Digital resources in the Social Sciences and Humanities OpenEdition Our platforms OpenEdition Books OpenEdition Journals Hypotheses Calenda Libraries OpenEdition Freemium Follow us

SIC 5.4.3: Multiscope

Multiscope

Version: 1.0 (2024-04-01)

Multiscope is a web-based dashboard for training and evaluating multi-label text classification models. It provides a complete pipeline for multi-label text classification that covers data splitting, model training and evaluation. The user has the option to both train and test models on datasets of choice.

Specifications

Which Operating Systems does it work on?

Windows

Unix

MacOs

Linux

Licence

Free

Distribution

Local application

User Interface

GUI

Docker instance

No or not known

Does the tool allow you to export the results?

Yes

How is an input document or how is the data to be structured for the tool?

The tool allows for .xlsx, .csv and .json files to be processed.

Excel files and CSV files should contain three columns, namely “text” (contains strings of texts; 1 text per row), “labels” (each row contains a list of label strings or integers) and “subset” (“train”, “val” or “test”).

JSON files should adhere to the following structure:        

                {

                  data:{

‘train’:    [{‘id’: ID_1, ‘text’: TEXT_1, ‘labels’: [LABELS]}, …, {‘id’: ID_N, ‘text’: TEXT_N, ‘labels’: [LABELS]} ]

‘val’:      [{‘id’: ID_1, ‘text’: TEXT_1, ‘labels’: [LABELS]}, …, {‘id’: ID_N, ‘text’: TEXT_N, ‘labels’: [LABELS]} ]  (if present)

‘test’:     [{‘id’: ID_1, ‘text’: TEXT_1, ‘labels’: [LABELS]}, …, {‘id’: ID_N, ‘text’: TEXT_N, ‘labels’: [LABELS]} ]  (remove ‘labels’ if not present in test set)

                        }

                }            

Access

https://github.com/clips/multiscope

How to install, documentation, user guides, workflow description, link to tutorials, … 

The official documentation can be found here: https://www.uantwerpen.be/en/research-groups/clips/research/computational-linguistics/compling-resources/clips-technical-repo/#:~:text=11.%20Multiscope%3A%20A%20User-Friendly%20Multi-Label%20Text%20Classification%20Dashboard

To install the tool, follow the instructions on the GitHub page: https://github.com/clips/multiscope.

Additional information

Category

Content analysis

Author(s)

Jens Lemmens – University of Antwerp (CLiPS) – ORCID ID: 0000-0001-6240-1158

Walter Daelemans – University of Antwerp (CLiPS) – ORCID ID: 0000-0002-9832-7890

Identifier(s)

/

Related tools

Styloscope (stylometry detection pipeline): https://github.com/clips/styloscope

Citation

/