Adam Radziszewski

A Tiered CRF Tagger for Polish

In this paper we present a new approach to morphosyntactic tagging of Polish by bringing together Conditional Random Fields and tiered tagging. Our proposal also allows to take advantage of a rich set of morphological features, which resort to an external morphological analyser. The proposed algorithm is implemented as a tagger for Polish. Evaluation of the tagger shows significant improvement in tagging accuracy on two state-of-the-art taggers, namely PANTERA and WMBT.

WCCL: a morpho-syntactic feature toolkit

The paper presentsWCCL, a new formalism and toolkit for constructing morpho-syntactic features, a crucial task for many natural language processing algorithms. One existing solution, JOSKIPI, is analysed from two perspectives: features of the formalism as well as software engineering-related issues. Then we propose its successor. A short case study follows, exemplifying the improvement enabled by using rich features expressed with WCCL. The formalism is targeted at Polish, although it seems well suited for any inflectional language.

Large-Scale Experiments with NP Chunking of Polish

The published experiments with shallow parsing for Slavic languages are characterised with small size of the corpora used. With the publication of the National Corpus of Polish (NCP), a new opportunity was opened: to test several chunking algorithms on the 1-million token manually annotated subcorpus of the NCP. We test three Machine Learning techniques: Decision Tree induction, Memory-Based Learning and Conditional Random Fields. We also investigate the influence of tagging errors on the overall chunker performance, which happens to be quite substantial.

Taggers Gonna Tag: An Argument against Evaluating Disambiguation Capacities of Morphosyntactic Taggers

Usually tagging of inflectional languages is performed in two stages: morphological analysis and morphosyntactic disambiguation. A number of papers have been published where the evaluation is limited to the second part, without asking the question of what a tagger is supposed to do. In this article we highlight this important question and discuss possible answers. We also argue that a fair evaluation requires assessment of the whole system, which is very rarely the case in the literature. Finally we show results of the full evaluation of three Polish morphosyntactic taggers. The discrepancy between our results and those published earlier is striking, showing that these issues do make a practical difference.

Fextor: A Feature Extraction Framework for Natural Language Processing: A Case Study in Word Sense Disambiguation, Relation Recognition and Anaphora Resolution

Feature extraction from text corpora is an important step in Natural Language Processing (NLP), especially for Machine Learning (ML) techniques. Various NLP tasks have many common steps, e.g. low level act of reading a corpus and obtaining text windows from it. Some high-level processing steps might also be shared, e.g. testing for morpho-syntactic constraints between words. An integrated feature extraction framework removes wasteful redundancy and helps in rapid prototyping.

Using Low-Cost Annotation to Train a Reliable Czech Shallow Parser

Bushbank is a relatively new concept — a type of annotated corpus where annotation is driven by use of automatic tools and the task of human annotators is limited to accepting or rejecting parts of their output. This creates a possibility to obtain annotated corpora of considerable size at relatively low cost.

Parsing with Agreement

Shallow parsing has been proposed as a means of arriving at practically useful structures while avoiding the difficulties of full syntactic analysis. According to Abneys principles, it is preferred to leave an ambiguity pending than to make a likely wrong decision. We show that continuous phrase chunking as well as shallow constituency parsing display evident drawbacks when faced with freer word order languages. Those drawbacks may lead to unnecessary data loss as a result of decisions forced by the formalism and therefore diminish practical value of shallow parsers for Slavic languages. We present an alternate approach to shallow parsing of noun phrases for Slavic languages which follows the original Abneys principles. The proposed approach to parsing is decomposed into several stages, some of which allow for marking discontinuous phrases.

Morphosyntactic Constraints in the Acquisition of Linguistic Knowledge for Polish

Many approaches to the construction of language tools and acquisition of linguistic knowledge from corpora assume the application of some robust shallow parser. Construction of such a parser is difficult in the case of inflective languages with relaxed word order like Polish. The goal of the work presented here is to analyse the extent of knowledge that can be expressed in the form of morphosyntactic constraints referring to morphological properties of word forms, and its applications in the automatic extraction of syntactic and semantic knowledge. Basic properties of an extended version of the language of morphosyntactic constraints called JOSKIPI are briefly presented. The application of morphosyntactic constraints as background knowledge for extraction of disambiguation rules for Polish is discussed. A new approach to extraction of lexical semantic relations is presented: it relies on the constraints in identifying lexico-morphosyntactic dependencies among word forms in the text. Finally, a combination of the constraints and statistical analysis in the acquisition of multiword expressions is outlined.

Classification of Predicate-Argument Relations in Polish Data

This paper discusses the problem of syntactic relation recognition in Polish data. We consider subject, object and copula relations between VP and NP or AdjP chunks. The problem has been studied for English, while it has received very little attention in the context of Slavic languages. Slavic languages, including Polish, are characterised with relatively free word order, which makes the task more challenging than in the case of English.

WCCL Match – A Language for Text Annotation

In this paper we present a formalism for text annotation called WCCL Match. The need for a new formalism originates from our works related to Question Answering for Polish. We examined several existing formalisms to conclude that none of them fulfills our requirements. The new formalism was designed on top of an existing language for writing morphosyntactic functional expressions, namely WCCL. The major features of WCCL Match are: creation of new annotations, modification of existing ones, support for overlapping annotations, explicit access to tagset attributes and referring to context outside of captured annotation. We discuss three applications of the formalism: recognition of proper names, question analysis and question-to-query transformation. The implementation of WCCL Match is language-independent and can be used for almost any natural language.