neuartige Herangehensweise an das Trainieren von daten-gesteuerten Dependenzparsern auf unvollständigen Annotationen. Unsere Parser sind einfache Varianten von zwei bekannten Dependenzparsern, nämlich des transitions-basierten Malt-Parsers sowie des graph-basierten MST-Parsers. Während frühere Arbeiten zum Parsing mit unvollständigen Daten die Aufgabe meist in Frameworks für unüberwachtes oder schwach überwachtes maschinelles Lernen gebettet haben, behandeln wir sie im Wesentlichen mit überwachten Lernverfahren. Insbesondere schlagen wir "agnostische" Parser vor, die jegliche Fragmentierung der Trainingsdaten vor ihren daten-gesteuerten Lernkomponenten verbergen. Wir stellen Versuchsergebnisse mit Trainingsdaten vor, die mithilfe von Annotationsprojektion gewonnen wurden. Annotationsprojektion ist ein Verfahren, das es uns erlaubt, innerhalb eines Parallelkorpus Annotationen von einer Sprache auf eine andere zu übertragen. Bedingt durch begrenzten crosslingualen Parallelismus und fehleranfällige Wortalinierung ist die Ausgabe des Projektionsschrittes jedoch üblicherweise verrauscht und unvollständig. Gerade dies macht projizierte Annotationen zu einer angemessenen Testumgebung für unsere fragment-fähigen Parser. Unsere Ergebnisse belegen, dass (i) Dependenzparser, die auf großen Mengen von projizierten Annotationen trainiert wurden, größere Genauigkeit erzielen als die zugrundeliegenden direkten Projektionen, und dass (ii) die Genauigkeit unserer agnostischen, fragment-fähigen Parser der Genauigkeit der Originalparser (trainiert auf streng gefilterten, komplett projizierten Bäumen) annähernd gleichgestellt ist. Schließlich zeigen wir mit künstlich fragmentierten Gold-Standard-Daten, dass (iii) der Verlust an Genauigkeit selbst dann bescheiden bleibt, wenn bis zu 50% aller Kanten in den Trainingsdaten fehlen.

This thesis investigates the role of linguistically-motivated generative models of syntax and semantic structure in natural language processing (NLP). Syntactic well-formedness is crucial in language generation, but most statistical models do not account for the hierarchical structure of sentences. Many applications exhibiting natural language understanding rely on structured semantic representations to enable querying, inference and reasoning. Yet most semantic parsers produce domain-specific or inadequately expressive representations. We propose a series of generative transition-based models for dependency syntax which can be applied as both parsers and language models while being amenable to supervised or unsupervised learning. Two models are based on Markov assumptions commonly made in NLP: The first is a Bayesian model with hierarchical smoothing, the second is parameterised by feed-forward neural networks. The Bayesian model enables careful analysis of the structure of the conditioning contexts required for generative parsers, but the neural network is more accurate. As a language model the syntactic neural model outperforms both the Bayesian model and n-gram neural networks, pointing to the complementary nature of distributed and structured representations for syntactic prediction. We propose approximate inference methods based on particle filtering. The third model is parameterised by recurrent neural networks (RNNs), dropping the Markov assumptions. Exact inference with dynamic programming is made tractable here by simplifying the structure of the conditioning contexts. We then shift the focus to semantics and propose models for parsing sentences to labelled semantic graphs. We introduce a transition-based parser which incrementally predicts graph nodes (predicates) and edges (arguments). This approach is contrasted against predicting top-down graph traversals. RNNs and pointer networks are key components in approaching graph parsing as an incremental prediction problem. The RNN architecture is augmented to condition the model explicitly on the transition system configuration. We develop a robust parser for Minimal Recursion Semantics, a linguistically-expressive framework for compositional semantics which has previously been parsed only with grammar-based approaches. Our parser is much faster than the grammar-based model, while the same approach improves the accuracy of neural Abstract Meaning Representation parsing.

