Spontaneous Speech Events in Two Speech Databases of Human-Computer and Human-Human Dialogs in Spanish

Abstract

Previous works in English have revealed that disfluencies follow regular patterns and that incorporating them into the language model of a speech recognizer leads to lower perplexities and sometimes to a better performance. Although work on disfluency modeling has been applied outside the English community (e.g., in Japanese), as far as we know there is no specific work dealing with disfluencies in Spanish. In this paper, we follow a data driven approach in exploring the potential benefit of modeling disfluencies in a speech recognizer in Spanish. Two databases of human-computer and human-human dialogs are considered, which allow the absolute and relative frequencies of disfluencies in the two situations to be compared. The rate of disfluencies in human-human dialogs is found to be very close to that found for similar databases in English. Due to setup factors, the rate of disfluencies found in human-computer dialogs was remarkably higher than that reported for similar databases in English. In any case, from the point of view of speech recognition, the high frequencies of disfluencies and the distinct features of the acoustic events related to them support the need for explicit acoustic models. The regularities observed in the distribution of filled pauses and speech repairs reveal that including them in the language model of the speech recognizer may be also helpful. The extent to which the number of events depends on utterance length and on the speaker is also explored. Statistics are shown that follow previous studies for English, and a sizeable space is devoted to comparing our results with them. Finally, various possible cues for the automatic detection of speech repairs--a key issue from the point of view of speech understanding--are explored: silent pauses, filled pauses, lengthenings, cut off words and discourse markers. As previously observed for English, none of them was found to be reliable by itself. More information, especially at the acoustic-prosodic level, is no doubt needed to reliably detect speech repairs.

Cut-offs and co-occurring gestures: Similarities between speakers’ first and second languages

“…There were also more cut-offs in the L2 narrations, although, the difference in the rate of cut-offs was not statistically different between the two languages. A comparison of the various types of disfluencies falls outside the scope of this study but previous work (Lopez-Ozieblo 2017;Machuca, Llisterri and Ríos Carratalá 2015;Rodríguez and Torres 2006) has identified a tendency for filled pauses and elongations, in addition to cut-offs, in Spanish L1 speakers. On the other hand, Lecumberri, Cooke and Wester (2017) observed similar patterns in the cut-offs of English and Spanish native speakers.…”

“Um, I can tell you're lying”: Linguistic markers of deception versus truth-telling in speech

“…A study of speech recognition in Spanish confirmed the frequency and acoustic predictability of fillers, and demonstrated that incorporation of such utterances as lexical items (rather than noise) in models of automatic speech recognition improves the recognizer's performance (Rodriguez & Torres, 2006).…”

Developing and Using Big Data Archives to Quantify Disfluency and Stuttering in Bilingual Children

“…20,21 Findings of these studies include an overall disfluency rate of 4 to 6 disfluencies per 100 words in Spanish, which is similar to reports of disfluency rates reported in adult-directed speech in English. 9,17 The common Spanish filled pauses ah, eh, and mm were found to occur at a rate of 1.67 per 100 words. In another corpus set of mostly formal language monologues, speakers produced an average of 5 disfluencies per 100 words (not including silent pauses); around 1.5 per 100 words were fillers.…”

“…10 In short, typical disfluencies are common in English spontaneous speech, and this finding has been replicated in other languages, including Chinese, Croatian, Dutch, German, Norwegian, Japanese, Russian, Spanish, Swedish, and Tok Pisin, suggesting the universality of these types of disfluencies in language production. [11][12][13][14][15][16][17][18][19] Several corpora-based studies have informed our knowledge of disfluencies in spontaneous speech dialogues in Spanish. In general, these studies report similar types of disfluencies as those reported for monolingual English speakers, including retracings, filled pauses, and hesitations/silent pauses.…”