Method and system for automatic determination of stress position in word forms

• US 10,043,510 B2
• Filed: 12/01/2016
• Issued: 08/07/2018
• Est. Priority Date: 12/28/2015
• Status: Active Grant

• Current Assignee: Direct Cursus Technology LLC
• Original Assignee: Yandex Europe AG (Yandex N.V.)
• Inventors: Zelenkov, Yury Grigorievich
• Primary Examiner(s): Lerner, Martin
• Application Number:
  US15/366,133
• Publication Number:
  US 20170185584A1
• Time in Patent Office: 614 Days
• Field of Search: 704 1, 704 8, 704 9, 704 10, 704258, 704260, 704266
• CPC Class Codes: G10L 13/10 Prosody rules derived from text; Stress or intonation

Abstract

A method and a computing device for building a reference system for determining a stress position of a new word form, the method comprising: sorting, in a reverse lexicographic order, a plurality of word forms being marked with a particular stress position; clustering the plurality of sorted word forms into a plurality of clusters, comprises a plurality of terminal clusters, each terminal cluster comprising word forms having both: (i) a same ending being a terminal common ending, and (ii) a same stress position, combination of the terminal common ending and said same stress position being unique; building, using the plurality of terminal clusters, the reference system having a reference to at least one terminal cluster of the plurality of terminal clusters, the at least one terminal cluster comprising an indication of the particular stress position proper to word forms which are included in that respective terminal cluster.

Claims

1. A method for building a reference system for determining, by a computing device, a stress position of a new word form, the method comprising:

sorting, in a reverse lexicographic order, a plurality of word forms, each word form of the plurality of word forms being marked with a particular stress position, in order to generate a plurality of sorted word forms,

clustering the plurality of sorted word forms into a plurality of clusters of word forms such that the plurality of clusters of word forms comprises a plurality of terminal clusters, each terminal cluster of the plurality of terminal clusters comprising word forms having both;

(i) a same ending being a terminal common ending, and (ii) a same stress position, a combination of the terminal common ending and said same stress position being unique;

building, using the plurality of terminal clusters, the reference system for determining the stress position of the new word form, the reference system having a reference to at least one terminal cluster of the plurality of terminal clusters, the at least one terminal cluster comprising an indication of the particular stress position proper to word forms which are included in that respective terminal cluster.

2. The method of claim 1, wherein the terminal common ending, within any terminal cluster, is an ending of a word forms comprising in an immediately preceding superior level cluster and also an additional letter.

3. The method of claim 1, wherein clustering the plurality of sorted word forms into a plurality of clusters of word forms further comprises organizing the plurality of clusters into a hierarchical tree-structure of clusters, the organizing being performed such that:

(i) the plurality of clusters of word forms comprises;

(a) a plurality of root clusters, each root cluster having at least one immediately following lower level cluster, and

(b) the plurality of terminal clusters, each terminal cluster of the plurality of terminal clusters having no lower level cluster;

(ii) at least some clusters of the hierarchical tree-structure, in respect to each other, are immediately preceding superior level clusters and immediately following lower level clusters, and

(iii) an ending of a word form in a immediately following lower level cluster has a same sequence of letters as in a immediately preceding superior level cluster and also an additional letter.

4. The method of claim 3, wherein the hierarchical tree-structure of clusters further comprises a plurality of internal clusters, each internal cluster being the immediately following lower level cluster of an immediately preceding superior level cluster and the immediately preceding superior level cluster in respect to at least one immediately following lower level cluster.

5. The method of claim 1 wherein word forms, the word forms having the same ending being the terminal common ending, have at least two different stress positions, the method further comprises generating at least two terminal clusters, each of said at least two terminal clusters comprising word forms having:

said terminal common ending, and

one respective same stress position, and

a number of occurrences of said one respective same stress position.

6. The method of claim 5, further comprising receiving a request for defining the stress position of the new word form and, responsive to receiving said request:

using a new ending of the new word form for finding, in the reference system for endings, said at least two terminal clusters, and

applying to the new word form that stress position which corresponds to a stress position of word forms being in that one of said at least two terminal clusters, which terminal cluster has a highest number of occurrences of a particular stress position.

7. The method of claim 1, further comprising, before the sorting the plurality of word forms, acquiring, from a supplying device, the plurality of word forms.

