Michael Cysouw

Frontmatter was published in Maltese on page I.

This paper studies some factors governing the presence or absence of word-final schwa in German. To obtain data as homogeneous as possible we focus on three adverbs outside morphological paradigms, namely, heut(e) ‘today’, gern(e) ‘willingly’, and bald(e) ‘soon’, in one particular text type, the letters written by one and the same person, the writer Johann Wolfgang von Goethe (1749–1832). Apart from lexical differences between the items studied and change over time, various phonological factors are shown to be important, most prominently the accent pattern of the following word (schwa tends to be present if the first syllable of the following word is accentuated), foot structure, and the initial segment of the following word. Statistical analyses, both for the individual factors and their (potential) interactions, reveal significant patterns at work behind the variation. For gern(e) the most important factors are purely phonological while for heut(e) the type of the following boundary and the position in the sentence is crucial.

The choice associated with words is a fundamental property of natural languages. It lies at the heart of quantitative linguistics, computational linguistics and language sciences more generally. Information theory gives us tools at hand to measure precisely the average amount of choice associated with words: the word entropy. Here, we use three parallel corpora, encompassing ca. 450 million words in 1916 texts and 1259 languages, to tackle some of the major conceptual and practical problems of word entropy estimation: dependence on text size, register, style and estimation method, as well as non-independence of words in co-text. We present two main findings: Firstly, word entropies display relatively narrow, unimodal distributions. There is no language in our sample with a unigram entropy of less than six bits/word. We argue that this is in line with information-theoretic models of communication. Languages are held in a narrow range by two fundamental pressures: word learnability and word expressivity, with a potential bias towards expressivity. Secondly, there is a strong linear relationship between unigram entropies and entropy rates. The entropy difference between words with and without co-textual information is narrowly distributed around ca. three bits/word. In other words, knowing the preceding text reduces the uncertainty of words by roughly the same amount across languages of the world.

The choice associated with words is a fundamental property of natural languages. It lies at the heart of quantitative linguistics, computational linguistics, and language sciences more generally. Information-theory gives us tools at hand to measure precisely the average amount of choice associated with words—the word entropy. Here we use three parallel corpora—encompassing ca. 450 million words in 1916 texts and 1259 languages—to tackle some of the major conceptual and practical problems of word entropy estimation: dependence on text size, register, style and estimation method, as well as non-independence of words in co-text. We present three main results: 1) a text size of 50K tokens is sufficient for word entropies to stabilize throughout the text, 2) across languages of the world, word entropies display a unimodal distribution that is skewed to the right. This suggests that there is a trade-off between the learnability and expressivity of words across languages of the world. 3) There is a strong linear relationship between unigram entropies and entropy rates, suggesting that they are inherently linked. We discuss the implications of these results for studying the diversity and evolution of languages from an information-theoretic point of view.

In this chapter, we illustrate a method for identifying clusters of semantic roles by cross-linguistic comparison. On the basis of data from 25 languages drawn from the ValPaL (Valency Patterns Leipzig) database, we show how one can visualize coexpression tendencies using quantitative methods (in particular, multidimensional scaling). Traditionally, the coexpression of semantic microroles (such as the breaker and the broken thing of the ‘break’ verb, the helper and the helpee of the ‘help’ verb, etc.) has been studied for particular languages, with generalized macroroles such as “agent”, “actor”, and “undergoer” being compared across languages in a next step. We set up a conceptual space of 87 microroles based on their coexpression tendencies, i.e. the extent to which they are expressed identically (via flagging and indexing) across our languages. The individual coding means (cases, adpositions, index-sets) can then be mapped onto this conceptual space, revealing broader alignment patterns.

Instead of defining semantic roles on the basis of the interpretation of lexical predicates, I will show that it is possible to induce semantic roles from the usage of case-like markers across a wide variety of languages. The assumptions behind this proposal are, first, that semantic roles are strongly contextually determined and, second, that similarity in coding of contextual roles across many different languages shows which contexts evoke the same (or better: very similar) semantic roles. This approach to the investigation of semantic roles will be exemplified by a study of case-like marking in a parallel text across a sample of fifteen languages. On this basis, a semantic map of contextual roles can be established, and it will be shown that higher-level abstractions, like semantic roles or even macro-roles, can be statistically derived from this diversity of marking across many languages. Further, a typology of alignment systems can be derived statistically. Keywords: Semantic roles; case; parallel text; linguistic typology; alignment