Text-To-Speech (TTS) evaluations are habitually carried out without contextual and situational framing. Since humans adapt their speaking style to situation specific communicative needs, such evaluations may not generalize across situations. Without clearly defined framing, it is even unclear in which situations evaluation results hold at all. We test the hypothesized impact of framing on TTS evaluation in a crowdsourced MOS evaluation of four TTS voices, systematically varying (a) the intended TTS task (domestic humanoid robot, child's voice replacement, fiction audio books and long and information-rich texts) and (b) the framing of that task. The results show that framing differentiated MOS responses, with individual TTS performance varying significantly across tasks and framings. This corroborates the assumption that decontextualized MOS evaluations do not generalize, and suggests that TTS evaluations should not be reported without the type of framing that was employed, if any.

We introduce continuous phonological features as input to TTS with the dual objective of more precise control over phonological aspects and better potential for exploration of latent features in TTS models for speech science purposes. In our framework, the TTS is conditioned on continuous values between 0.0 and 1.0, where each phoneme has a specified position on each feature axis. We chose 11 features to represent US English and trained a voice with Matcha-TTS. Effectiveness was assessed by investigating two selected features in two ways: through a categorical perception experiment confirming the expected alignment of feature positions and phoneme perception, and through analysis of acoustic correlates confirming a gradual, monotonic change of acoustic features consistent with changes in the phonemic input features.

Nonhuman great apes have been claimed to be unable to learn human words due to a lack of the necessary neural circuitry. We recovered original footage of two enculturated chimpanzees uttering the word “mama” and subjected recordings to phonetic analysis. Our analyses demonstrate that chimpanzees are capable of syllabic production, achieving consonant-to-vowel phonetic contrasts via the simultaneous recruitment and coupling of voice, jaw and lips. In an online experiment, human listeners naive to the recordings’ origins reliably perceived chimpanzee utterances as syllabic utterances, primarily as “ma-ma”, among foil syllables. Our findings demonstrate that in the absence of direct data-driven examination, great ape vocal production capacities have been underestimated. Chimpanzees possess the neural building blocks necessary for speech.

Neural text-to-speech synthesis (TTS) captures prosodicfeatures strikingly well, notwithstanding the lack of prosodiclabels in training or synthesis. We trained a voice on a singleSwedish speaker reading in Swedish and English. The resultingTTS allows us to control the degree of English-accentedness inSwedish sentences. English-accented Swedish commonlyexhibits well-known prosodic characteristics such as erroneoustonal accents and understated or missed durational differences.TTS quality was verified in three ways. Automatic speechrecognition resulted in low errors, verifying intelligibility.Automatic language classification had Swedish as the majoritychoice, while the likelihood of English increased with ourtargeted degree of English-accentedness. Finally, a rank ofperceived English-accentedness acquired through pairwisecomparisons by 20 human listeners demonstrated a strongcorrelation with the targeted English-accentedness.We report on phonetic and prosodic analyses of theaccented TTS. In addition to the anticipated segmentaldifferences, the analyses revealed temporal and prominencerelated variations coherent with Swedish spoken by Englishspeakers, such as missing Swedish stress patterns and overlyreduced unstressed syllables. With this work, we aim to gleaninsights into speech prosody from the latent prosodic featuresof neural TTS models. In addition, it will help implementspeech phenomena such as code switching in TTS

In order to investigate the strengths and weaknesses of Audience Response System (ARS) in text-to-speech synthesis (TTS) evaluations, we revisit three previously published TTS studies and perform an ARS-based evaluation on the stimuli used in each study. The experiments are performed with a participant pool of 39 respondents, using a web-based tool that emulates an ARS experiment. The results of the first experiment confirms that ARS is highly useful for evaluating long and continuous stimuli, particularly if we wish for a diagnostic result rather than a single overall metric, while the second and third experiments highlight weaknesses in ARS with unsuitable materials as well as the importance of framing and instruction when conducting ARS-based evaluation.

We introduce the MEET corpus. The corpus was collected with the aim of systematically studying the effects of collocated (physical), remote (digital) and hybrid work meetings on collaborative decision-making. It consists of 10 sessions, where each session contains three recordings: a collocated, a remote and a hybrid meeting between three participants. The participants are working on a different survival ranking task during each meeting. The duration of each meeting ranges from 10 to 18 minutes, resulting in 380 minutes of conversation altogether. We also present the annotation scheme designed specifically to target our research questions. The recordings are currently being transcribed and annotated in accordance with this scheme.

Text-to-speech synthesis based on deep neuralnetworks can generate highly humanlike speech,which revitalizes the potential for analysis-bysynthesis in speech research. We propose that neuralsynthesis can provide evidence that a specificdistinction in its transcription system represents arobust acoustic/phonetic distinction in the speechused to train the model.We synthesized utterances with allophones inincorrect contexts and analyzed the resultsphonetically. Our assumption was that if we gainedcontrol over the allophonic variation in this way, itwould provide strong evidence that the variation isgoverned robustly by the phonological context usedto create the transcriptions.Of three allophonic variations investigated, thefirst, which was believed to be quite robust, gave usrobust control over the variation, while the other two,which are less categorical, did not afford us suchcontrol. These findings are consistent with ourhypothesis and support the notion that neural TTS canbe a valuable analysis-by-synthesis tool for speechresearch. 

We conduct two crowdsourcing experiments designed to examine the usefulness of audio cocktails to quickly find out information on the contents of large audio data. Several thousand crowd workers were engaged to listen to audio cocktails with systematically varied composition. They were then asked to state either which sound out of four categories (Children, Women, Men, Orchestra) they heard the most of, or if they heard anything of a specific category at all. The results show that their responses have high reliability and provide information as to whether a specific task can be performed using audio cocktails. We also propose that the combination of crowd workers and audio cocktails can be used directly as a tool to investigate the contents of large audio data.

The tongue is one of the organs most central to human speech. Here, the evolution and species-unique properties of the human tongue is traced, via reference to the apparent articulatory behavior of extant non-human great apes, and fossil findings from early hominids - from a point of view of articulatory phonetics, the science of human speech production. Increased lingual flexibility provided the possibility of mapping of articulatory targets, possibly via exaptation of manual-gestural mapping capacities evident in extant great apes. The emergence of the human-specific tongue, its properties, and morphology were crucial to the evolution of human articulate speech.

Swedish speech and language technology (LT) research goes back over 70 years. This has paid off: there is a national research infrastructure, as well as significant research projects, and Swedish is well-endowed with language resources (LRs) and tools. However, there are gaps that need to be filled, especially high-quality goldstandard LRs required by the most recent deep-learning methods. In the future, we would like to see closer collaborations and communication between the “traditional” LT research community and the burgeoning AI field, the establishment of dedicated academic LT training programmes, and national funding for LT research.