An Overview of Phonetic Encoding Algorithms
- PDF / 230,294 Bytes
- 15 Pages / 612 x 792 pts (letter) Page_size
- 90 Downloads / 204 Views
E SCALE SYSTEMS CONTROL
An Overview of Phonetic Encoding Algorithms V. S. Vykhovanets∗,a , J. Du∗∗,b , and S. A. Sakulin∗∗,c ∗
Trapeznikov Institute of Control Sciences, Russian Academy of Sciences, Moscow, Russia ∗∗ Bauman Moscow State Technical University, Moscow, Russia e-mail: a [email protected], b [email protected], c [email protected] Received September 12, 2017 Revised September 12, 2017 Accepted May 25, 2020
Abstract—This paper presents an overview of the phonetic encoding algorithms designed to determine the similarity of words in sound (pronunciation). Phonetic encoding algorithms are divided into the algorithms for comparing words and the algorithms for determining the distance between words. Word comparison algorithms, such as SoundEx, NYSIIS, Daitch–Mokotoff, Metaphone, and Polyphone, as well as algorithms for determining the distance between words, such as Levenshtein, Jaro, and N -grams, are described. For each algorithm, the advantages and shortcomings are discussed, and an analog for the Russian language is given. For eliminating the common shortcomings of phonetic encoding algorithms, the idea suggested in this paper is to use not the letter sequences of words, but the sequences of their elementary sounds. In this case, word recognition, record linkage, and word indexing by sounds are expected to improve. Keywords: phonetic encoding algorithm, phonetic distance, record linkage, word indexing by sound DOI: 10.1134/S0005117920100082
1. INTRODUCTION Phonetic encoding algorithms are algorithms for word indexing by sound, which use a letter sequence of a word and pronunciation rules to convert them into text (code, index, or key). If the encoding texts for two different words coincide or are close to one another, the resulting conclusion is that these words are similar in sound. The first phonetic encoding algorithm for the English language, SoundEx [69], was used in the 1930s to encode surnames in a population census. The algorithm is based on an encoding method designed to eliminate spelling and typographical errors in names. For example, SoundEx will calculate the same encoding text “S530” for such words as “Smith,” “Smithe,” and “Smyth,” i.e., will identify these words as identical in sound. With the development of computer technologies, many other phonetic encoding algorithms appeared, including the ones for other natural languages. Examples of such algorithms are NYSIIS [77], the advanced SoundEx [75], Metaphone [40], and others. Phonetic encoding algorithms include not only algorithms for comparing words, but also algorithms for determining the distance between words in the case of searching by sound. The algorithms for calculating the Levenshtein distance [2] and the Jaro distance [30], as well as the distance based on N -grams [33] are most widespread in applications. Phonetic encoding algorithms are widely used in the fields where the comparison of acoustic data with text samples is required. Some example include speech recognition, word spelling check and correction, database search, extrac
Data Loading...
