Humorous language play in teaching computational engineers.

Popescu, Carmen ; Smoleanu, Oana

1. INTRODUCTION

The present paper deals with the process of language comprehension, i.e. interpreting meaning from texts, in particular from humorous texts. The main assumption is that while humour can be successfully incorporated in the language learning process at all levels, the study of its mechanisms and its impact on computational linguistics can offer a broader understanding of the production or processing of texts for computational engineers, as well as for engineers in general.

Since the early 1990's, there has been a growing interest in the computer generation of humorous texts and small humour-generation programmes have been implemented: the LIBJOG system (Raskin & Attardo, 1994), The Tom Swifty generator (Lessard & Levinson, 1992), the JAPE riddle generator (Binsted & Ritchie, 1994), the HCPP generator (Lessard et al., 2002), The WisCraic pun generator (Mc Kay, 2002), and the HAHA acronym manipulator (Stock & Strapparava, 2002; Attardo & Mele 2002). Ritchie (2004) also mentions the possibility of creating a joke-understander--'a programme which takes as input a possible joke and gives as output a rating (...) of its funniness'.

This area of interpretation seems interesting in the light of schema theory applied to discourse processing, which can further lead to discourse generation. Humour research enumerates 'script-based theories' as one of its important directions.

2. THE FRAMEWORK OF SCHEMA THEORY

Schema theory has its origins in the Gestalt psychology of the 1920's and 1930's. Its basic claim is that 'a new experience is understood by comparison with a stereotypical version of a similar experience held in memory. The new experience is then processed in terms of its deviation from the stereotypical version, or in conformity to it'. (Cook, 1994) The theory can be applied not only to sensory data processing but also to the processing of language. Artificial Intelligence work on text understanding was drawn on by discourse analysis and reading theory and it provided an insight into the way in which knowledge schemata (i.e. mental representations of typical situations, entities, events) are used (i.e. activated) by readers to predict and make sense of the text by drawing inferences. The construct of schema, the basic unit of prior knowledge, is described by Bartlett as'an active organisation of past reactions, or past experiences, which must always be supposed to be operating in well-adapted organic response'. A series of authors, (Semino, 1997; Rumelhart, 1980) consider schemata as playing an active role in all information processing, as basic elements that determine the organization of actions, goals, subgoals, resources, and guide 'the flow of the processing system' (Rumelhart, 1980, quoted in Semino, 1997). Each schema contains several variables, i.e. different realizations of the environment in different applications of the schema and constraints that apply to the variables, i.e. specifications about the entities (persons, objects, situations, events, sequences of events, sequences of actions) that are typically activated by the schema and guide the comprehender's search. For instance, in a text such as the following:

"One of Mr. Smith's neighbours approached one Sunday afternoon and asked: 'I wonder if you are using your grassmower this afternoon.' 'Yes, I am,' Mr. Smith replied coldly. Then the neighbourhood borrower replied: 'Do you think I could use it when you have finished?'"--the schema for BORROWING is activated. We identify four main variables--a borrower, a lender, the use of some object and the method of getting the use of the object needed. In point of constraints, the borrower and the lender will share the feature [+human]. This constraint will change if the students are aware that they are reading a fable, for example, accepting the conventions of this genre in which grasshoppers and ants can talk. Thus the constraint will be altered to [+/-human]. The fact that the lender may or may not be in the possession of the particular object that the borrower needs to use can be taken as the constraint of the variable 'lender' (i.e. +/-possess).

An important function that the variable constraints have is to 'provide default values, (i.e. elements of the schema taken as known or assumed and therefore redundant in communication) for variables that are not specified by a particular input' (Rumelhart & Ortony, 1977; Rumelhart, 1980 quoted in Semino, 1997). These default elements are filled in by the readers according to their expectations about the default elements of the schemata. Semino argues that different default expectations will lead to different interpretations. In the example of the BORROWING schema discussed above, the fact that the borrower needs the use of the object he asks for is a shared expectation about one of its default elements. This will lead to the interpretation of the text as a request-refusal-insistence sequence. If this expectation is not shared by some comprehenders, for some reason which can be traced back to their prior knowledge (for example a reader with a suspicious mind), a certain deviation of the schema can occur and different inferences will be drawn. For instance, that the borrower asks for an object that he does not need and therefore he is a liar. In its turn, this will trigger a different interpretation of the relationship between the lender and the borrower and will throw a different light upon their personalities, their goals and upon the predicted future course of events. Moreover, it will affect the expectations about another variable mentioned, i.e. the method of getting the object. Such a method will no longer qualify as belonging to the bonafide (serious) mode of communication, i.e. 'polite demand', but on the contrary to the non-bona fide communication mode, typical of lying, playing and joke telling (Attardo, 1994). This may lead to a change in the initial schema, so as to adjust it to the new interpretation. In this way the situation can be accounted for, in other words can be comprehended.

Therefore, this whole process of assigning a different default value to some constraint variables in the text mentioned above can also occur, so to say, in retrospect, and therefore can be looked at as a two-way process. That is to say, a modification in a default value can affect the schema just as the activation of an alternative schema can lead to the reinterpretation of one of the default values. When the comprehender identifies his/her false expectation as the source of the schema change, they will know they will have to reprocess the entire text.

