Zdeněk Žabokrtský

TectoMT: modular NLP framework

In the present paper we describe TectoMT, a multi-purpose open-source NLP framework. It allows for fast and efficient development of NLP applications by exploiting a wide range of software modules already integrated in TectoMT, such as tools for sentence segmentation, tokenization, morphological analysis, POS tagging, shallow and deep syntax parsing, named entity recognition, anaphora resolution, tree-to-tree translation, natural language generation, word-level alignment of parallel corpora, and other tasks. One of the most complex applications of TectoMT is the English-Czech machine translation system with transfer on deep syntactic (tectogrammatical) layer. Several modules are available also for other languages (German, Russian, Arabic).Where possible, modules are implemented in a language-independent way, so they can be reused in many applications.

Arabic syntactic trees: from constituency to dependency

This research note reports on the work in progress which regards automatic transformation of phrase-structure syntactic trees of Arabic into dependency-driven analytical ones. Guidelines for these descriptions have been developed at the Linguistic Data Consortium, University of Pennsylvania, and at the Faculty of Mathematics and Physics and the Faculty of Arts, Charles University in Prague, respectively.The transformation consists of (i) a recursive function translating the topology of a phrase tree into a corresponding dependency tree, and (ii) a procedure assigning analytical functions to the nodes of the dependency tree.Apart from an outline of the annotation schemes and a deeper insight into these procedures, model application of the transformation is given herein.

Combining czech dependency parsers

In this paper we describe in detail two dependency parsing techniques developed and evaluated using the Prague Dependency Treebank 2.0. Then we propose two approaches for combining various existing parsers in order to obtain better accuracy. The highest parsing accuracy reported in this paper is 85.84 %, which represents 1.86 % improvement compared to the best single state-of-the-art parser. To our knowledge, no better result achieved on the same data has been published yet.

Tamil dependency parsing: results using rule based and corpus based approaches

Very few attempts have been reported in the literature on dependency parsing for Tamil. In this paper, we report results obtained for Tamil dependency parsing with rule-based and corpus-based approaches. We designed annotation scheme partially based on Prague Dependency Treebank (PDT) and manually annotated Tamil data (about 3000 words) with dependency relations. For corpus-based approach, we used two well known parsers MaltParser and MSTParser, and for the rule-based approach, we implemented series of linguistic rules (for resolving coordination, complementation, predicate identification and so on) to build dependency structure for Tamil sentences. Our initial results show that, both rule-based and corpus-based approaches achieved the accuracy of more than 74% for the unlabeled task and more than 65% for the labeled tasks. Rule-based parsing accuracy dropped considerably when the input was tagged automatically.

CzEng 0.9: Large Parallel Treebank with Rich Annotation

CzEng 0.9: Large Parallel Treebank with Rich Annotation We describe our ongoing efforts in collecting a Czech-English parallel corpus CzEng. The paper provides full details on the current version 0.9 and focuses on its new features: (1) data from new sources were added, most importantly a few hundred electronically available books, technical documentation and also some parallel web pages, (2) the full corpus has been automatically annotated up to the tectogrammatical layer (surface and deep syntactic analysis), (3) sentence segmentation has been refined, and (4) several heuristic filters to improve corpus quality were implemented. In total, we provide a sentence-aligned automatic parallel treebank of about 8.0 million sentences, 93 million English and 82 million Czech words. CzEng 0.9 is freely available for non-commercial research purposes.

Synthesis of czech sentences from tectogrammatical trees

In this paper we deal with a new rule-based approach to the Natural Language Generation problem. The presented system synthesizes Czech sentences from Czech tectogrammatical trees supplied by the Prague Dependency Treebank 2.0 (PDT 2.0). Linguistically relevant phenomena including valency, diathesis, condensation, agreement, word order, punctuation and vocalization have been studied and implemented in Peri using software tools shipped with PDT 2.0. BLEU score metric is used for the evaluation of the generated sentences.

Valency lexicon of czech verbs VALLEX: recent experiments with frame disambiguation

VALLEX is a linguistically annotated lexicon aiming at a description of syntactic information which is supposed to be useful for NLP. The lexicon contains roughly 2500 manually annotated Czech verbs with over 6000 valency frames (summer 2005). In this paper we introduce VALLEX and describe an experiment where VALLEX frames were assigned to 10,000 corpus instances of 100 Czech verbs – the pairwise inter-annotator agreement reaches 75%. The part of the data where three human annotators agreed were used for an automatic word sense disambiguation task, in which we achieved the precision of 78.5%.

Anaphora in czech: large data and experiments with automatic anaphora resolution

The aim of this paper is two-fold. First, we want to present a part of the annotation scheme of the Prague Dependency Treebank 2.0 related to the annotation of coreference on the tectogrammatical layer of sentence representation (more than 45,000 textual and grammatical coreference links in almost 50,000 manually annotated Czech sentences). Second, we report a new pronoun resolution system developed and tested using the treebank data, the success rate of which is 60.4 %.

Resolving noun phrase coreference in czech

In this work, we present first results on noun phrase coreference resolution on Czech data. As the data resource for our experiments, we employed yet unfinished and unpublished extension of Prague Dependency Treebank 2.0, which captures noun phrase coreference and bridging relations. Incompleteness of the data influenced one of our motivations --- to aid annotators with automatic pre-annotation of the data. Although we introduced several novel tree features and tried different machine learning approaches, results on a growing amount of data shows that the selected feature set and learning methods are not able to sufficiently exploit the data.

Morphological meanings in the prague dependency treebank 2.0

In this paper we report our work on the system of grammatemes (mostly semantically-oriented counterparts of morphological categories such as number, degree of comparison, or tense), the concept of which was introduced in Functional Generative Description, and is now further elaborated in the context of Prague Dependency Treebank 2.0. We present also a new hierarchical typology of tectogrammatical nodes.