Языки мира: языки манде
This talk provides an analysis for the functioning of the temporal distance in past category which has emerged in Gban (South Mande). The category consists of three subcategories: hodiernal (“today’s past”; default interpretation ‘(earlier) today’), hesternal (“yesterday’s past”; default interpretation ‘yesterday’) and remote (default interpretation ‘the day before yesterday or earlier’). When influenced by a temporal location adverbial modifier, the interpretation of sentences with temporal distance forms changes in a nontrivial way.
Mainstream approaches to the typology of reported discourse have
been based on the notion of a direct-indirect continuum: reported speech constructions
are traditionally analyzed as conforming to or deviating from the
“ideals” of European direct and indirect speech. This study argues that continuum-
based approaches fail to distinguish between two dimensions of variation
that are systematically discriminated in a number of African languages and
should therefore be treated separately. First, different constructions can be
recruited for speech reporting, ranging from paratactic to subordinate structures.
Second, languages differ in the way pronouns in speech reports are interpreted.
In European languages two different deictic strategies are associated with different
syntactic types of speech report (‘indirect’ and ‘direct’ deixis is correlated
with subordination and parataxis, respectively). In Kakabe, we argue, the choice
of pronominal values is independent of the construction’s syntax. Dissociating
the construction’s structural properties from the behavior of indexicals allows us
to describe the Kakabe strategies of speech reporting as well as account for the
seemingly puzzling behavior of reported commands. Our data shows that speech
reporting strategies of Kakabe should be treated as a type in its own right: a type
characterized by loose syntax and flexible pronominal indexicality.
An annotated corpus of Guinean Maninka, Corpus Maninka de Référence (CMR), was published in April 2016. It includes two subcorpora: one contains texts originally written in Latin-based graphics (792,778 words), and the other one is composed of texts in N'ko alphabet (3,105,879 words). Both subcorpora are searchable in both Latin-based graphics and in N'ko. In the building CMR, the Daba software package (earlier developed for the Corpus Bambara de Référence) has been used. As the search tool, NoSketchEngine has been used, it was adapted to the right-to-left direction of the N'ko writing. All texts in N'ko were obtained in electronic format, most of them were converted from pre-Unicode fonts. The morphological annotation is based on the Malidaba electronic dictionary which is in an intermediary stage of compillation; much efforts is needed to bring it to a minimally acceptable state.
This article provides a brief overview of Daba software package created in the course of building corpora for Manding languages. Key software features are motivated by the tasks and problems characteristic of many African languages. The corpus-building model proposed here was initially developed for Bambara Reference Corpus which is available online and is freely accessible. The morphological analysis procedure and corpus annotation scheme are discussed in detail. Daba uses a morpheme-based morphological annotation scheme inspired by the interlinear glossed form of presentation of linguistic examples. A scheme mapping Daba’s morpheme-based morphological information onto traditional word-based corpus annotation is provided. Since Bambara is characterized by a low level of written language standardization special attention is paid to the issues of representing variability in corpus annotation.