The tendency for speakers to align or accommodate their verbal and non-verbal behaviour to their interlocutors is a fundamental mechanism in spoken interaction, strongly associated with successful communication and social bonding. Despite its ubiquity and documentation across various modalities and linguistic levels (e.g., lexical, prosodic), a lack of comparable, multi-layered linguistic resources and methodological agreement prevents a deeper understanding of its cognitive mechanisms. Multidimensional view of speech alignment might enhance its application in areas like language training or human-machine interaction. This paper addresses these gaps by presenting the development of a multilingual corpus of L1 Slovak and L2 English speech, extending a comparable corpus in L1 English. The corpus utilizes a modified cooperative board game, Forbidden Island, to elicit semi-spontaneous, multi-party conversation and introduces a complementary pair game to specifically target and prime syntactic alignment. The resource includes psychological metadata (e.g., personality, anxiety, perceived dominance) and enables a reproducible methodology for investigating the relationship between entrainment patterns and individual characteristics. By providing a non-Germanic language perspective and a direct L1–L2 comparison framework at prosodic, lexical, pragmatic and syntactic levels, this corpus offers a rich resource for advancing the theoretical understanding, replication, and practical application of speech alignment.

Presence of appropriate acoustic cues of affective features in the synthesized speech can be a prerequisite for the proper evaluation of the semantic content by the message recipient. In the recent work the authors have focused on the research of expressive speech synthesis capable of generating naturally sounding synthetic speech at various levels of arousal. The synthesizer should be able to produce speech in Slovak in different styles from extremely urgent warnings, insisting messages, alerts, through comments, and neutral style speech to soothing messages and very calm speech. A three-step method was used for recording both - the high-activation and low-activation expressive speech databases. The acoustic properties of the obtained databases are discussed. Several synthesizers with different levels of arousal were designed using these databases and their outputs are compared to the original voice of the voice talent. A possible ambiguity of acoustic cues is pointed out and the relevance of the semantic meaning of the sentences both in the sentence set for the speech database recording and in the set for subjective synthesizer testing is discussed.