Modulated String Searching

Abstract: In his 1987 paper entitled "Generalized String Matching", Abrahamson introduced {\em pattern matching with character classes} and provided the first efficient algorithm to solve it. The best known solution to date is due to Linhart and Shamir (2009). Another broad yet comparatively less studied class of string matching problems is that of numerical string searching, such as, e.g., the `less-than' or $L_1$-norm string searching. The best known solutions for problems in this class are based on FFT convolution after some suitable re-encoding. The present paper introduces {\em modulated string searching} as a unified framework for string matching problems where the numerical conditions can be combined with some Boolean/numerical decision conditions on the character classes. One example problem in this class is the {\em locally bounded $L_1$-norm} matching problem on character classes: here the "match" between a character at some position in the text and a set of characters at some position in the pattern is assessed based on the smallest $L_1$ distance between the text character and one of those pattern characters. The two positions "match" if the (absolute value of the) difference between the two characters does not exceed a predefined constant. The pattern has an occurrence in an alignment with the text if the sum of all such differences does not exceed a second predefined constant value. This problem requires a pointwise evaluation of the quality of each match and has no known solution based on the previously mentioned algorithms.