ZARZA We are Zarza, the prestigious firm behind major projects in information technology.
Listen "Interview with Dr. Trude Heift, Keynote speaker "
Episode Synopsis
Trude Heift Interview Transcripts
Dr. Heift’s website can be found at http://www.sfu.ca/~heift/
WorldCALL 3 website: http://www.j-let.org/~wcf/modules/tinyd0/
Marcel: The first thing I asked her to do was to introduce herself and talk a little about her work and research areas.
Prof. Heift: I was born and raised in Germany and after obtaining a bachelor’s degree and a teaching certification I taught for three years in California before moving to Canada where I completed a Master’s and a Ph.D. degree in linguistics. In 1998, I started a professor position in the linguistics department as SFU [Simon Fraser University], where I currently teach both German language courses as well as theoretical courses in linguistics, including courses on SLA and CALL. My main research interest lies in computer assisted language learning bridging applied and computational linguistics. I am interested in the design as well as the evaluation of intelligent language tutoring systems. These are computer environments for second language learning that make use of natural language processing and techniques of artificial intelligence. From a second language acquisition and CALL perspective, I focus on studies of human-computer interaction such as studying navigation patterns, learner strategies and responses with intelligent systems as well as corrective feedback and error analysis. From a computational point of view, I am interested in automatic analysis of learner language and learner modeling. Learner models try to gather and structure information about the learner by determining their current knowledge state and taking into account learner and task variables. Most of this work involves empirical studies with the ultimate goal of creating a language learning environment that best facilitates learning through truly individualized instruction. In this context, I am investigating what techniques different learners employ when using a language learning system. Most of my CALL software usability studies are carried out with the E-Tutor, an online language learning system for German that is used in German language courses at several North American universities. The system consists of a robust parsing system and provides error-specific learner feedback by performing a linguistic analysis of learner input. In addition, it contains an open learner model that also adjusts and modulates feedback suited to learner expertise by considering individualized help options.
Marcel: Next, I asked her to talk little generally at first about feedback for language learners and what types of feedback are most useful. In addition, I asked her to talk a little about feedback in CALL and how it is challenging.
Prof. Heift: Issues regarding the role and contribution of corrective feedback for language learning have been central to second language acquisition theory and pedagogy. Corrective feedback has received much attention in the oral classroom lately, in particular studies that investigate the effectiveness of recasts. Recasts involve a teacher’s reformulation of a student’s utterance, minus the error, sometimes also referred to as paraphrase. Some studies for instance found that recasts appear to be most effective in contexts where it is clear to the learner that the recast is a reaction to the accuracy of the form, not the content of the original utterance. More general, studies further indicate that the efficacy of corrective feedback in the oral classroom is determined by a number of factors. For example, some research shows that the success of corrective feedback is affected by its format, the type of error, and certain learner characteristics. Of the learner characteristics taken into consideration, verbal intelligence, relative proficiency within levels at school or university, and the learner’s attitudes toward correction proved to be most influential. However, despite this vast interest in studying the role of corrective feedback in the oral classroom, very little research has been conducted for the CALL environment. Due to a difference in modes of instruction, however, the studies and the outcomes will most likely vary for the two respective learning environments and thus independent research in both areas is needed. The limited research that does exist for grammar instruction in CALL, and that includes my own work, generally found that metalinguistic feedback (feedback that explains the type of an error) is more effective than traditional CALL feedback mostly associated with WRONG: TRY AGAIN. The challenging part, however, is to be able to generate this more sophisticated type of feedback. Many simple drills, for instance, are usually based on string-matching algorithms; that is, the student response is compared, letter for letter, against an answer key. Yet one obviously cannot store the infinitely many sentences required for meaningful practice for purposes of comparison and thus more sophisticated answer processing techniques have to be found.
Marcel: And now for my final question. In the outline for her talk, Dr. Heift mentions ICALL. I thought perhaps many listeners may not be so familiar with it so I asked her to explain a little about ICALL and AI and how their use makes teaching and learning different from more traditional methods and from regular CALL?
