This study focuses on automatic analysis of annotated transcribed speech. The annotation system considered has been recently introduced to address the several limitations of classical syntactic annotations when faced to natural speech transcriptions. It introduces many different components such as embedding, piles, kernels, pre-kernels, discursive markers etc.. All those components are tightly coupled in a complex tree structure and can hardly be considered separately because of their close intrication. Hence, a joint analysis is required but no analysis tool to handle them all together was available yet. In this study, we introduce such an automatic parser of annotated transcriptions of speech and present the corresponding framework based on multi-trees. This framework permits to jointly handle separate aspects of speech such as macro and micro syntactic levels, which are traditionnaly considered separately. Several applications are proposed, including analysis of the transcribed speech by classical parsers designed for written language
