Cross-lingual audio search

• US 8,805,871 B2
• Filed: 08/29/2012
• Issued: 08/12/2014
• Est. Priority Date: 06/28/2011
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Ajmera, Jitendra, Verma, Ashish
• Primary Examiner(s): BOWEN, RICHARD L
• Application Number:
  US13/598,296
• Publication Number:
  US 20130006604A1
• Time in Patent Office: 713 Days
• Field of Search: None
• US Class Current: 707/769
• CPC Class Codes: G06F 16/632 Query formulation
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

Methods and arrangements for executing an audio search. A search query in a first language variety is accepted, the search query being in a form of at least one of: text and audio. A corpus of material in the first language variety is accessed, and first language variety baseforms are obtained via data obtained from a second language variety. A first language variety phonetic model is built, and the first language variety phonetic model and the second language variety are employed in executing an audio search based on the accepted search query.

Claims

1. A method comprising:

accepting a search query in a first language variety, the search query being in a form of at least one of;

text and audio;

accessing a corpus of material in the first language variety;

determining similarity of a second language variety with respect to a first language variety;

choosing the second language variety based on determining that the first language variety baseforms can be obtained via data from the second language variety, and at least one selection criterion;

the at least one selection criterion comprising a ranking from among ranked pairs of language varieties, the ranked pairs being ranked on a basis of determined similarity between language varieties;

obtaining first language variety baseforms via data obtained from the second language variety;

thereupon building a first language variety phonetic model, based on the first language variety baseforms obtained via data obtained from the second language variety; and

executing an audio search based on the accepted search query, via employing the first language variety phonetic model and the second language variety.

2. The method according to claim 1, further comprising:

pre-processing search data in the first language variety;

said obtaining comprising obtaining first language variety baseforms via employing models trained on the second language variety; and

said executing comprises searching the search data for the accepted search query via comparing the accepted search query and the search data.

3. The method according to claim 2, wherein:

said obtaining comprises obtaining baseforms for words in the first language variety from the corpus; and

said pre-processing comprises pre-processing search data in the first language variety via employing phonetic models from both the first language variety and the second language variety.

4. The method according to claim 1, wherein said accessing comprises accessing corpus of first language text material from public sources and transliterating the text material into a script of the second language variety.

5. The method according to claim 1, wherein said executing comprises executing the audio search via employing a statistical model trained using dictionary data of the second language variety.

6. The method according to claim 1, wherein said executing comprises executing the audio search via employing at least one acoustic model of the second language variety and the phonetic language model of the first language variety.

7. The method according to claim 1, wherein said executing comprises generating a phonetic index for search data.

8. The method according to claim 7, wherein said generating comprises generating a phonetic lattice of the search query in the first language variety.

9. The method according to claim 8, wherein the phonetic index of the search data comprises a phonetic lattice representation of the search data.

10. The method according to claim 8, wherein:

said accepting comprises accepting a search query in the form of text; and

said generating of a phonetic lattice comprises generating at least one baseform of the accepted search query and converting the at least one baseform to a phonetic lattice.

11. The method according to claim 8, wherein:

said accepting comprises accepting a search query in the form of audio; and

said generating of a phonetic lattice comprises generating a phonetic lattice of the search query directly via employing an acoustic model of the second language variety and the phonetic language model of the first language variety.