CALL: Computer-Assisted Language Learning

CALL: Computer-Assisted Language Learning

Computer-Assisted (Language) Learning “Little” programs Purpose-built learning programs (courseware) Using existing technology for educational purposes CALL and NLP Learner corpora 2/14

“Little” programs From earliest days of “microcomputers”, enthusiasts saw ways to implement programs to help learners – Programmed in low-level languages, eg Basic – Crude implementations based on activities which were already part of (language) learning – e.g. vocabulary drills, gap-filling exercises 3/14

“Little” programs Often admirable attempts to use new technology Usually programs were “one-off” – No separation of algorithms and data – Each exercise was a self-contained program Quite easy to “modularise” – have a generic program which would “load” a data file, containing quiz questions and answers 4/14

Issues Content / design determined by technological or pedagogical concerns/issues? – Find some use for technology that is available, or – Design programs to do what you really want Flexibility and reuse – Lot of effort goes into design, so best if design allows for multiple reuse – Notion of “authoring” packages – Allowing multiple correct answers Student-driven learning – Student can work at own time and pace – Role of teacher (if any) very different – Some systems designed for “teach-yourself” scenario 5/14

Typical CALL programs at this level Multiple-choice tests Matching activities Item list learning and testing – Vocabulary test (L1 L2, L2 L1, picture naming) – Writing system (eg Japanese, Chinese characters) Gap filling drills – Grammatical forms (agreement, tenses) – Vocabulary Note difficulty of allowing creative language use, due to need to check right answer – E.g. “compete this sentence with an appropriate adjective” – Alternative allowable answers must be explicitly predicted 6/14

Purpose-built CALL programs “Courseware” Much more than computerized exercises “Typical CALL programs present a stimulus to which the learner must respond. The stimulus may be presented in any combination of text, still images, sound, and motion video. The learner responds by typing at the keyboard, pointing and clicking with the mouse, or speaking into a microphone. The computer offers feedback, indicating whether the learner’s response is right or wrong and, in the more sophisticated CALL programs, attempting to analyse the learner’s response and to pinpoint errors. Branching to help and remedial activities is a common feature of CALL programs.” (wikipedia) 7/14

Stimulus – text, picture, sound, video Lesson plan Learner’s input – typed, spoken, other GUI Is response appropriate? Explanations etc. Feedback to user Student model 8/14

Using existing technology Use in the classroom of technology designed for other purposes – Playing computer games in the L2 – Using L2 word processors, spell checkers and other packages – Speech recognition as pronunciation training – Use of synthetic speech to create spoken language material – Use of MT (mainly to illustrate language differences) 9/14

CALL and NLP What is the role of parsing technology in CALL? – Parsers can allow creative writing to be part of CALL package – Parser as a grammar checker – Parser as an error checker 10/14

Parser as a grammar checker Especially with beginners and intermediate learners, since range of structures and vocabulary is more limited Parsers can (usually) not only say whether a sentence is grammatical, but also why (and where) it is ungrammatical Errors can trigger feedback messages, and can send information to the student model – For example, errors in agreement might indicate that student hasn’t yet grasped this concept, so needs some more instruction 11/14

Parser as an error checker Parsing mechanism can also be used to look for particular (expected) errors “Grammar of errors”: parser has rules which specifically capture ungrammatical sentences If input can be parsed, then there is an error Otherwise, sentence is “correct” (ie no error detected) 12/14

Grammar checking for language learners Long experience of language teaching tells us what errors to expect – Some errors are due to inherent complexities of the language – Other errors are due to interference from a particular L1 13/14

Learner corpora Language teaching meets corpus linguistics several efforts to collect corpora of learners’ writing – Notably International Corpus of Learner English (ICLE), Louvain University several efforts to collect corpora of learners’ writing – Study of “interlanguage” 14/14