Prof. Heift: The use of parses in CALL is commonly referred to as intelligent CALL or ICALL, however, it might be more accurately described as parser-based CALL because its intelligence lies in the use of parsing, a technique that enables the computer to encode complex grammatical knowledge, such as humans use to assemble sentences, recognize errors, and make corrections. A parser produces a formal linguistic representation of natural language input by identifying the grammatical functions of the parts of a sentence. With respect to the use of parser-based CALL in teaching and learning, a number of researchers have identified the significant interactive qualities of CALL as one advantage of using the computer in the language classroom. True interaction, however, requires intelligent behavior on the part of the computer. Without intelligence, the system is merely another method of presenting information, one not necessarily preferable to a static medium like print. Instead of multiple choice questions, relatively uninformative answer keys, and gross mainstreaming of characteristic students of workbooks, parser-based CALL is aiming at interactive computer systems possessing a high degree of artificial intelligence and capable of processing natural language input. The strength of NLP, therefore, is that it allows for a sophisticated error analysis where student tasks can go beyond multiple choice questions and/or fill in the blanks and it thus provides the analytical complexity underpinning a parser-based system. However, in addition to focusing on automatic analysis of learner language to provide error-specific feedback, techniques of artificial intelligence have also been applied to student modeling. A student model is any information which a teaching program has, which is specific to the particular student being taught. The reason for maintaining such information is to help the program to decide on appropriate teaching actions with the ultimate goal to achieve an individualized learning environment. The information itself can range from a simple count of how many incorrect and correct answers have been given, to some complicated data structure which, in addition to the student’s knowledge of the subject matter, also includes learner and task variables such as, for instance, learning styles. Accordingly, a student model enables the tutoring system to not only observe, record, and analyze surface phenomena of the learning activity such as text entered by language learners, but also to reason, or at least speculate about the underlying causes of correct as well as incorrect responses. Student models are challenging in a number of ways. For instance, some of the central questions are how to capture, what kind of information, and how to maintain and implement it. Such difficulties have since been the subject of debate and research toward solving some of these thorny problems is well under way.
Dr. Heift’s website can be found at http://www.sfu.ca/~heift/
WorldCALL 3 website: http://www.j-let.org/~wcf/modules/tinyd0/
Marcel: The first thing I asked her to do was to introduce herself and talk a little about her work and research areas.
Prof. Heift: I was born and raised in Germany and after obtaining a bachelor’s degree and a teaching certification I taught for three years in California before moving to Canada where I completed a Master’s and a Ph.D. degree in linguistics. In 1998, I started a professor position in the linguistics department as SFU [Simon Fraser University], where I currently teach both German language courses as well as theoretical courses in linguistics, including courses on SLA and CALL. My main research interest lies in computer assisted language learning bridging applied and computational linguistics. I am interested in the design as well as the evaluation of intelligent language tutoring systems. These are computer environments for second language learning that make use of natural language processing and techniques of artificial intelligence. From a second language acquisition and CALL perspective, I focus on studies of human-computer interaction such as studying navigation patterns, learner strategies and responses with intelligent systems as well as corrective feedback and error analysis. From a computational point of view, I am interested in automatic analysis of learner language and learner modeling. Learner models try to gather and structure information about the learner by determining their current knowledge state and taking into account learner and task variables. Most of this work involves empirical studies with the ultimate goal of creating a language learning environment that best facilitates learning through truly individualized instruction. In this context, I am investigating what techniques different learners employ when using a language learning system. Most of my CALL software usability studies are carried out with the E-Tutor, an online language learning system for German that is used in German language courses at several North American universities. The system consists of a robust parsing system and provides error-specific learner feedback by performing a linguistic analysis of learner input. In addition, it contains an open learner model that also adjusts and modulates feedback suited to learner expertise by considering individualized help options.
Marcel: Next, I asked her to talk little generally at first about feedback for language learners and what types of feedback are most useful. In addition, I asked her to talk a little about feedback in CALL and how it is challenging.
