LANGUAGE PHONETIC PROCESSING BASED ON FINE-GRAINED MAPPING OF PHONETIC COMPONENTS

• US 20200135177A1
• Filed: 10/31/2018
• Published: 04/30/2020
• Est. Priority Date: 10/31/2018
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Li, Min, Li, Yunyao, Hailpern, Marina D., Noeman, Sara
• Granted Patent: US 10,811,003 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 40/20 Natural language analysis semantic analysis of natural language G06F40/30
• CPC Class Codes: G10L 15/005 Language recognition
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams
• CPC Class Codes: G10L 2015/025 Phonemes, fenemes or fenones being the recognition units

Abstract

In one embodiment, a computer-implemented method includes obtaining a pronunciation of a first word of a particular language and identifying a phonetic component of the pronunciation. The method includes obtaining a phonetic component mapping table for the type of phonetic component identified in the pronunciation of the first word and assigning a phonetic value to the identified phonetic component using the phonetic component mapping table. For a second word, the method includes obtaining a pronunciation of a second word, identifying a phonetic component of the pronunciation, and assigning a phonetic value to the identified phonetic component. In addition, the method includes calculating a phonetic distance between the identified phonetic component of the first word and the identified phonetic component of the second word, using the assigned phonetic values of the respective identified phonetic components of the first word second word, and storing the calculated phonetic distance in a file.

Claims

1. A computer-implemented method for determining a phonetic distance between two words of a particular language, the computer-implemented method comprising:

obtaining a pronunciation of a first word of a particular language;

identifying a phonetic component of the pronunciation of the first word, wherein the phonetic component corresponds to a type of phonetic component of the particular language;

obtaining a phonetic component mapping table for the type of phonetic component identified in the pronunciation of the first word;

assigning a phonetic value to the identified phonetic component using the phonetic component mapping table;

obtaining a pronunciation of a second word of the particular language, wherein the first word and the second word are different;

identifying a phonetic component of the pronunciation of the second word;

assigning a phonetic value to the identified phonetic component of the second word using the phonetic component mapping table;

calculating a phonetic distance between (i) the identified phonetic component of the first word and (ii) the identified phonetic component of the second word, using (a) the assigned phonetic value of the identified phonetic component of the first word and (b) the assigned phonetic value of the identified phonetic component of the second word; and

storing the calculated phonetic distance in association with the identified phonetic component of the first word.

2. The computer-implemented method of claim 1, wherein the phonetic component mapping table has a plurality of dimensions, the dimensions determined by the type of phonetic component of the particular language,

wherein each dimension represents a phonetic encoding characteristic of the type of phonetic component,

wherein each dimension is divided into partitions, each partition corresponding to a variance of the phonetic encoding characteristic as defined by the rules of the language for the type of phonetic component.

3. The computer-implemented method of claim 1, comprising,

determining a phonetic similarity distance between the first word and the second word,

wherein the phonetic similarity distance comprises calculating a sum of a plurality of phonetic distances between the first word and the second word, each phonetic distance representing a given type of phonetic component.

4. The computer-implemented method of claim 1, wherein the particular language has at least two different types of phonetic components.

5. The computer-implemented method of claim 4, wherein each different type of phonetic component is mapped to a phonetic component mapping table corresponding to that type of phonetic component.

6. The computer-implemented method of claim 4, wherein the pronunciation of the first word has a plurality of phonetic components, wherein each of the plurality of phonetic components is one of the at least two different types of phonetic components of the particular language.

7. A computer-implemented method for ranking a series of candidate words with pronunciation similar to that of a seed word, the computer-implemented method comprising:

obtaining a pronunciation of a seed word of a particular language;

identifying a phonetic component of the pronunciation of the seed word, wherein the phonetic component corresponds to a type of phonetic component of the particular language;

obtaining a phonetic component mapping table for the type of phonetic component identified in the pronunciation of the seed word,

assigning a phonetic value to the identified phonetic component using the phonetic component mapping table;

obtaining a pronunciation of a given one of a plurality of candidate words of the particular language, wherein the plurality of candidate words and the seed word are different;

identifying a phonetic component of the pronunciation of the given one of the plurality of candidate words;

assigning a phonetic value to the identified phonetic component using the phonetic component mapping table;

for each type of phonetic component identified in the seed word, calculating a phonetic distance between (i) the identified phonetic component of the seed word and (ii) the identified phonetic component of the candidate word, using (a) the assigned phonetic value of the identified phonetic component of the seed word and (b) the assigned phonetic value of the identified phonetic component of the candidate word;

determining a phonetic similarity distance between the seed word and the candidate word,

wherein the phonetic similarity distance comprises calculating a sum of a plurality of phonetic distances between the seed word and the candidate word, each phonetic distance representing a given type of phonetic component;

generating a series of candidate words, wherein each candidate word in the series of candidate words has a pronunciation similar to that of the seed word based on a value of the determined phonetic similarity distance between the seed word and each candidate word; and

ranking the candidate words that have a pronunciation similar to the seed word in order of the value of the determined phonetic similarity distance between the seed word and each candidate word.

8. The computer-implemented method of claim 7, wherein the phonetic component mapping table has a plurality of dimensions, the dimensions determined by the type of phonetic component of the particular language,

wherein each dimension represents a phonetic encoding characteristic of the type of phonetic component,

wherein each dimension is divided into partitions, each partition corresponding to a variance of the phonetic encoding characteristic as defined by the rules of the language for the type of phonetic component.

