This paper introduces a new bilingual Czech-English verbal valency lexicon (called CzEngVallex) representing a relatively large empirical database. It includes 20,835 aligned valency frame pairs (i.e., verb senses which are translations of each other) and their aligned arguments. This new lexicon uses data from the Prague Czech-English Dependency Treebank and also takes advantage of the existing valency lexicons for both languages: the PDT-Vallex for Czech and the EngVallex for English. The CzEngVallex is available for browsing as well as for download in the LINDAT/CLARIN repository.The CzEngVallex is meant to be used not only by traditional linguists, lexicographers, translators but also by computational linguists both for the purposes of enriching theoretical linguistic accounts of verbal valency from a cross-linguistic perspective and for an innovative use in various NLP tasks.

Event structure: taxonomy and semantic roles CategoriesDecompositions obey a certain  -different verb classes have different decomposition formats (DFs): all verbs of the same category have the same DF.Verbs of Action are characterized by the following configuration of components: (1) K4. Activity | X acted with the Goal in mind K6. Causation | this caused K8. Result | new state came about & holds at the MS. This configuration is present in the decomposition of such verbs as vyteret' 'wipe', razrezat' 'cut ', vystirat' 'wash', postroit' 'build', pokrasit' 'paint ', svarit' 'boil ', vykopat' 'dig out' etc. There are different kinds of actions. Their decomposition formats differ from one another. But configuration (1) is present in all formats for actions. Thematic classesCategory components constitute the   of the decomposition. Thematic components are inserted in different places of the category frame. If we replace, e.g., the concrete state sleep -by its natural hyperonym   we are able to identify razbudit' as a verb belonging to the thematic class  verbs. For vyteret' 1.2 'wipe' its thematic class  is substantiated by the following configuration:(2) K0. Initial state| the (functional) state of Y was not normal /desirable K8. Result | the (functional) state of Y is normal /desirable. Other verbs of treatment -žarit' 'stew', varit' 'boil', gladit', 'iron'. Decompositions provide a semantic basis both for category and thematic classification of verbs. Meaning shifts-how can they be presented as operations on LSDs. 3.3.1. Deagentivization, a   (3) a. Ivan razbudil menja grubym pinkom [razbudil 'woke up' -action] Ivan NOM wake PAST me ACC rude INSTR kick INSTR 'Ivan woke me up with a rude kick.' b. Zvonok v dver' razbudil menja [razbudil 'woke up' -happening] ringing NOM in door wake PAST me ACC 'The ringing of the doorbell woke me up.' 9 PBML 92 DECEMBER 2009 Templates (#3а) and (#3b) below present two abbreviated LSDs of the verb razbudit' (corresponding to its different lexemes; T-category of the lexeme and thematic classes of the participants are given in brackets; components in parenthesis are optional). (#3а) X razbudil Y [action : ordinary] = K0. Initial state | before t < MS Y was in a state: Y slept K4. Activity | at t < MS X acted with the Goal in mind [X is a person] K5. (Manner of action | acted upon Y: applying Z) K6. Causation | this was causing [causation as a process] / caused [causation as event] K7. (Process in Object | synchronous; telic) K8. Result | new state of Y came about & holds at the MS: Y does not sleep K9, K10. Entailment, Implication |-(#3b) X razbudil Y [happening] = K0. Initial state | before t < MS Y was in a state: Y slept K4. Causer | X took place [X is an event] K5. (Manner of action |-) K6. Causation | this caused [causation as event] K8. Effect | new state of Y came about & holds at the MS: Y does not sleep K9. Entailment |-K10. Implication | this is bad for YThe differenc...

This paper presents a resource and the associated annotation process used in a project of interlinking Czech and English verbal translational equivalents based on a parallel, richly annotated dependency treebank containing also valency and semantic roles, namely the Prague Czech-English Dependency Treebank. One of the main aims of this project is to create a high-quality and relatively large empirical base which could be used both for linguistic comparative research as well as for natural language processing applications, such as machine translation or cross-language sense disambiguation. This paper describes the resulting lexicon, CzEngVallex, and the process of building it, as well some interesting observations and statistics already obtained.

The aim of this paper is to introduce the Czech subjectivity lexicon 1 , a new lexical resource for sentiment analysis in Czech. We describe particular stages of the manual refinement of the lexicon and demonstrate its use in the state-of-the art polarity classifiers, namely the Maximum Entropy classifier. We test the success rate of the system enriched with the dictionary on different data sets, compare the results and suggest some further improvements of the lexicon-based classification system.

Abstract:In the paper, we present our efforts to annotate evaluative language in the Prague Dependency Treebank 2.0. The project is a follow-up of the series of annotations of small plaintext corpora. It uses automatic identification of potentially evaluative nodes through mapping a Czech subjectivity lexicon to syntactically annotated data. These nodes are then manually checked by an annotator and either dismissed as standing in a non-evaluative context, or confirmed as evaluative. In the latter case, information about the polarity orientation, the source and target of evaluation is added by the annotator. The annotations unveiled several advantages and disadvantages of the chosen framework. The advantages involve more structured and easy-to-handle environment for the annotator, visibility of syntactic patterning of the evaluative state, effective solving of discontinuous structures or a new perspective on the influence of good/bad news. The disadvantages include little capability of treating cases with evaluation spread among more syntactically connected nodes at once, little capability of treating metaphorical expressions, or disregarding the effects of negation and intensification in the current scheme.