 |
| Paid Research - Membership Required |
| Automated Translation Technology |
| |
| Keywords |
| access, automated translation, discovery, knowledge management, knowledge transfer, machine translation, natural language processing, translation memory, statistical machine translation, rules-based machine translation, OLMT, MT, consumability |
Abstract |
Of all natural language processing (NLP) technologies, automated translation harbors the most potential as a disruptive technology for global communications and commerce.
Content volume is increasing faster than any company or government can manage, much less translate into all the languages of their employees or external users. Faced with mass quantities of information written in other languages, most organizations choose not to translate most of that content (we call this “zero translation”). Many take the path of high-quality but relatively expensive human translation (HT), picking and choosing just a small subset of their corporate content to translate. A growing number choose the faster, cheaper, but imperfect option of machine translation (MT).
Machine translation is a computer application that analyzes text in a source language and produces an equivalent text in another tongue. Functionally speaking, MT parses text into parts of speech such as nouns, verbs, and adjectives. Then it processes these linguistic components according to linguistic rules, statistical algorithms, or a combination of these methods.
This report consists of three parts, each aimed at helping information publishers determine the suitability of machine translation for their applications:
- Demand. This section will educate organizations considering the use of automated translation technology by answering questions such as "what is MT?" and "how can I determine whether it meets my needs?" It defines machine translation, describes the demands driving the development of automated translation, outlines how it is currently being used in industry and government, and categorizes the major MT features. We also estimate the size of the market for automated translation technology.
- Buying Guide. This part will be useful to organizations evaluating MT as a way to provide information to their customers or constituencies – either by translating foreign-language content on-demand or by incorporating MT technology into their content life cycles. It outlines a step-wise approach for choosing an MT solution based on system architecture, language pair needs, standards, application programming interfaces for integration, and performance factors such as translation throughput.
- Technology Probe. This section is only for the stout of heart. Anyone truly interested in learning more about the science of automated translation technology will find value in our dissection of the four MT approaches most favored by developers. It describes each model, lists its advantages and disadvantages, and outlines the major metrics used by industry, academe, and government to assess the performance of these four approaches.
|
Benefits |
| In General: When used in conjunction with our “Automated Translation Suppliers” report, this document should be most useful to anyone evaluating the use of automated translation solutions. This report will not be useful to information consumers wondering whether they should use automated translation technology. |
| For Buyers: Any organization that produces information for global consumption may find MT useful for filling in the gaps of human translation (HT) or increasing the effectiveness of their HT budgets. Evaluators, decision makers, information technologists, and content management specialists need this information to determine the suitability of MT for their applications and the ability to integrate automated translation into their technology infrastructure. |
| For Suppliers: Some LSPs use MT for pre-processing jobs. Evaluators will be interested in optimization and interfaces for integrating with their core operating and workflow systems. |
Physical Details |
| Authors: Donald A. DePalma and Robert J. Kuhns |
| Date: 15 November 2006 |
| ISBN: 1-933555-31-9 |
| Pages: 59 |
Companies |
| Canadian Parliament, Carahsoft, CIA, Collins, CrossLanguage, DARPA, Duden, European Union, Fujitsu, IBM, I.R.I.S., Language Engineering Corp., LEC, Language Weaver, LogoVista, Meaningful Machines, MetaTexis, Microsoft Research, MultiCorpora, NATO, NIST, Nstein, ORFO, PROMT, SDL, Stellent, SYSTRAN, Verbalis, WorldLingo |
Table of Contents |
- Topic
- Using Machine Translation to Close the Translation Gap
- Structure and Navigation of the Report
- Who Should Read This Report?
- Demand
- The Choice: Human, Machine, or Zero Translation
- Today More Web Users “Pull” MT than Receive Published MT Content
- How Good Does Automated Translation Output Have to Be?
