Improving phonetic alignment by handling secondary sequence structures

In traditional alignment analyses, sequences are only compared with regard to their primary structure. Here, the term primary structure refers to the order order of segments, whereby segments are understood as the smallest units of a sequence which directly correspond to the characters of the alphabet from which the sequence is drawn. Apart from the primary structure, sequences can, however, also have a secondary structure. Apart from segmentizing sequences into their primary units, one can further segmentize them into larger units of subsequences consisting of one or more primary segments. A secondary segmentation which is very common in linguistics is, e.g., the segmentation of words into syllables apart from the primary segmentation of words into phonemes. The traditional alignment modes such as global, local, or semiglobal alignment (cf. the overview in Durbin et al. 2002) align sequences only with respect to their primary structure. Thus, given the sequence "THE CATFISH HUNTS" and "THE CAT FISHES", they all yield an alignment in which the subsequence "CATFISH" of the first sequence is matched with the subsequence "CAT FISH" of the second sequence (see Table 1). In contrast to these primary alignments, a secondary alignment displays the similarities of sequences with regard to both their primary and their secondary structure, aligning letters which belong to the same word in one sequence only with those letters in the other sequence which also belong to a single word (see Table 1).