Appendix 1: Docuscope’s Architecture of Strings
¶ 1 Leave a comment on paragraph 1 0 The following long note introduces the internal structure of Docuscope and gives details of some of the LATs we discuss in this article. As will become clear, our analysis of Shakespeare’s language based on results from Docuscope is not dependent on the high-level architecture of the program, and it is to some extent also independent of the theoretical-linguistic model used by the creators of Docuscope.
¶ 2 Leave a comment on paragraph 2 0 Docuscope divides all the strings it recognizes between three high-level ‘clusters.’ These clusters correspond to a theoretical model of the effects texts seek to have on their readers developed from a Hallidayan theoretical base (Kaufer et al. 2004: 51-5). The model groups rhetorical effects as follows:
- ¶ 3 Leave a comment on paragraph 3 0
- Internal Perspectives: those strings used to communicate the interior mind of the writer, or a character, to the reader (for example, grammatical first person features, expressive and subjective vocabulary, complex tense/aspect constructions which imply a relation between two different times posited on a narratorial consciousness)
- Relational Perspectives: strings used to connect readers to the representations within a text. For example, assumed shared reasoning and social ties, which refer out of the text, as well as strings which orient readers within a text, pointing forwards or back to items within a text.
- External Perspectives: these strings refer out of the text, but to the physical world (rather than the exterior shared values of ‘Relational Perspectives’) – they include types of description of physical objects, spatial location of described objects, representation of movement through space and time
¶ 4 Leave a comment on paragraph 4 0 A basic hypothesis of the creators of Docuscope is that texts will vary in the frequency with which they employ string types from each of these clusters, depending on the writer’s purpose. Broad expectations would be that fiction and autobiography will be high in cluster 1, while instructional writing (technical manuals for example), will be high in clusters 2 and 3. This is hardly an impressive or surprising finding, of course. The real value of Docuscope comes in the fine-grained analysis possible when comparisons are made at a far more detailed level of string category. Within Docuscope, the three high-level clusters are further divided into six ‘families,’ and then into ‘Dimensions’ and multiple ‘language action types’ (LATs), which allow a high degree of interpretive distinction to be made in the analysis of texts. An example is given below (adapted from Kaufer et al. 2004: 59-88):
- ¶ 6 Leave a comment on paragraph 6 0
- Dimension 1: First person
— LAT: Grammatical first person: e.g. first person pronouns – ‘I,’ ‘me,’ ‘my,’ ‘mine’
— LAT: Self-disclosure: e.g. first person plus simple past – ‘I went’
— LAT: Autobiographical reference: e.g. first person plus habitual past verb phrase – ‘I used to go’
- Dimension 2: Inner thinking
— LAT: Private thinking: e.g. private cognition or thinking verbs – ‘contemplate,’ ‘decide,’ ‘discover’
— LAT: Disclosures: e.g. verbs of speaking, some adverbs – ‘confessed,’ ‘acknowledged,’ ‘personally,’ ‘frankly,’ ‘tellingly’
— LAT: Confidence: e.g. ‘that’-complement, situational ‘it,’ existential ‘there’: ‘I know that the box is upstairs’; ‘It’s a boy!’; ‘There’s an apartment down the street that you can afford’
— LAT: Uncertainty: e.g. adverbials – ‘allegedly,’ ‘to the best of my knowledge,’ ‘nearly,’ ‘almost’
- Dimension 3: LAT: Think positive: e.g. ‘loving,’ ‘succulent’; attitudinally marked prepositions: ‘up’
- Dimension 4: LAT: Think negative: e.g. ‘too many,’ ‘too much’; attitudinally marked prepositions: ‘down’
¶ 7 Leave a comment on paragraph 7 0 Note how the shift from ‘Family’ to LAT allows the analyst to make some very fine distinctions in the stylistic effects produced by texts: within the family ‘First Person,’ Docuscope distinguishes bare first person pronouns such as ‘I,’ ‘me,’ ‘mine,’ which produce a simple point of view within a text, from ‘pronoun + tensed verb’ strings (‘I went…,’ ‘I’ll go…’), which produce a particularised consciousness, self-realised in terms of time. Look at the difference between these examples (adapted from Kaufer et al. 2004: 60):
- ¶ 8 Leave a comment on paragraph 8 0
- I often use facts about Einstein’s laws in my work
- I often used facts about Einstein’s laws in my work
¶ 9 Leave a comment on paragraph 9 0 In the first example, ‘I’ appears with a simple present tense, and establishes that the text is written by a specific individual, but not much else. The second example, however, complicates the point of view presented considerably: the self constructed in the sentence is one which looks back on a past state of selfhood, analyses itself, and discloses something about that analysis.
¶ 10 Leave a comment on paragraph 10 0 Although its Cluster-based model of text function has been carefully developed over a number of years, and has much empirical support from the success of Docuscope as a text analysis device, it remains a hypothesis, open to modification or complete rejection. In fact, all of our conclusions about Shakespeare so far are based on a micro-analysis of relative LAT frequency, rather than assumptions about how the LATs may group into Families or Clusters — our conclusions would not change if theoretical objections altered the make-up or number of Clusters, or rejected them completely.