System and method for processing out of vocabulary compound words

• US 10,380,242 B2
• Filed: 05/14/2015
• Issued: 08/13/2019
• Est. Priority Date: 05/14/2015
• Status: Active Grant

• Current Assignee: Cerence Operating Company (Cerence Inc.)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Minnis, Stephen, Bos-Plachez, Corinne
• Primary Examiner(s): Yehl, Walter
• Application Number:
  US14/712,074
• Publication Number:
  US 20160336004A1
• Time in Patent Office: 1,552 Days
• CPC Class Codes: G06F 40/232 Orthographic correction, e.g. spell checking or vowelisation
• CPC Class Codes: G06F 40/242 Dictionaries
• CPC Class Codes: G10L 13/10 Prosody rules derived from text; Stress or intonation

Abstract

A system and method for out-of-vocabulary compound word handling is provided. Embodiments may include storing a plurality of compound word rules and compound word dictionaries in a database. Embodiments may also include evaluating membership criteria associated with a received compound word, wherein membership criteria includes at least one of dictionary based or part of speech (POS) based criteria. Embodiments may further include applying one or more filtering rules to the received compound word.

Claims

1. A computer-implemented method for out-of-vocabulary compound word handling comprising:

storing a plurality of compound word rules and compound word dictionaries in a database;

evaluating membership criteria associated with a received compound word, wherein membership criteria includes at least one of dictionary based or part of speech (POS) based criteria;

applying one or more filtering rules to the received compound word;

identifying the received compound word as an out-of-vocabulary compound word based upon, at least in part, the one or more filtering rules, wherein the one or more filtering rules includes determining whether one of a first word and a last word of the received compound word is one of the first word and the last word in the compound word dictionaries;

predicting a compound prominence pattern associated with the received compound word based upon, at least in part, identifying the received compound word as the out-of-vocabulary compound word, wherein the compound prominence pattern includes a default compound prominence when a no confidence situation exists for predicting the compound prominence pattern associated with the received compound word in the out-of-vocabulary compound word handling for text-to-speech synthesis, wherein the default compound prominence includes one of right side prominence and left side prominence for a portion of the compound word; and

generating an output representative of the received compound word based upon, at least in part, the default compound prominence.

2. The computer-implemented method of claim 1, wherein evaluating membership criteria includes reviewing one or more orthographic patterns.

3. The computer-implemented method of claim 1, wherein applying one or more filtering rules includes generalizing over a list of known compounds in the database.

4. The computer-implemented method of claim 1, further comprising:

predicting a compound pronunciation associated with the compound word.

5. The computer-implemented method of claim 4, further comprising:

predicting a compound part of speech associated with the compound word.

6. The computer-implemented method of claim 5, further comprising:

generating a model based upon, at least in part, the predicted compound prominence pattern, compound pronunciation, and compound part of speech.

7. The computer-implemented method of claim 1, wherein the compound word is a compound word with more than two words.

8. The computer-implemented method of claim 1 further comprising ruling out a co-occurrence of the received compound word based upon, at least in part, applying the one or more filtering rules to the received compound word.

9. The computer-implemented method of claim 1 wherein the no confidence situation exists when available statistical confidence for the received compound word is 50% for both right side prominence and left side prominence.

10. The computer-implemented method of claim 1 wherein the output is an audible output.

11. A non-transitory computer-readable storage medium having stored thereon instructions, which when executed by a processor result in one or more operations for out-of-vocabulary compound word handling, the operations comprising:

storing a plurality of compound word rules and compound word dictionaries in a database;

evaluating membership criteria associated with a received compound word, wherein membership criteria includes at least one of dictionary based or part of speech (POS) based criteria;

applying one or more filtering rules to the received compound word;

identifying the received compound word as an out-of-vocabulary compound word based upon, at least in part, the one or more filtering rules, wherein the one or more filtering rules includes determining whether one of a first word and a last word of the received compound word is one of the first word and the last word in the compound word dictionaries;

predicting a compound prominence pattern associated with the received compound word based upon, at least in part, identifying the received compound word as the out-of-vocabulary compound word, wherein the compound prominence pattern includes a default compound prominence when a no confidence situation exists for predicting the compound prominence pattern associated with the received compound word in the out-of-vocabulary compound word handling for text-to-speech synthesis, wherein the default compound prominence includes one of right side prominence and left side prominence for a portion of the compound word; and

generating an output representative of the received compound word based upon, at least in part, the default compound prominence.

12. The non-transitory computer-readable storage medium of claim 11, wherein evaluating membership criteria includes reviewing one or more orthographic patterns.

13. The non-transitory computer-readable storage medium of claim 11, wherein applying one or more filtering rules includes generalizing over a list of known compounds in the database.

14. The non-transitory computer-readable storage medium of claim 11, further comprising:

predicting a compound pronunciation associated with the compound word.

15. The non-transitory computer-readable storage medium of claim 14, further comprising:

predicting a compound part of speech associated with the compound word.

16. The non-transitory computer-readable storage medium of claim 15, further comprising:

generating a model based upon, at least in part, the predicted compound prominence pattern, compound pronunciation, and compound part of speech.

17. The non-transitory computer-readable storage medium of claim 14, wherein the compound word has more than two words.

19. The system of claim 17, wherein evaluating membership criteria includes reviewing one or more orthographic patterns.

20. The system of claim 17, wherein applying one or more filtering rules includes generalizing over a list of known compounds in the database.

18. A system configured to perform out-of-vocabulary compound word handling comprising:

one or more processors configured to allow for storing a plurality of compound word rules and compound word dictionaries in a database, the one or more processors further configured to evaluate membership criteria associated with a received compound word, wherein membership criteria includes at least one of dictionary based or part of speech (POS) based criteria, the one or more processors further configured to identify the received compound word as an out-of-vocabulary compound word based upon, at least in part, the one or more filtering rules, wherein the one or more filtering rules determining whether one of a first word and a last word of the received compound word is one of the first word and the last word in the compound word dictionaries, the one or more processors further configured to apply one or more filtering rules to the received compound word, the one or more processors further configured to predict a compound prominence pattern associated with the received compound word based upon, at least in part, identifying the received compound word as the out-of-vocabulary compound word, wherein the compound prominence pattern includes a default compound prominence when a no confidence situation exists for predicting the compound prominence pattern associated with the received compound word in the out-of-vocabulary compound word handling for text-to-speech synthesis, wherein the default compound prominence includes one of right side prominence and left side prominence for a portion of the compound word, the one or more processors further configured to generate an output representative of the received compound word based upon, at least in part, the default compound prominence.