Svetla Koeva

Summarizing Short Texts Through a Discourse-Centered Approach in a Multilingual Context

The chapter presents the architecture of a system targeting summaries of short texts in six languages. At the core of a summary, which comprises clauses and sentences extracted from the original text, is the structure of the discourse and its relationship with its coreferential links. The approach shows a uniform design for all languages, while language specificity is attributed to the resources that fuel the component modules. The design described here includes a number of feedback loops used to fine-tune the parameters by comparing the output of the modules against annotated corpora. “Average” summaries over some human-produced ones are used to evaluate the accuracy of each of the monolingual systems. The study also presents some quantitative data on the corpora used, showing a comparison among languages and results that, mostly, prove to be above the state of the art.

Multi-word Term Extraction for Bulgarian

The goal of this paper is to compile a method for multi-word term extraction, taking into account both the linguistic properties of Bulgarian terms and their statistical rates. The method relies on the extraction of term candidates matching given syntactic patterns followed by statistical (by means of Log-likelihood ratio) and linguistically (by means of inflectional clustering) based filtering aimed at improving the coverage and the precision of multi-word term extraction.

Ontology-Supported Text Classification Based on Cross-Lingual Word Sense Disambiguation

The paper reports on recent experiments in cross-lingual document processing (with a case study for Bulgarian-English-Romanian language pairs) and brings evidence on the benefits of using linguistic ontologies for achieving, with a high level of accuracy, difficult tasks in NLP such as word alignment, word sense disambiguation, document classification, cross-language information retrieval, etc. We provide brief descriptions of the parallel corpus we used, the multilingual lexical ontology which supports our research, the word alignment and word sense disambiguation systems we developed and a preliminary report on an ongoing development of a system for cross-lingual text-classification which takes advantage of these multilingual technologies. Unlike the keyword-based methods in document processing, the concept-based methods are supposed to better exploit the semantic information contained in a particular document and thus to provide more accurate results.

Semantic and syntactic patterns of multiword names: A cross-language study

Named entities (NEs) constitute a great challenge for computational linguistics and one of the major research topics during the last decade. They can be divided in categories describing people, location, time, organization and others. In this paper we will restrict our discussion to proper names that belong to three main classes: personal, location and organization names, and that can be either single-word nouns or multiword expressions. First, we are going to define common (language-independent) semantic patterns for proper names and then we will present the corresponding syntactic patterns in English, Bulgarian, French, Greek, and Serbian. We will compare these patterns regarding grammatical categories of dependent constituents, definiteness, distribution of clitics, word order and various alternations. Our ultimate goal is to build a universal framework for Named Entity Recognition (NER).

Detecting Gaps in Language Resources and Tools in the Project CESAR

In this paper the first preliminary results of the analysis of marks collected within the tables of META-NET series of Language White Papers of CESAR project languages are demonstrated. Although they are preliminary results, we can consider them useful for showing us where real gaps in language resources and tools can be detected.