9. The computer-implemented method of claim 8, wherein the phonetic value is assigned to the identified phonetic component according to the location of the identified phonetic component on the phonetic component mapping table,

wherein the location of the identified phonetic component is determined by the respective phonetic encoding characteristics of each dimension on the phonetic component mapping table.

10. The computer-implemented method of claim 7, wherein the particular language has at least two different types of phonetic components.

11. The computer-implemented method of claim 7, wherein the pronunciation of the seed word is comprised of portions having at least one phonetic component.

12. The computer-implemented method of claim 11, wherein the pronunciation of the seed word has a plurality of portions, wherein each portion of the seed word has at least two different types of phonetic components of the particular language.

13. The computer-implemented method of claim 7, wherein the phonetic similarity distance S is calculated according to the formula:

14. The computer-implemented method of claim 7, wherein the generated series of candidate words is within a similarity threshold.

15. The computer-implemented method of claim 14, wherein the similarity threshold is set to generate a number of candidate words in the range of 10 words to 100 words.

16. The computer-implemented method of claim 7, wherein the language is Chinese.

17. The computer-implemented method of claim 16, wherein a Chinese word is comprised of a series of Chinese characters, each Chinese character having at least one Pinyin phonetic component selected from the group of components consisting of:

an initial, a final, and a tone.

18. The computer-implemented method of claim 17, wherein the phonetic similarity distance between the seed word and the candidate word is calculated using the following formula:

19. A computer program product for ranking a series of candidate words with pronunciation similar to that of a seed word, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, wherein the computer readable storage medium is not a transitory signal per se, the program instructions executable by a computer to cause the computer to perform a method comprising:

obtaining, by computer, a pronunciation of a seed word of a particular language;

identifying, by computer, a phonetic component of the pronunciation of the seed word, wherein the phonetic component corresponds to a type of phonetic component of the particular language;

obtaining, by computer, a phonetic component mapping table for the type of phonetic component identified in the pronunciation of the seed word;

assigning, by computer, a phonetic value to the identified phonetic component using the phonetic component mapping table;

obtaining, by computer, a pronunciation of a given one of a plurality of candidate words of the particular language, wherein the plurality of candidate words and the seed word are different;

identifying, by computer, a phonetic component of the pronunciation of the given one of a plurality of candidate words;

assigning, by computer, a phonetic value to the identified phonetic component using the phonetic component mapping table;

for each type of phonetic component identified in the seed word, calculating, by computer, a phonetic distance between (i) the identified phonetic component of the seed word and (ii) the identified phonetic component of the candidate word, using (a) the assigned phonetic value of the identified phonetic component of the seed word and (b) the assigned phonetic value of the identified phonetic component of the candidate word;

determining, by computer, a phonetic similarity distance between the seed word and the candidate word,

wherein the phonetic similarity distance comprises calculating a sum of a plurality of phonetic distances between the seed word and the candidate word, each phonetic distance representing a given type of phonetic component;

generating, by computer, a series of candidate words, wherein each candidate word in the series of candidate words has a pronunciation similar to that of the seed word based on a value of the determined phonetic similarity distance between the seed word and each candidate word; and

ranking, by computer, the candidate words that have a pronunciation similar to the seed word in order of the value of the determined phonetic similarity distance between the seed word and each candidate word.

20. A computer-implemented method for ranking a series of candidate words with pronunciation similar to that of a seed word, wherein the candidate words and the seed word are of the Chinese language, the computer-implemented method comprising:

obtaining a Pinyin pronunciation of a seed word of the Chinese language,

wherein the seed word is comprised of a series of characters,

wherein each character has a Pinyin pronunciation;

identifying a Pinyin phonetic component of the Pinyin pronunciation of one character of the seed word,

wherein the Pinyin phonetic component is selected from the group of Pinyin phonetic components consisting of;

an initial, a final, and a tone;

obtaining a Pinyin component mapping table selected from the group consisting of;

a Pinyin initial mapping table, a Pinyin final mapping table, and a Pinyin tone mapping table;

assigning a phonetic value to the identified Pinyin phonetic component of the Pinyin pronunciation of the character of the seed word using the respective Pinyin phonetic component mapping table;

obtaining a Pinyin pronunciation of a given one of a plurality of candidate words of the Chinese language, wherein the given one of the candidate words is comprised of a series of characters, wherein each character has a Pinyin pronunciation;

identifying a Pinyin phonetic component of the Pinyin pronunciation of a character of the given one of the candidate words;

assigning a phonetic value to the identified Pinyin phonetic component of the Pinyin pronunciation of the character using the respective Pinyin phonetic component mapping table;

for each type of phonetic component identified in the character of the seed word, calculating a phonetic distance between (i) the identified Pinyin phonetic component of the character of the seed word and (ii) the identified Pinyin phonetic component of the character of the candidate word using (a) the assigned phonetic value of the identified Pinyin phonetic component of the character of the seed word and (b) the assigned phonetic value of the identified Pinyin phonetic component of the character of the candidate word;

determining a phonetic similarity distance between the seed word and the candidate word,

wherein the phonetic similarity distance comprises calculating a sum of a plurality of phonetic distances between the seed word and the candidate word, each phonetic distance representing a given type of phonetic component;

generating a series of candidate words, wherein each candidate word in the series of candidate words has a Pinyin pronunciation similar to that of the seed word based on a value of the determined phonetic similarity distance between the seed word and each candidate word; and

ranking the candidate words that have a Pinyin pronunciation similar to the seed word in order of the value of the determined phonetic similarity distance between the seed word and each candidate word.

21. The computer-implemented method of claim 20, wherein the phonetic similarity distance between the seed word and the candidate word is calculated using the following formula: