TENSE GENERATION IN AN INTELLIGENT TUTOR FOR FOREIGN LANGUAGE TEACHING: SOME ISSUES IN THE DESIGN OF THE VERB EXPERT Danilo FUM (*), Paolo Giangrandi(°), Carlo Tasso (o) (*) Dipartimento dell~ducazione, Universita' di Trieste, Italy (o) Laboratorio di Intelligenza Artificiale, Universita' di Udine, Italy via Zanon, 6 - 33100 UDINE, Italy e.mail: tasso%[email protected] ABSTRACT Intelligent Tutoring Systems, in the specific domain of foreign language teaching The paper presents some of the results (Barchan, Woodmansee, and Yazdani, 1985; obtained within a research project aimed at Cunningham, Iberall, and Woolf, 1986; developing ET (English Tutor), an intelligent Schuster and Finin, 1986; Weischedel, Voge, tutoring system which supports Italian and James, 1978; Zoch, Sabah, and Alviset, students in learning the English verbs. We 1986). concentrate on one of the most important modules of the system, the domain (i.e. verb) An Intelligent Tutoring System (ITS, for expert which is devoted to generate, in a cog- short) is a program capable of providing nitively transparent way, the right tense for students with tutorial guidance in a given the verb(s) appearing in the exercises subject (Lawler and Yazdani, 1987; Sleeman presented to the student. An example which and Brown, 1982; Wenger, 1987). A full- highlights the main capabilities of the verb fledged ITS: (a) has specific domain expert is provided. A prototype version of ET expertise; (b) is capable of modeling the has been fully implemented. student knowledge in order to discover the reason(s) of his mistakes, and (c) is able to make teaching more effective by applying different tutorial strategies. ITS technology 1. INTRODUCTION seems particularly promising in fields, like language teaching, where a solid core of facts In the course of its evolution, English has lost is actually surrounded by a more nebulous most of the complexities which still area in which subtle discriminations, personal characterize other Indo-European languages. points of view, and pragmatic factors are Modern English, for example, has no involved (Close, 1981). declensions, it makes minimum use of the subjunctive mood and adopts 'natural' gender In this paper we present some of the results instead of the grammatical one. The obtained within a research project aimed at language, on the other hand, has become developing ET (English Tutor), an ITS which more precise in other ways: cases have thus helps Italian students to learn the English verb been replaced by prepositions and fixed word system. An overall description of ET, of its order while subtle meaning distinctions can be structure and mode of operation has been conveyed through a highly sophisticated use given elsewhere (Fum, Giangrandi, and of tense expressions. Learning correct verb Tasso, 1988). We concentrate here on one of usage is however extremely difficult for non the most important modules of the system, the native speakers and causes troubles to people domain (i.e. verb) expert which is devoted to who study English as a foreign language. In generate, in a cognitively transparent way, the order to overcome the difficulties which can right tense for the verb(s) appearing in the be found in this and several other grammatical exercises presented to the student. The paper areas, various attempts have been made to analyzes some issues that have been dealt utilize Artificial Intelligence techniques for with in developing the verb expert focusing developing very sophisticated systems, called 124 - on the knowledge and processing mecha- cians and people interested in computational nisms utilized. The paper is organized as accounts of language usage (see, for example: follows. Section two introduces our approach Ehrich, 1987; Fuenmayor, 1987; to the problem of tense generation in the Matthiessen, 1984). There is however no context of a tutor for second language agreement on, and no complete theoretical teaching. Section three briefly illustrates the account of, the factors which contribute to ET general architecture and mode of tense generation. The different proposals operation. Section four constitutes the core of which exist in the literature greatly vary the paper and presents the design re- according to the different features that are quirements, knowledge bases and reasoning actually identified as being critical and their algorithms of the verb expert together with an level of explicitness, i.e. which features are example which highlights its main given directly to the tense selection process capabilities. The final section deals with the and which must be inferred through some relevance of the present proposal both in the form of reasoning framework of linguistic studies on verb generation and of intelligent tutoring systems Our interest in this topic focuses on for language teaching. developing a system for tense selection capable of covering most of the cases which can be found in practice and usable for 2. THE TENSE GENERATION teaching English as a foreign language. A PROBLEM basic requirement which we have followed in designing ET is its cognitive adequacy: not An important part of the meaning of a only the final result (i.e. the tense which is sentence is constituted by temporal generated), but also the knowledge and information. Every complete sentence must reasoning used in producing it should mirror contain a main verb and this verb, in all Indo- those utilized by a human expert in the field European languages, is temporally marked. (i.e. by a competent native speaker). The ITS The tense of the verb indicates the relation must thus be an 'articulated' or 'glass-box' between the interval or instant of time in expert. which the situation (i.e. state, event, activity etc.) described in the sentence takes place and the moment in which the sentence is uttered, 3. THE ET SYSTEM and may also indicate subtle temporal relations between the main situation and other ET is an intelligent tutoring system devoted to situations described or referenced in the same support Italian students in learning the usage sentence. Other information can be derived of English verbs. The system, organized from the mood and aspect of the verb, from around the classical architecture of an ITS the lexical category which the verb is a (Sleeman and Brown 1982), consists member of and, more generally, from several essentially of: kinds of temporal expressions that may - the Tutor, which is devoted to manage the appear in the sentence. Moreover, the choice teaching activity and the interaction with the of the tense is determined by other student, information, not directly related with temporal - the Student Modeler which is able to meaning, such as speaker's intention and evaluate the student's competence in the perspective, rhetoric characteristics of specific domain, and discourse, etc. Very complex relations exist - the Domain (i.e. verb) Expert which is an among all these features which native articulated expert in the specific domain dealt speakers take into account in understanding a with by the system. sentence or in generating an appropriate tense for a given clause or sentence. In what follows, in order to better understand the discussion of the Domain Expert, a The problem of choosing the right verb tense sketchy account of the system mode of in order to convey the exact meaning a operation is given. sentence is intended to express has aroused the interest of linguists, philosophers, logi- - 125 - At the beginning of each session, the Tutor While the sentences that are presented to the starts the interaction with the student by student are in natural language form, the verb presenting him an exercise on a given topic. expert receives in input a schematic The same exercise is given to the Domain description of the sentence. Expert which will provide both the correct Every sentence of the exercise is constituted solution and a trace of the reasoning by one or more clauses playing a particular employed for producing it. At this point, the role in it (major clauses and minor clauses at Student Modeler compares the answer of the various levels of subordination). Each clause student with that of the expert in order to is represented inside the system through a identify the errors, if any, present in the series of attribute-value pairs (called exercise former and to formulate some hypotheses descriptors) that highlight the information about their causes. On the basis of these hy- relevant for the tense selection process. This potheses, the Tutor selects the next exercise information includes, for example, the kind of which will test the student on the critical clause (main, coordinate, subordinate), aspects pointed out so far and will allow the whether the clause has a verb to be solved, Modeler to gather further information which the voice and form of the clause, the kind of could be useful for refining the hypotheses event described by the clause, the time previously drawn. Eventually, when some interval associated with the event described in misconceptions have been identified, the the clause, etc. Some of the exercise refined and validated hypotheses will be used descriptors must be manually coded and in order to explain the errors to the student inserted in the exercise data base whereas the and to suggest possible remediations. When a others (mainly concerning purely linguistic topic has been thoroughly analyzed, the Tutor features) can be automatically inferred by a will possibly switch to other topics. preprocessor devoted to parsing the exercise text. For instance, the schematic description of: 4. THE DOMAIN EXPERT ET > EXERCISE-1: The Domain Expert is devoted to generate the 7 (live) in this house for ten years. Now the fight answers for the exercises proposed to roof needs repairing.' the student. Usually, exercises are constituted by a few English sentences in which some of is the following (with the items automatically the verbs (open items) are given in infinitive inferred by the parser preceded by the symbol form and have to be conjugated into an @): appropriate tense.