- Market Demand for Local Language Causes Human Translation Gap
- How and Where Organizations Deploy Automated Translation
- Where You Can Find MT in 2006
- More Ambitious Uses of MT Assume Friendly Audiences
- What Does the Future Hold for Automated Translation Applications?
- MT Market: Fifty Years in the Making, But Still Short of Expectations
- Mismatch: Market Expectations versus Economic Value
- The Business Side of Automated Translation
- Strengths of Machine Translation
- Weaknesses of Machine Translation
- Opportunities for Machine Translation
- Threats to Machine Translation
- Conclusion
- Buying Guide
- The Techno-Religious Question: Rules, Statistics, or None of the Above?
- Is Your Content Ready for Automated Translation?
- Software License Costs Range from Pennies to Real Money
- In Which Language Do You Need Your Content?
- Where Do You Need Machine Translation to Run?
- Integration with the Corporate Technology Stack
- Integration with Global Content Life Cycle
- How Can You Improve MT Output?
- The Final Issue: Performance, Sure, But Your Mileage May Vary
- Where Will Automated Translation Suppliers Go in the Next Few Years?
- Conclusions from Evaluating These MT Vendors
- Technology Probe
- Automated Translation Covers Multiple Technologies
- Rules-Based Systems: Pros and Cons
- Statistical MT Systems: Pros and Cons
- Hybrid MT Systems: Pros and Cons
- Context-Based Systems: Pros and Cons
- How MT Specialists Evaluate Automated Translation Software
- BLEU (Bilingual Evaluation Understudy) from IBM Research
- NIST Metric from the National Institute of Standards and Technologies
- F-Measure from New York University
- Which Measure Is Best?
- Appendix
- Table of Figures
- Figure 1: Half of Consumers Use MT to Understand Anglophone Websites
- Figure 2: Zero Translation Dominates, With Most Content Never Getting Translated
- Figure 3: Criteria for When HT and MT Are Appropriate
- Figure 4: Before and After Free Online Translation – Pull MT
- Figure 5: Some Organizations Minimally
|
| Paid Research - Membership Required |
| Automated Translation Technology |
| |
| Keywords |
| access, automated translation, discovery, knowledge management, knowledge transfer, machine translation, natural language processing, translation memory, statistical machine translation, rules-based machine translation, OLMT, MT, consumability |
Abstract |
Of all natural language processing (NLP) technologies, automated translation harbors the most potential as a disruptive technology for global communications and commerce.
Content volume is increasing faster than any company or government can manage, much less translate into all the languages of their employees or external users. Faced with mass quantities of information written in other languages, most organizations choose not to translate most of that content (we call this “zero translation”). Many take the path of high-quality but relatively expensive human translation (HT), picking and choosing just a small subset of their corporate content to translate. A growing number choose the faster, cheaper, but imperfect option of machine translation (MT).
Machine translation is a computer application that analyzes text in a source language and produces an equivalent text in another tongue. Functionally speaking, MT parses text into parts of speech such as nouns, verbs, and adjectives. Then it processes these linguistic components according to linguistic rules, statistical algorithms, or a combination of these methods.
This report consists of three parts, each aimed at helping information publishers determine the suitability of machine translation for their applications:
- Demand. This section will educate organizations considering the use of automated translation technology by answering questions such as "what is MT?" and "how can I determine whether it meets my needs?" It defines machine translation, describes the demands driving the development of automated translation, outlines how it is currently being used in industry and government, and categorizes the major MT features. We also estimate the size of the market for automated translation technology.
- Buying Guide. This part will be useful to organizations evaluating MT as a way to provide information to their customers or constituencies – either by translating foreign-language content on-demand or by incorporating MT technology into their content life cycles. It outlines a step-wise approach for choosing an MT solution based on system architecture, language pair needs, standards, application programming interfaces for integration, and performance factors such as translation throughput.
