2022 Fiscal Year Annual Research Report
Feature visualizer and detector for scientific texts
| Project/Area Number |
19K00850
|
|---|
| Research Institution | The University of Aizu |
|---|
Principal Investigator |
BLAKE John 会津大学, コンピュータ理工学部, 上級准教授 (80635954)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
Mozgovoy Maxim 会津大学, コンピュータ理工学部, 上級准教授 (60571776)
|
|---|
| Project Period (FY) |
2019-04-01 – 2023-03-31
|
| Keywords | rhetorical features / language features / feature visualisation / genre awareness |
|---|
| Outline of Annual Research Achievements |
We have developed Feature Detection and Feature Visualisation tools as described in the initial proposal. The Feature Visualisation tool comprises an annotated dataset of short research articles and a bank of multimodal materials which are displayed in the user interface of the Feature Visualiser. Here users can access research articles categorized by category, and then within each article, they can visualize particular rhetorical or language aspects, e.g. modality, tense and cohesion. Users then have the option to display additional multimodal explanations to understand the specific rhetorical or language features.
In addition, two Feature Detection tools werecreated that can process student-submitted work. The first focuses on tenses. This was separated into a discrete tool, given its applicability to multiple user groups. The tool colorizes finite verb phrases according to one of twelve pedagogic tenses (i.e. the commonly taught forms, e.g. present perfect progressive). The main feature detection tool enables users to gain feedback on deep grammatical features, namely information structure. The end weight, the information focus and information flow are automatically annotated, helping learners differentiate between unmarked, highly frequent usage and marked, rare usage.
|
| Remarks |
This version of the Feature Detector uses the Django framework, and incorporates two novel functionalities, namely automatic annotation of the coherence and cohesion of each paragraph.
|
