Predicting the cost associated with translating textual content

• US 10,417,646 B2
• Filed: 03/09/2010
• Issued: 09/17/2019
• Est. Priority Date: 03/09/2010
• Status: Active Grant

• Current Assignee: SDL Inc (RWS Holdings Plc)
• Original Assignee: SDL Inc (RWS Holdings Plc)
• Inventors: Soricut, Radu, Viswanathan, Narayanaswamy, Marcu, Daniel
• Primary Examiner(s): Campbell, Shannon S
• Assistant Examiner(s): Nelson, Freda A
• Application Number:
  US12/720,536
• Publication Number:
  US 20110225104A1
• Time in Patent Office: 3,479 Days
• Field of Search: 705400, 705 1426, 705 1439, 705 40, 704 4, 704 5, 704 9, 704 2- 3, 707770
• CPC Class Codes: G06F 40/58 Use of machine translation, e.g. for multi-lingual retrieval, for server-side translation for client devices or for real-time translation
• CPC Class Codes: G06F 9/4843 by program, e.g. task dispatcher, supervisor, operating system
• CPC Class Codes: G06Q 10/067 Enterprise or organisation modelling
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0283 Price estimation or determination

Abstract

A prediction of the cost associated with translating textual content in a source language can be determined. A first quantity estimation of first textual content may be determined. The first textual content is to be translated via human translation. A second quantity estimation of second textual content may also be determined. The second textual content is to be translated via machine translation. An indication of a target language is obtained, wherein the source language and the target language form a language pair. The prediction of the cost associated with translating the first textual content and the second textual content from the source language to the target language is then determined. The prediction is based at least in part on the first quantity estimation, the second quantity estimation, and the language pair.

Claims

1. A method comprising:

receiving an estimate of a first quantity of a first type of textual content of documents of a source language expected to be translated in the future, the first type of textual content including text expected to be translated via human translation for which a near-perfect translation is desired;

receiving an estimate of a second quantity of a second type of textual content of the documents expected to be translated in the future, the second type of textual content different from the first type of textual content, the second type of textual content including text expected to be translated via machine-generated translation for which a potentially imperfect translation is acceptable;

executing instructions stored in a memory by a processor to obtain an indication of a target language, the source language and the target language forming a language pair;

training a machine translation system for the language pair using a training dataset associated with similar subject matter as documents of the second type of textual content expected to be translated in the future;

comparing a machine-generated target-language corpus with a human-generated target-language corpus, relative to the second type of textual content of the documents expected to be translated in the future;

mapping features such as similarities and differences between the machine-generated target-language corpus and the human-generated target-language corpus using the comparison;

predicting a quality level of the trained machine translation system associated with translational accuracy of future machine-generated translations relative to human-generated translations using the mapping;

executing instructions stored in a memory by the processor to determine a prediction of a first cost associated with translating the first type of textual content and a second cost associated with the second type of textual content from the source language to the target language, the prediction based at least in part on the first quantity estimation, the second quantity estimation, the predicted quality level, and the language pair;

selecting the trained machine translation system to perform machine translations of the second type of textual content based on the predicted quality level of translation of the second type of textual content by the machine translation system and the predicted first and second cost;

receiving text for translation;

receiving a selection of a quantity of the received text that is of the second type of textual content to be translated by a machine translation system rather than by a human; and

performing a machine translation of the selected quantity of the received text of the second type of textual content using the selected machine translation system.

2. The method of claim 1, wherein each predicted quality level is associated with a translation cost, the amount of the translation cost corresponding to the quality of translation.

3. The method of claim 1, wherein the predicted quality level is determined without translating the second type of textual content.

4. The method of claim 1, wherein a plurality of predicted quality levels are each associated with a different translation cost and a different portion of the second type of textual content.

5. The method of claim 1, wherein estimating the first quantity includes identifying existing textual content to be translated via human translation.

6. The method of claim 1, wherein estimating the first quantity includes estimating expected textual content to be translated via human translation.

7. The method of claim 1, wherein estimating the second quantity includes identifying existing textual content to be translated via machine translation.

8. The method of claim 1, wherein estimating the second quantity includes estimating expected textual content to be translated via machine translation.