- Technology Probe. This section is only for the stout of heart. Anyone truly interested in learning more about the science of automated translation technology will find value in our dissection of the four MT approaches most favored by developers. It describes each model, lists its advantages and disadvantages, and outlines the major metrics used by industry, academe, and government to assess the performance of these four approaches.
|
Benefits |
| In General: When used in conjunction with our “Automated Translation Suppliers” report, this document should be most useful to anyone evaluating the use of automated translation solutions. This report will not be useful to information consumers wondering whether they should use automated translation technology. |
| For Buyers: Any organization that produces information for global consumption may find MT useful for filling in the gaps of human translation (HT) or increasing the effectiveness of their HT budgets. Evaluators, decision makers, information technologists, and content management specialists need this information to determine the suitability of MT for their applications and the ability to integrate automated translation into their technology infrastructure. |
| For Suppliers: Some LSPs use MT for pre-processing jobs. Evaluators will be interested in optimization and interfaces for integrating with their core operating and workflow systems. |
Physical Details |
| Authors: Donald A. DePalma and Robert J. Kuhns |
| Date: 15 November 2006 |
| ISBN: 1-933555-31-9 |
| Pages: 59 |
Companies |
| Canadian Parliament, Carahsoft, CIA, Collins, CrossLanguage, DARPA, Duden, European Union, Fujitsu, IBM, I.R.I.S., Language Engineering Corp., LEC, Language Weaver, LogoVista, Meaningful Machines, MetaTexis, Microsoft Research, MultiCorpora, NATO, NIST, Nstein, ORFO, PROMT, SDL, Stellent, SYSTRAN, Verbalis, WorldLingo |
Table of Contents |
- Topic
- Using Machine Translation to Close the Translation Gap
- Structure and Navigation of the Report
- Who Should Read This Report?
- Demand
- The Choice: Human, Machine, or Zero Translation
- Today More Web Users “Pull” MT than Receive Published MT Content
- How Good Does Automated Translation Output Have to Be?
- Market Demand for Local Language Causes Human Translation Gap
- How and Where Organizations Deploy Automated Translation
- Where You Can Find MT in 2006
- More Ambitious Uses of MT Assume Friendly Audiences
- What Does the Future Hold for Automated Translation Applications?
- MT Market: Fifty Years in the Making, But Still Short of Expectations
- Mismatch: Market Expectations versus Economic Value
- The Business Side of Automated Translation
- Strengths of Machine Translation
- Weaknesses of Machine Translation
- Opportunities for Machine Translation
- Threats to Machine Translation
- Conclusion
- Buying Guide
- The Techno-Religious Question: Rules, Statistics, or None of the Above?
- Is Your Content Ready for Automated Translation?
- Software License Costs Range from Pennies to Real Money
- In Which Language Do You Need Your Content?
- Where Do You Need Machine Translation to Run?
- Integration with the Corporate Technology Stack
- Integration with Global Content Life Cycle
- How Can You Improve MT Output?
- The Final Issue: Performance, Sure, But Your Mileage May Vary
- Where Will Automated Translation Suppliers Go in the Next Few Years?
- Conclusions from Evaluating These MT Vendors
- Technology Probe
- Automated Translation Covers Multiple Technologies
- Rules-Based Systems: Pros and Cons
- Statistical MT Systems: Pros and Cons
- Hybrid MT Systems: Pros and Cons
- Context-Based Systems: Pros and Cons
- How MT Specialists Evaluate Automated Translation Software
- BLEU (Bilingual Evaluation Understudy) from IBM Research
- NIST Metric from the National Institute of Standards and Technologies
- F-Measure from New York University
- Which Measure Is Best?
- Appendix
- Table of Figures
- Figure 1: Half of Consumers Use MT to Understand Anglophone Websites
- Figure 2: Zero Translation Dominates, With Most Content Never Getting Translated
- Figure 3: Criteria for When HT and MT Are Appropriate
- Figure 4: Before and After Free Online Translation – Pull MT
- Figure 5: Some Organizations Minimally
|
|
 |