『マルテの手記』をテキストマイニングする！ : リルケ散文の計量的分析を用いた相補的研究の試み

Rainer Maria Rilke’s ‘The Notebooks of Malte Laurids Brigge’ (1910) consists of 71 chapters with differing lengths and styles that were written by the protagonist, who wants to be a poet, divided into Part 1 and Part II. Malte continuously repeats the phrase ‘I must change’. Previous studies have attempted to find some ‘differences’ between Part 1 and Part 2 and have argued that there is a ‘change’ in Malte. If so, what kind of change did Malte undergo? And how is the change demonstrated? Since there is no third-person narrator who objectively describes change across the work, it is difficult to recognise explicit change, as the sentences are all spoken in first person. The purpose of this paper is to clarify if there is any change that occurs in this work, and if so, what kind of change it is, by using text mining techniques (multilingual text mining software MTmineR). First, because the work is divided into two parts, a word cloud was used to compare the first half to the second half. In the first half, ‘I’ was the most frequent word, and in the second half, ‘he’. Next, using the random forest method, we examined the style of previous memories and of present events, and found that the style of short novels and of critiques were used in both (the average error rate of Out of Bag data was 5.49%). After that, the 71 chapters were divided by the random forest method into two groups (35 chapters were identified as short stories and 36 as critiques). Results showed that both styles were used in the first half as well as in the second half. Based on these results, the short novel group and the critique group were compared according to content words using the word cloud. The most frequent word in the short novel group was the verb ‘see’, and in the critique group, the adverb ‘such’. ‘That’ was widely used in both groups, and network analysis found that in the short novel group, ‘that’ is often associated with words such as ‘remember’ and ‘feel’. This result partially supports the hypothesis that the first half was written with a subjective narrative, but the second half shifted to a chronological narrative. However, it can be seen that this work is not necessarily divided into two equally weighted halves, but rather two phases of time, past and present, and the two styles of narrative and critique evolve while intersecting in various ways. As described above, a new perspective can be provided by using the text mining technique. However, text mining techniques assist and augment traditional literature research, do not exclude literature research, and do not replace the practice of reading comprehension. What is needed now is to enrich the literary space through the use of text mining, while keeping the limitations of this method in mind.