3. APPLICATION

An interesting multiple choice exercise can be used in the language class in order to highlight the alterations of variables and scripts in the processing of texts:

The neighbourhood borrower approached Mr. Smith one Sunday afternoon and asked: "I wonder if you are using your grassmower this afternoon?" "Yes, I am," Mr. Smith replied coldly. Then the neighbourhood borrower replied:

1. "Great! Then you won't be needing your gulf clubs. I'll just borrow them."

2. "Do you think I could use it when you have finished?"

3. "You know the grass is greener on the other side."

The selection by the students of answer 1, the correct punchline, proves that the reader got the joke; after forming some expectations, s/he chose the surprise element present in this option, reinterpreted the beginning and 'reconciled ' with it.

The selection of answer 2, a straightforward ending, shows that the reader did not take the text to be a joke, rejected the incongruous element and therefore no 'reconciliation' was necessary. This reader looks for an interpretation that can 'save' the text and make it relevant, true and coherent.

The selection of answer 3, a nonsensical ending (a proverb that may be misleading because of its contained reference to the grass, already mentioned in the text) indicated that the reader went for the surprise element. This may be due to an intuition of the text being a joke, without actually understanding it as such.

This type of exercise can serve as the starting point for presenting and discussing script based theories and their role in artificial intelligence work on the interpretation of texts.

4. CONCLUSION

The presentation of schema theory, accompanied by language practice based on humorous texts such as the ones discussed above could have a double benefit for the language education of engineers: firstly, they could improve their foreign language skills in a more interesting and relaxed teaching context, as some humour studies claim. Reading comprehension is the skill that would benefit most from the incorporation of jokes in the language classroom. Secondly, students could familiarize themselves with incongruities, double meanings, revisions of expectations, and multiplicity of viewpoints, that the interpretation of humorous texts require. Importantly, they could be guided into understanding the relevance of these instances of language play to applications in computational linguistics, which could result in the development of programmes to be used in language education and not only.

As mentioned above, several applications of a computational model of humour are suggested in the literature, e.g.: a machine assisted joke generator or a humour understander that could be successfully used in language education (Mc Kay 2002, O'Mara et al., 2002). Another possible use of a computational model of humour is to make a computer system more friendly and pleasant by the use of humour generation in a user-interface, in the form of humorous remarks from the system which could ease interaction with humans by relaxing the user/learner (Binsted, 1994) stressed by "temperamental" computers. That is why we argue that computational engineers and other professionals in the field might be interested in this rather new area, that joins computers, humorous language play and language education. Cognitive psychology and (computational) linguistics are the disciplines that can offer valuable background information to be used in the development of such programmes. Humour in this context can be regarded as a motivational incentive and can create a link between the transactional and the interpersonal use of language.

In the end, we have to admit that the strategy proposed needs to be developed and improved and further evidence needs to be obtained about the beneficial effect of humour in the language education of computational engineers.

5. REFERENCES

Attardo, S. (1994). Linguistic Theories of Humour. No. 1 in Humour Research, Berlin: Mouton de Gruyter, ISBN 3110142554

Attardo, S. & Mele, F. (2002). HAHAcronym evaluation and application opportunities, deliverable D2, HAHAcronym: Humorous Agent for Humorous Acronyms, IST Programme, Future and Emerging Technologies, European Commission

Binsted, K. & Ritchie, G. (1994). An implemented model of punning riddles, in Proceedings of the Twelfth National Conference on Artificial Intelligence (AAAI-94), Seattle; Cook, G. (1994). Discourse and Literature, Oxford, Oxford University Press, ISBN 0-19-437185-9

Lessard, G. & Levinson, M. (1992). Computational modeling of linguistic humour: Tom Swifties, in ALLC/ACH Joint Annual Conference, Oxford

Lessard, et al. (2002). Proceedings of the April's Fools' day Workshop on Computational Humour, No. 20 in Twente Workshops of Language Technology, Enschede, Netherlands: University of Twente

Mc Kay, J. (2002). Generation of idiom-based witticisms to aid second language learning. Master's thesis, Division of Informatics, University of Edinburgh, Scotland

O'Mara, D. A. et al. (2002). Linguistic-humour and the development of language skills in AAC, in Proceedings of 10th Biennial Conference of the International society for Augmentative and Alternative Communication, Odense, Denmark

Raskin, V. & Attardo, S. (1994). Non-literalness and non-bonafide in language: approaches to formal and computational treatments of humour, Pragmatics and Cognition, 2(1):3169

Ritchie, G. (2003). The JAPE riddle generator: technical specification. Informatics research report EDI-INF-RR0158, School of Informatics, University of Edinburgh

Ritchie, G (2004). The linguistic analysis of jokes, Routledge, London and New York, ISBN 0-415-30983-2

Rumelhart, D.E. 1980. 'Schemata: the building blocks of language'. In R.J. Spiro, B. Bruce and W. Brewer (eds.). Theoretical issues on rendering comprehension. Hillsdale, NJ: Erlbaum, pp 33-35

Semino, E. (1997). Language and World Creation in Poems and other Texts. Harlow: Longman, ISBN 0582 303540

Stock, O. & Strapparava, C. (2002). Humorous agent for humorous acronyms: the HAHAcronym project, in Proceedings of the April's Fools' day Workshop on Computational Humour No. 20 in Twente Workshops of Language Technology, Enschede, Netherlands: University of Twente