FITI Experiment Results

This is the data of experiment results for the paper "Xiaochen Li, Domenico Bianculli, Lionel Briand. Tracing Content Requirements in Financial Documents using Multi-granularity Text Analysis. Requirements Engineering (REEN), 2025", including the following files:

FITI results folder:

1. dataset-info.csv: the basic statistics on our dataset
2. feature-info.csv: the detail of features used to train FITI
3. rq-accuracy.csv: results for RQ1
4. rq-component.csv: results for RQ2
5. rq-train-size-p.csv, rq-train-size-r.csv, rq-train-size-f.csv: results for RQ3
6. rq-selection.csv: results for RQ4
7. rq-inspection.csv: results for RQ5

FITI prediction folder: It contains the raw output of FITI for each prospectus file in the test set, which was used to compute the statistical results presented in the evaluation section. The name of each file corresponds to a prospectus. In each file, it shows the sentences identified by FITI for each information type with the following format:

Information type ID

page number: sentence ID, the sentence

Dataset (prospectuses) folder: It contains the non-annotated prospectuses, which were used in this study. They are also available on https://www.bourse.lu at the time this work started.