Discovering the structure of natural language sentences by semi-supervised methods Rudolf Rosa In this thesis, we focus on the problem of automatically syntactically ana- lyzing a language for which there is no syntactically annotated training data. We explore several methods for cross-lingual transfer of syntactic as well as morphological annotation, ultimately based on utilization of bilingual or multi- lingual sentence-aligned corpora and machine translation approaches. We pay particular attention to automatic estimation of the appropriateness of a source language for the analysis of a given target language, devising a novel measure based on the similarity of part-of-speech sequences frequent in the languages. The effectiveness of the presented methods has been confirmed by experiments conducted both by us as well as independently by other respectable researchers. 1

Analyse probabiliste est l'un des domaines de recherche les plus attractives en langage naturel En traitement. Analyseurs probabilistes succès actuels nécessitent de grandes treebanks qui Il est difficile, prend du temps et coûteux à produire. Par conséquent, nous avons concentré notre l'attention sur des approches moins supervisés. Nous avons proposé deux catégories de solution: l'apprentissage actif et l'algorithme semi-supervisé. Stratégies d'apprentissage actives permettent de sélectionner les échantillons les plus informatives pour annotation. La plupart des stratégies d'apprentissage actives existantes pour l'analyse reposent sur la sélection phrases incertaines pour l'annotation. Nous montrons dans notre recherche, sur quatre différents langues (français, anglais, persan, arabe), que la sélection des phrases complètes ne sont pas une solution optimale et de proposer un moyen de sélectionner uniquement les sous-parties de phrases. Comme nos expériences ont montré, certaines parties des phrases ne contiennent aucune utiles information pour la formation d'un analyseur, et en se concentrant sur les sous-parties incertains des phrases est une solution plus efficace dans l'apprentissage actif. / Probabilistic parsing is one of the most attractive research areas in natural language processing. Current successful probabilistic parsers require large treebanks which are difficult, time consuming, and expensive to produce. Therefore, we focused our attention on less-supervised approaches. We suggested two categories of solution: active learning and semi-supervised algorithm. Active learning strategies allow one to select the most informative samples for annotation. Most existing active learning strategies for parsing rely on selecting uncertain sentences for annotation. We show in our research, on four different languages (French, English, Persian, and Arabic), that selecting full sentences is not an optimal solution and propose a way to select only subparts of sentences. As our experiments have shown, some parts of the sentences do not contain any useful information for training a parser, and focusing on uncertain subparts of the sentences is a more effective solution in active learning.

The process of determining the structural relationships between words in both natural and machine languages is known as parsing. Parsers are used as core components in a number of Natural Language Processing (NLP) applications such as online tutoring applications, dialogue-based systems and textual entailment systems. They have been used widely in the development of machine languages. In order to understand the way parsers work, we will investigate and describe a number of widely used parsing algorithms. These algorithms have been utilised in a range of different contexts such as dependency frameworks and phrase structure frameworks. We will investigate and describe some of the fundamental aspects of each of these frameworks, which can function in various ways including grammar-driven approaches and data-driven approaches. Grammar-driven approaches use a set of grammatical rules for determining the syntactic structures of sentences during parsing. Data-driven approaches use a set of parsed data to generate a parse model which is used for guiding the parser during the processing of new sentences. A number of state-of-the-art parsers have been developed that use such frameworks and approaches. We will briefly highlight some of these in this thesis. There are three specific important features that it is important to integrate into the development of parsers. These are efficiency, accuracy, and robustness. Efficiency is concerned with the use of as little time and computing resources as possible when processing natural language text. Accuracy involves maximising the correctness of the analyses that a parser produces. Robustness is a measure of a parser’s ability to cope with grammatically complex sentences and produce analyses of a large proportion of a set of sentences. In this thesis, we present a parser that can efficiently, accurately, and robustly parse a set of natural language sentences. Additionally, the implementation of the parser presented here allows for some trading-off between different levels of parsing performance. For example, some NLP applications may emphasise efficiency/robustness over accuracy while some other NLP systems may require a greater focus on accuracy. In dialogue-based systems, it may be preferable to produce a correct grammatical analysis of a question, rather than incorrectly analysing the grammatical structure of a question or quickly producing a grammatically incorrect answer for a question. Alternatively, it may be desirable that document translation systems translate a document into a different language quickly but less accurately, rather than slowly but highly accurately, because users may be able to correct grammatically incorrect sentences manually if necessary. The parser presented here is based on data-driven approaches but we will allow for the application of constraint rules to it in order to improve its performance.