Tēzaurs.lv – the experience of building a multifunctional lexical resource

Abstract

In this paper, we describe our findings from developing the lexicographic platform Tēzaurs.lv, extending it from a traditional explanatory dictionary into a multifunctional resource for structured lexical data. Tēzaurs.lv is the largest Latvian dictionary with more than 390,000 entries, which emerged as a compilation from nearly 300 prior dictionaries and other sources. Recently, it has been extended with Latvian WordNet data, effectively making it also a synonym dictionary and a translation dictionary. Each entry can contain multiple lexemes with their grammatical information and inflection tables, enabling search on inflection forms and spelling variants. For the new requirements, we have developed a lexical database system and a collaborative online editor toolkit, which are also used for two other major Latvian dictionaries. While previously the data model and tools were based on what the end user would see in a dictionary entry, the current infrastructure is designed with a highly structured lexical data model. This avoids duplication and helps to ensure consistency if entries or word senses are edited or merged, and it supports the usage of this data in computational linguistics.