9. The method of claim 1, further comprising generating a report that includes a schedule of cost options associated with the prediction of the cost, the cost options based at least in part on different quantities of the first type of textual content being translated and different quantities of the second type of textual content being translated.

10. The method of claim 1, further comprising determining a machine translation system to perform the machine translation, the determination of the machine translation system based on content associated with the second type of textual content.

11. A system for determining a prediction of a cost associated with translating textual content expected to be translated in the future, but before the translations are performed, the system comprising:

a first assessment module stored in a memory of a smartphone and executable by the smartphone to obtain a first quantity estimation of a first type of textual content of documents expected to be translated in the future, to be included in content pages of the smartphone, the first type of textual content to be translated via human translation;

a second assessment module stored in a memory of the smartphone and executable by the smartphone to obtain a second quantity estimation of a second type textual content of documents expected to be translated in the future, to be included in the content pages, the second type of textual content different from the first type of textual content, the second type of textual content to be translated via machine translation on the smartphone;

a language module stored in a memory and executable by a processor to obtain an indication of a target language, a source language and the target language forming a language pair; and

a cost prediction module stored in a memory and executable by the smartphone to;

compare a machine-generated target-language corpus with a human-generated target-language corpus, relative to the second type of textual content of the documents expected to be translated in the future,

map features such as similarities and differences between the machine-generated target-language corpus and the human-generated target-language corpus using the comparison,

determine a quality level associated with translational accuracy of future machine-generated translations using the mapping, and

determine the prediction of a cost associated with translating the first type of textual content and the second type of textual content from the source language to the target language, the prediction based at least in part on the first quantity estimation, the second quantity estimation, and the language pair.

12. The system of claim 11, further comprising a quality estimation module stored in a memory and executable by a processor to predict a quality level attainable via machine translation of at least a portion of the second type of textual content.

13. The system of claim 12, wherein a higher predicted quality level corresponds to a lower prediction of the cost associated with translating the second type of textual content, relative to a lower predicted quality level.

14. The system of claim 12, wherein the quality level is predicted without translating the second type of textual content.

15. The system of claim 12, wherein different portions of the second type of textual content are each associated with a different quality estimation.

16. The system of claim 11, further comprising a text evaluation module stored in a memory and executable by a processer to perform the following:

identify existing textual content to be translated via machine translation, the existing textual content to be translated via machine translation forming at least a portion of the second quantity estimation obtained by the second assessment module; and

estimate quality of forthcoming textual content to be translated via machine translation based on a determination of trust levels associated with translational accuracy of future machine-generated translations that lack corresponding human-generated translations, the forthcoming textual content to be translated via machine translation forming at least a portion of the second quantity estimation obtained by the second assessment module.

17. The system of claim 11, further comprising a report generation module stored in a memory and executable by a processor to generate a report that includes a schedule of cost options associated with the prediction of the cost, the cost options based at least in part on different quantities of the first type of textual content being translated and different quantities of the second textual content being translated.

18. The system of claim 11, further comprising a content language module stored in a memory and executable by a processor to determine a machine translation system to perform the machine translation, the determination of the machine translation system based on content associated with the second type of textual content.

19. A non-transitory computer-readable storage medium having instructions embodied thereon, the instructions being executable by a processor to perform a method for determining a prediction of a cost associated with translating textual content in a source language, the method comprising:

receiving a first quantity estimation of first textual content of documents expected to be translated in the future, the first textual content expected to be translated via human translation;

receiving a second quantity estimation of second textual content, the second textual content different from the first textual content of the documents expected to be translated in the future, the second textual content expected to be translated via machine translation;

obtaining an indication of a target language, the source language and the target language forming a language pair;

comparing a machine-generated target-language corpus with a human-generated target-language corpus, relative to the second type of textual content of the documents expected to be translated in the future;

mapping features such as similarities and differences between the machine-generated target-language corpus and the human-generated target-language corpus using the comparison;

determining a quality level associated with translational accuracy of future machine-generated translations using the mapping; and

determining a prediction of a first cost associated with translating the first textual content by a human translator and a second cost associated with translating the second textual content using machine translation, from the source language to the target language, the prediction based at least in part on the first quantity estimation, the second quantity estimation and determined quality level of machine-generated translations, and the language pair.