What is MT good for? Language and Example translations Language - PowerPoint PPT Presentation

Language and What is MT good for? Language and Example translations Language and Computers Computers Computers The simple case Topic 5: Machine Topic 5: Machine Topic 5: Machine Translation Translation Translation ◮ When you need the gist of something and there are no Introduction Introduction Introduction human translators around: Examples for Translations Examples for Translations Examples for Translations ◮ It will help to look at a few examples of real translation ◮ translating e-mails & webpages Background: Background: Background: Dictionaries Dictionaries Dictionaries ◮ obtaining information from sources in multiple before talking about how a machine does it. Transformer Transformer Transformer Language and Computers (Ling 384) languages (e.g., search engines) approaches approaches ◮ Take the simple Spanish sentence and its English approaches Linguistic Linguistic Linguistic Topic 5: Machine Translation ◮ If you have a limited vocabulary and a small range of translation below: knowledge-based knowledge-based knowledge-based systems systems systems sentence types: Direct transfer systems Direct transfer systems Direct transfer systems (1) Yo hablo espa˜ nol. Interlingua-based systems Interlingua-based systems Interlingua-based systems ◮ translating weather reports Adriane Boyd ∗ I speak 1 st , sg Spanish Machine Machine Machine ◮ translating technical manuals learning-based learning-based learning-based Department of Linguistics, OSU systems systems systems ◮ translating terms in scientific meetings ‘I speak Spanish.’ Alignment Alignment Alignment Autumn 2005 ◮ determining if certain words or ideas appear in What makes MT What makes MT What makes MT hard? suspected terrorist documents → help pin down which hard? hard? ◮ Words in this example pretty much translate one-for-one Evaluating MT documents need to be looked at closely Evaluating MT ◮ But we have to make sure hablo matches with Yo , i.e., Evaluating MT systems systems systems that the subject agrees with the form of the verb. ◮ If you want your human translators to focus on References References References ∗ The course was created by Markus Dickinson, Detmar Meurers and Chris Brew. interesting/difficult sentences while avoiding lookup of unknown words and translation of mundane sentences. 1 / 67 4 / 67 7 / 67 Outline Language and Language and Example translations Language and Is MT needed? Computers Computers Computers A slightly more complex case Topic 5: Machine Topic 5: Machine Topic 5: Machine Translation Translation Translation Introduction Introduction Introduction Introduction Examples for Translations Examples for Translations Examples for Translations Background: Background: Background: Background: Dictionaries Dictionaries Dictionaries Dictionaries The order and number of words can differ: ◮ Translation is of immediate importance for multilingual Transformer Transformer Transformer countries (Canada, India, Switzerland, . . . ), approaches approaches approaches Transformer approaches (2) a. Tu hablas espa˜ nol? international institutions (United Nations, International Linguistic Linguistic Linguistic You speak 2 nd , sg Spanish knowledge-based knowledge-based knowledge-based Monetary Fund, World Trade Organization, . . . ), systems systems systems Linguistic knowledge-based systems Direct transfer systems Direct transfer systems Direct transfer systems ‘Do you speak Spanish?’ multinational or exporting companies. Interlingua-based systems Interlingua-based systems Interlingua-based systems Machine ◮ The European Union used to have 11 official languages, Machine Machine b. Hablas espa˜ nol? learning-based learning-based learning-based Machine learning-based systems systems systems systems since May 1, 2004 it has 20. All federal laws and other Speak 2 nd , sg Spanish Alignment Alignment Alignment documents have to be translated into all languages. What makes MT What makes MT What makes MT What makes MT hard? ‘Do you speak Spanish?’ hard? hard? hard? Evaluating MT Evaluating MT Evaluating MT systems systems systems Evaluating MT systems References References References References 2 / 67 5 / 67 8 / 67 Language and Language and Language and What is Machine Translation? What is MT not good for? What goes into a translation Computers Computers Computers Topic 5: Machine Topic 5: Machine Topic 5: Machine Translation Translation Translation Introduction Introduction Introduction Examples for Translations Examples for Translations Examples for Translations ◮ Things that require subtle knowledge of the world Background: Background: Background: Translation is the process of: Some things to note about these examples and thus what Dictionaries Dictionaries Dictionaries and/or a high degree of (literary) skill: we might need to know to translate: Transformer Transformer Transformer ◮ moving texts from one (human) language ( source approaches approaches approaches ◮ translating Shakespeare into Navajo ◮ Words have to be translated. → dictionaries language ) to another ( target language ), Linguistic Linguistic Linguistic ◮ diplomatic negotiations knowledge-based knowledge-based knowledge-based ◮ in a way that preserves meaning. ◮ Words are grouped into meaningful units. (cf., our systems systems systems ◮ court proceedings Direct transfer systems Direct transfer systems Direct transfer systems ◮ . . . discussion of syntax for grammar checkers). Interlingua-based systems Interlingua-based systems Interlingua-based systems Machine translation (MT) automates (part of) the process: Machine Machine Machine ◮ Word order can differ from language to languge. ◮ Things that may be a life or death situation: learning-based learning-based learning-based ◮ Fully automatic translation systems systems systems ◮ The forms of words within a sentence are systematic, ◮ Pharmaceutical business Alignment Alignment Alignment ◮ Computer-aided (human) translation What makes MT What makes MT What makes MT ◮ Automatically translating frantic 911 calls for a caller e.g., verbs have to be conjugated, etc. hard? hard? hard? who speaks only Spanish Evaluating MT Evaluating MT Evaluating MT systems systems systems References References References 3 / 67 6 / 67 9 / 67