In Defense of Contextual Vocabulary Acquisition How to Do Things with Words in Context

Researchers in “contextual vocabulary acquisition” differ over the kinds of context involved in vocabulary learning, and the methods and benefits thereof. This paper presents a computational theory of contextual vocabulary acquisition, identifies the relevant notion of context, exhibits the assumptions behind some classic objections, and defends our theory against these objections. 1 A Computational Theory of Contextual Vocabulary Acquisition Contextual vocabulary acquisition (CVA) is the deliberate acquisition of a meaning for a word in a text by reasoning from context, where “context” includes: (1) the reader’s “internalization” of the surrounding text, i.e., the reader’s “mental model” of the word’s “textual context” (hereafter, “co-text” [3]) integrated with (2) the reader’s prior knowledge (PK), but it excludes (3) external sources such as dictionaries or people. CVA is what you do when you come across an unfamiliar word in your reading, realize that you don’t know what it means, decide that you need to know what it means in order to understand the passage, but there is no one around to ask, and it is not in the dictionary (or you are too lazy to look it up). In such a case, you can try to figure out its meaning “from context”, i.e., from clues in the co-text together with your prior knowledge. Our computational theory of CVA—implemented in a the SNePS knowledge representation and reasoning system [28]—begins with a stored knowledge base containing SNePS representations of relevant PK, inputs SNePS representations of a passage containing an unfamiliar word, and draws inferences from these two (integrated) information sources. When asked to define the word, definition algorithms deductively search the resulting network for information of the sort that might be found in a dictionary definition, outputting a definition frame whose slots are the kinds of features that a definition might contain (e.g., class membership, properties, actions, spatio-temporal information, etc.) and whose slot-fillers contain information gleaned from the network [6–8,20,23,24]. We are investigating ways to make our system more robust, to embed it in a naturallanguage-processing system, and to incorporate morphological information. Our research group, including reading educators, is also applying our methods to the developA. Dey et al. (Eds.): CONTEXT 2005, LNAI 3554, pp. 396–409, 2005. c ©Springer-Verlag Berlin Heidelberg 2005 In Defense of Contextual Vocabulary Acquisition 397 ment of what we hope will be a better pedagogical curriculum than the current state of the art for teaching CVA. To this end, we have been studying the CVA literature from a variety of disciplines that, generally speaking, seem to ignore each other’s literature (including computational linguistics, reading education, second-language education, and psychology [22]). Two often-cited papers by reading scientists [2,26] have claimed that not only are certain contexts less than useful for doing CVA, but that most “natural” (as opposed to artificial) contexts are not helpful at all. Their arguments make several assumptions inconsistent with our computational theory. Thus, their objections do not apply to us. 2 Are All Contexts Created Equal? 2.1 The Role of Prior Knowledge Beck et al.’s [2], subtitled “All Contexts Are Not Created Equal”, claims that “it is not true that every context is an appropriate or effective instructional means for vocabulary development”. They begin by pointing out that “the context that surrounds a word in text can give clues to the word’s meaning” (my italics). But a passage is not a clue without some other information to interpret it as a clue. Therefore (A1) Textual clues must be supplemented with other information in order for a meaning to be computed. This supplemental information must come from the reader’s PK. Such PK (which need not be true) might include general “world” or cultural knowledge, “commonsense” knowledge, specialized “domain” knowledge, and perhaps the “background” knowledge the author assumes the reader will have. However, not all of the reader’s PK may be consciously available at the time of reading, and each reader will bring to bear upon his or her interpretation of the text idiosyncratic PK [10,12]. I will use ‘co-text’ to refer to the text surrounding an unfamiliar word, reserving ‘context’ or ‘wide context’ to refer to the reader’s available PK “integrated” with the reader’s “internalization” (or “mental model”) of the co-text. Its integration with the reader’s PK involves belief revision: New beliefs would be inferred as conclusions from arguments in which at least one premise comes from the internalized co-text and at least one premise comes from PK. Typically, withdrawn beliefs are PK beliefs inconsistent with co-text propositions [20]. The “context” that the reader uses to compute a word’s meaning is not just the co-text but this wider context. The reader’s internalization of the text involves some interpretation (e.g., resolving pronoun anaphora) or the immediate and unconscious drawing of an inference (e.g., that ‘he’ refers back to a male or that ‘John’ is a proper name typically referring to a male human) [10]. Consider the following natural passage (my italics): “The archives of the medical department of Lourdes are filled with dossiers that detail well-authenticated cases of what are termed miraculous healings” [17]. Is this to be understood as saying (a) that the archives are filled with dossiers, and that these dossiers detail cases of miraculous healings? Or is it to be understood as saying (b) that the archives are filled with dossiers, and dossiers in general are things that detail cases of miraculous healings? The difference in interpretation has to do with whether “detail . . . miraculous healings” is a restrictive relative clause (case (a)) or a non-restrictive relative clause (case (b)).