Probabilistic grammar and the Portuguese Stress Corpus

This paper proposes a weight-based probabilistic approach to stress in Portuguese. Previous analyses have argued that weight-sensitivity in the language is categorical and restricted to word-final syllables. I show that weight effects in Portuguese are gradient and can be found across all three positions in the stress domain (three-syllable window). I also compare two domains of weight computation, namely, the syllable and the interval [5], and show that interval-based statistical models are more internally consistent. A thorough phonological analysis of stress in a given language requires a comprehensive and detailed corpus. Unlike frequency corpora, such a corpus must contain quantitative and qualitative segmental information as well as stress marking, syllabification, syllable shapes and weight profiles, among other variables. The present study introduces the Portuguese Stress Corpus (PSC, author). PSC was developed to provide researchers with a large and reliable annotated lexicon of Portuguese. The corpus contains nearly all non-verbs in the language (n = 154,611), and is based on the wordlist found in the Houaiss Dictionary [4], the most comprehensive dictionary in the Portuguese language. Background Weight, Syllables and Intervals: In languages where stress is weight-sensitive, syllables with greater weight are more likely to be prominent, i.e., to attract stress [2]. In interval theory [5], greater weight entails greater duration in a given interval, defined as the rhythmic unit that spans from a vowel up to (but not including) the following vowel; i.e., V-to-(V). Segments preceding the leftmost vowel are not included in any interval. Intervals (ι) have no a priori constituency, and predict different rhythmic units when compared to syllables (σ): onset segments in a given syllable are computed as part of the preceding interval. For example, the string CVCσCCVCσ is equivalent to 〈C〉VCCCιVCι in interval theory. The onset effects found in the Portuguese Stress Corpus motivate intervals, as they are negatively correlated with stress. Portuguese: Previous analyses ([1], [6], among others) have argued that weight effects on stress in Portuguese non-verbs are restricted to the word-final syllable (stress in verbs is not phonologically conditioned): stress is final if the word-final syllable is heavy. Otherwise, stress falls on the penult syllable (regardless of weight). Both final and penult stress patterns are (mostly) regular. Antepenult stress is irregular, and no pre-antepenult stress is allowed. Approximately 72% of the non-verbs in the language (N=163,626) have regular/predictable stress. Researchers have employed different factors to account for stress regularities (e.g., foot binarity, foot type, metrical alignment) and irregularities (e.g., extrametricality, catalexis, theme vowel influence) in the language. Under previous analyses, irregular and regular words are by definition treated differently. Methodology: In the present study, stress in the Portuguese Stress Corpus was modelled using Binomial Logistic Regressions (glm() in R). Given that three stress positions are possible, two binomial models were required to predict stress: one model predicts antepenult stress vs. penult or final stress ((1-a), (2-a)). Because all words with antepenult stress are considered to be irregular, this model predicts the following: given a word, how likely is it to bear antepenult stress as opposed to penult or final stress? Another model predicts penult vs. final stress ((1-b) and (2-b)); i.e., the two regular positions in the language. Because syllables and intervals are compared, a total of four models are used. Syllables have three constituents (onset, nucleus and coda). Intervals, on the other hand, are a single string of segments (no constituency). As a result, 9 predictors are used in (1-a), 6 in (1-b); 3 in (2-a), and 2 in (2-b).