The article describes the Corpus of Spoken Lithuanian, its structure, compilation stages (collection of the recordings, transcription, and grammatical annotation), and the methodology of data collection and digitalization; in addition, it discusses the possibilities of corpus application in the research of natural language usage and the research, which has already been carried out, using the corpus data.
At present (2017), the corpus, which is freely accessible for internet users, contains 226,174 word forms. The users of the online corpus version can perform search of a word or a word form and obtain data on the frequency of the form in the whole corpus or its part as well as see grammatical information about it.
In 2016-2017, the Corpus of Spoken Lithuanian was supplemented by new data resulting from the implementation of the project “Contemporary Spoken Lithuanian: A Corpus-based Analysis of Grammar and Lexis” (LIP-085/2016) financed by the Research Council of Lithuania under the programme of the State Lithuanian Studies and Dissemination Programme for 2016–2024. The project will also create a new internet access, which will provide more possibilities for the users. The updated corpus consists of 256 conversations (383,587 words) produced by 1,086 speakers (659 females and 427 males), whose age ranges from 3 to 81 years. When developing the Corpus of Spoken Lithuanian, much attention was paid to its composition, i.e. the proportions of the corpus. In order to improve the universality and suitability of the corpus for a more varied analysis, the principle of a balanced corpus was maintained; therefore, several criteria were taken into consideration when collecting the data: the nature of spoken language (private vs public speech) and structure (dialogues vs polilogues), different communication situations (direct vs indirect (e.g. a telephone conversation), demographic indicators, and social relations among the interlocutors. Therefore, in 2018, users of the updated version of the corpus will be able to filter results according to different categories, such as gender, age, place and structure of the conversation, and perform a more detailed search. It is expected that when the users are provided with more possibilities to analyse corpus data on the internet, the amount of spoken language research will increase comprising different areas of lexis and grammar.
Spoken-language research at Carnegie Mellon