8. The method of claim 7, wherein the acquiring the plurality of word forms comprises acquiring at least one word form of the plurality of word forms being marked with the particular stress position.

9. The method of claim 7, wherein the acquiring the plurality of word forms is acquiring from at least one literature source.

10. The method of claim 1, wherein word forms are word forms of a particular language.

11. The method of claim 10, wherein word forms are Russian-language word forms.

12. The method of claim 1, further comprising receiving a request for defining the stress position of the new word form and, responsive to receiving said request:

using a new ending of the new word form for finding, in the reference system for endings, a corresponding terminal cluster having matching terminal common ending, and

applying to the new word form that stress position which corresponds to a stress position of word forms being included into the corresponding terminal cluster.

13. The method of claim 12, wherein the using the new ending of the new word form is any one, selected from:

(i) using the new ending of the new word form as a key, and (ii) using a reversed sequence of letters in the new ending of the new word form as a sequence of keys.

14. A computing device for building a reference system for determining a stress position of a new word form, the computing device comprising a processor and an information storage medium storing computer-readable instructions that, when executed by the processor, cause the processor to perform:

sorting, in a reverse lexicographic order, a plurality of word forms, each word form of the plurality of word forms being marked with a particular stress position, in order to generate a plurality of sorted word forms,

clustering the plurality of sorted word forms into a plurality of clusters of word forms such that the plurality of clusters of word forms comprises a plurality of terminal clusters, each terminal cluster of the plurality of terminal clusters comprising word forms having both;

(i) a same ending being a terminal common ending, and (ii) a same stress position, a combination of the terminal common ending and said same stress position being unique;

building, using the plurality of terminal clusters, the reference system for determining the stress position of the new word form, the reference system having a reference to at least one terminal cluster of the plurality of terminal clusters, the at least one terminal cluster comprising an indication of the particular stress position proper to word forms which are included in that respective terminal cluster.

15. The computing device of claim 14, wherein the terminal common ending, within any terminal cluster, is an ending of a word forms comprising in an immediately preceding superior level cluster and also an additional letter.

16. The computing device of claim 14, wherein clustering the plurality of sorted word forms into a plurality of clusters of word forms further comprises organizing the plurality of clusters into a hierarchical tree-structure of clusters, the organizing being performed by the processor such that:

(i) the plurality of clusters of word forms comprises;

(a) a plurality of root clusters, each root cluster having at least one immediately following lower level cluster, and

(b) the plurality of terminal clusters, each terminal cluster of the plurality of terminal clusters having no lower level cluster;

(ii) at least some clusters of the hierarchical tree-structure, in respect to each other, are immediately preceding superior level clusters and immediately following lower level clusters, and

(iii) an ending of a word form in a immediately following lower level cluster has a same sequence of letters as in a immediately preceding superior level cluster and also an additional letter.

17. The computing device of claim 16, wherein the hierarchical tree-structure of clusters further comprises a plurality of internal clusters, each internal cluster being the immediately following lower level cluster of an immediately preceding superior level cluster and the immediately preceding superior level cluster in respect to at least one immediately following lower level cluster.

18. The computing device of claim 14 wherein word forms, the word forms having the same ending being the terminal common ending, have at least two different stress positions, and wherein the computer-readable instructions, when executed by the processor, further cause the processor to generate at least two terminal clusters, each of said at least two terminal clusters comprising word forms having:

said terminal common ending, and

one respective same stress position, and

a number of occurrences of said one respective same stress position.

19. The computing device of claim 18, wherein the computer-readable instructions, when executed by the processor, further cause the processor to receive a request for defining the stress position of the new word form and, responsive to receiving said request:

to use a new ending of the new word form for finding, in the reference system for endings, said at least two terminal clusters, and

to apply to the new word form that stress position which corresponds to a stress position of word forms being in that one of said at least two terminal clusters, which terminal cluster has a highest number of occurrences of a particular stress position.

20. The computing device of claim 14, wherein the computer-readable instructions, when executed by the processor, further cause the processor to receive a request for defining the stress position of the new word form and, responsive to receiving said request:

to use a new ending of the new word form for finding, in the reference system for endings, a corresponding terminal cluster having matching terminal common ending, and

to apply to the new word form that stress position which corresponds to a stress position of word forms being included into the corresponding terminal cluster.