Abstract
This calculator should provide quick support to corpus users when calculating basic statistical tasks most commonly encountered in research. The app is divided into 7 modules reflecting specific research problems: The first module "1 word in 1 corpus" does not in fact do any statistical testing as such but serves as a tool for adequate frequency interpretation.
It should help you answer the question: What exactly does it mean, when a feature I am interested in has a frequency X in a corpus? The second module "2 words in 1 corpus" compares two frequencies (e.g. two competing variants in one corpus) and determines, how significant their difference really is and whether their difference is not just a result of a random variation. The typical use case of the third module "2 words in 2 corpora" is the identification of keywords - units that are significantly more common in one of two corpora (considering the size of the two corpora).
It is, however, useful for any comparison of frequencies across different corpora. The fourth module "1 feature - more samples" helps in establishing the level of precision and reliability of a random sample analysis.
If the resulting range of variation (for the feature under scrutiny) is too broad, it might be advisable to improve the precision by adding more samples. Module "Many features - 1 sample" is used to estimate the occurrence of groups of features (e.g. word senses) in an analysed sample or concordance.
It can be used to show whether one group is truly more frequent than the other or whether a group can be considered to be actually attested. The sixth module labeled "zTTR" compares the lexical richness (the length of the sample in relation to the number of lexical units it contains) of texts.
Its advantage is that the resulting zTTR index is comparable even between texts of different length. When comparing multi-word units between two languages, we often face a problem of n-gram length (non-)correspondence.
The seventh module "N-gram correspondence" is used to investigate the N-grams correspondence by showing what is the actual counterpart of e.g. a list of the most frequent bigrams in one language when compared to another language.