Background Processing regimes Processing regimes Detmar Meurers - PowerPoint PPT Presentation

Word order errors & Word order errors & Background Processing regimes Processing regimes Detmar Meurers and Detmar Meurers and Vanessa Metcalf ◮ NLP technology can be used in Computer-Aided Vanessa Metcalf Background Background Language Learning tools that The topic The topic Towards a treatment of word order errors in Word order and FLT Word order and FLT ◮ provide individual feedback on learner errors, Approaches Approaches ◮ foster learner awareness of language forms & categories. Computer-Aided Language Learning List and match List and match Deep processing Deep processing ◮ Very few ICALL systems are used in FLT practice today When to use deep processing — and when not to A downside of mal-rules A downside of mal-rules Our perspective Our perspective (Nagata 2002; Heift 2001). and approach and approach ◮ Problem: lack of interdisciplinary research combining Two types of Two types of Detmar Meurers and Vanessa Metcalf word order errors word order errors computational, linguistic, and FLT/SLA expertise. Phrasal verbs Phrasal verbs Adverb placement Adverb placement The Ohio State University ◮ Our general approach: Summary Summary ◮ Link CL research to genuine FLT needs. References References ◮ Develop task-based systems in support of traditional Large-scale Grammar Development and Grammar Engineering teaching, essentially an intelligent workbook approach. Research Workshop of the Israel Science Foundation ◮ TAGARELA System for Portuguese (Amaral and University of Haifa, Israel, 25.–28. June, 2006 Meurers 2005, 2006) → integration into Portuguese Language Program at OSU in Spring 07 ◮ WERTi System for English (Metcalf and Meurers 2006) → started prototype development 1 / 22 2 / 22 Word order errors & Word order errors & Background Background Processing regimes Processing regimes Detmar Meurers and Detmar Meurers and From word-based to word-order errors in ICALL Word order and Foreign Language Teaching Vanessa Metcalf Vanessa Metcalf Background Background The topic The topic ◮ It is hard to learn word order: Word order and FLT Word order and FLT Approaches ◮ Language learners are known to produce a range of Approaches List and match List and match word order errors (cf., e.g., Odlin 1989). Deep processing Deep processing ◮ ICALL research has largely focused on diagnosing A downside of mal-rules ◮ Word order differs significantly across languages A downside of mal-rules Our perspective Our perspective word-based learner errors (i.e., morpho-syntax). and approach → transfer errors (cf., e.g., Selinker 1972; Odlin 2003) and approach Two types of Two types of ◮ Such approaches can rely on parsing algorithms to ◮ It is important to master word order, especially since word word order errors word order errors Phrasal verbs Phrasal verbs reign in the recursive potential of natural language. order errors can significantly complicate comprehension. Adverb placement Adverb placement Summary Summary ◮ Example from Hiroshima English Learners’ Corpus: ◮ How about word order mistakes, a type of error References References (1) He get to cleaned his son. regularly produced by language learners? → He get his son to cleaned. ◮ Exercise target: (2) He made his son clean the room. 3 / 22 4 / 22

Word order errors & Word order errors & Approaches to diagnosis word order errors Approaches to diagnosing word order errors Processing regimes Processing regimes Detmar Meurers and Detmar Meurers and Instance-based list and match Deep processing: Basics Vanessa Metcalf Vanessa Metcalf Background Background The topic The topic Word order and FLT Word order and FLT ◮ Basic idea: Match user input with listed expected forms. Approaches Approaches ◮ Use grammars, which are compact representations of ◮ matching all or some words, List and match List and match Deep processing the wide range of lexical and word order possibilities. Deep processing ◮ with a complete or partial order, A downside of mal-rules A downside of mal-rules ◮ based on surface forms or lemmata. Our perspective Our perspective and approach ◮ Efficient parsing algorithms are available to license a and approach ◮ Strength: simple and efficient processing Two types of potentially infinite set of strings based on finite grammars. Two types of word order errors word order errors ◮ Weakness: lack of generalization over tokens and patterns Phrasal verbs Phrasal verbs Adverb placement Adverb placement ◮ The additional erroneous word orders can be captured by: ◮ All words for which order is to be checked must be known. Summary Summary ◮ All grammatical orders must be preenvisaged and listed. ◮ extra phrase structure rules (so-called mal -rules, cf. References References e.g., Heift 1998; Fortmann and Forst 2004) → works well for heavily constrained activities, ◮ manipulation of chart edges, the hypotheses introduced ◮ e.g., “Build a Sentence” or “Translation” exercises in by phrase structure rules in a chart parser (Reuer 2003) German Tutor (Heift 2001) 5 / 22 6 / 22 Word order errors & Word order errors & Approaches to diagnosing word order errors Our perspective and approach Processing regimes Processing regimes Detmar Meurers and Detmar Meurers and Deep processing: A downside of mal -rules Vanessa Metcalf Vanessa Metcalf ◮ Word order errors are not uniform: Background Background ◮ Phrase structure grammars express two things at once ◮ some involve lexical triggers (one of a finite set of words The topic The topic Word order and FLT Word order and FLT ◮ generative potential (resource sensitivity, combinatorics) is known to occur) or indicative patterns, Approaches Approaches ◮ word order regularities ◮ others can only be spotted with deeper analysis. List and match List and match Deep processing Deep processing and both are determined at the level of a local tree. ◮ FLT activities are not uniform: A downside of mal-rules A downside of mal-rules Our perspective ◮ some can be set up to include specific lexical material Our perspective ◮ Licensing more word orders can significantly increase and approach and approach or patterns, the search space since the word order possibilities are Two types of Two types of word order errors ◮ in others it is hard to control lexical and structural variation. word order errors directly tied to the combinatorics. Phrasal verbs Phrasal verbs Adverb placement Adverb placement ⇒ Activity-based ICALL systems need a flexible approach ◮ Only local reordering between sisters in a local tree are Summary Summary to word order error detection and diagnosis. achievable through mal -rules. References References Ex. Extending the word order options of S → NP VP by ◮ We want to argue for: adding S → VP NP licenses a. and b., but not c. ◮ choosing processing methods depending on targeted word error type and activity design (3) a. Mary [ loves cats ] . ◮ in deep processing: moving beyond local trees as the b. * [ loves cats ] Mary. units corresponding to errors c. * loves Mary cats. 7 / 22 8 / 22