Prof. Heift: Issues regarding the role and contribution of corrective feedback for language learning have been central to second language acquisition theory and pedagogy. Corrective feedback has received much attention in the oral classroom lately, in particular studies that investigate the effectiveness of recasts. Recasts involve a teacher’s reformulation of a student’s utterance, minus the error, sometimes also referred to as paraphrase. Some studies for instance found that recasts appear to be most effective in contexts where it is clear to the learner that the recast is a reaction to the accuracy of the form, not the content of the original utterance. More general, studies further indicate that the efficacy of corrective feedback in the oral classroom is determined by a number of factors. For example, some research shows that the success of corrective feedback is affected by its format, the type of error, and certain learner characteristics. Of the learner characteristics taken into consideration, verbal intelligence, relative proficiency within levels at school or university, and the learner’s attitudes toward correction proved to be most influential. However, despite this vast interest in studying the role of corrective feedback in the oral classroom, very little research has been conducted for the CALL environment. Due to a difference in modes of instruction, however, the studies and the outcomes will most likely vary for the two respective learning environments and thus independent research in both areas is needed. The limited research that does exist for grammar instruction in CALL, and that includes my own work, generally found that metalinguistic feedback (feedback that explains the type of an error) is more effective than traditional CALL feedback mostly associated with WRONG: TRY AGAIN. The challenging part, however, is to be able to generate this more sophisticated type of feedback. Many simple drills, for instance, are usually based on string-matching algorithms; that is, the student response is compared, letter for letter, against an answer key. Yet one obviously cannot store the infinitely many sentences required for meaningful practice for purposes of comparison and thus more sophisticated answer processing techniques have to be found.
Marcel: And now for my final question. In the outline for her talk, Dr. Heift mentions ICALL. I thought perhaps many listeners may not be so familiar with it so I asked her to explain a little about ICALL and AI and how their use makes teaching and learning different from more traditional methods and from regular CALL?
Prof. Heift: The use of parses in CALL is commonly referred to as intelligent CALL or ICALL, however, it might be more accurately described as parser-based CALL because its intelligence lies in the use of parsing, a technique that enables the computer to encode complex grammatical knowledge, such as humans use to assemble sentences, recognize errors, and make corrections. A parser produces a formal linguistic representation of natural language input by identifying the grammatical functions of the parts of a sentence. With respect to the use of parser-based CALL in teaching and learning, a number of researchers have identified the significant interactive qualities of CALL as one advantage of using the computer in the language classroom. True interaction, however, requires intelligent behavior on the part of the computer. Without intelligence, the system is merely another method of presenting information, one not necessarily preferable to a static medium like print. Instead of multiple choice questions, relatively uninformative answer keys, and gross mainstreaming of characteristic students of workbooks, parser-based CALL is aiming at interactive computer systems possessing a high degree of artificial intelligence and capable of processing natural language input. The strength of NLP, therefore, is that it allows for a sophisticated error analysis where student tasks can go beyond multiple choice questions and/or fill in the blanks and it thus provides the analytical complexity underpinning a parser-based system. However, in addition to focusing on automatic analysis of learner language to provide error-specific feedback, techniques of artificial intelligence have also been applied to student modeling. A student model is any information which a teaching program has, which is specific to the particular student being taught. The reason for maintaining such information is to help the program to decide on appropriate teaching actions with the ultimate goal to achieve an individualized learning environment. The information itself can range from a simple count of how many incorrect and correct answers have been given, to some complicated data structure which, in addition to the student’s knowledge of the subject matter, also includes learner and task variables such as, for instance, learning styles. Accordingly, a student model enables the tutoring system to not only observe, record, and analyze surface phenomena of the learning activity such as text entered by language learners, but also to reason, or at least speculate about the underlying causes of correct as well as incorrect responses. Student models are challenging in a number of ways. For instance, some of the central questions are how to capture, what kind of information, and how to maintain and implement it. Such difficulties have since been the subject of debate and research toward solving some of these thorny problems is well under way.
In God we trust