In this work, we demonstrate the usage of SemTUI, an open-source tool to explore and define semanticbased data enrichment operations. The tool is designed to define sequences of data enrichment operations generalizing a link & extend paradigm by integrating external services for end-to-end tabular data annotation, data reconciliation, and data extension. These services are operated through a graphical user interface that helps users explore alternatives and revise the results.
In this paper, we present a demonstration of SemTUI [
An example of enrichment with the SemTUI GUI is described in Figure 1. SemTUI is designed as an open system, where more services can be integrated with limited efort and is released as open source under Apache 2.0 License. Links to GitHub repositories and two demonstration videos are available on the tool documentation page 5.
The main novelty in SemTUI consists in combining five main features, eventually enabling
sequences of data enrichment operations according to the link and extend paradigm: 1)
specification of data enrichment operations and inspection/editing of the results through a GUI
(with a special focus on reconciliation services); 2) support for end-to-end STI; 3) support for
an open set of reconciliation and extension services; 4) support for KG-based enrichment 5)
generalization of semantic-based enrichment to consider third-party APIs.
2The paper presents an improved method for cell entity annotation; for all the other STI tasks, Alligator reuses the
algorithms of s-elBat [
The primary competitors of SemTUI that ofer semantic table enrichment functionalities do
not encompass all five features. OpenRefine 6 served as an inspiration for our work, sharing
many similarities in approach despite its primary focus on data cleaning operations.
Kgextension [
For space constraints, Table 1 provides a comparative summary of SemTUI and other frameworks that deliver semantic enrichment functionalities. Features exclusive to SemTUI, which are not supported by the other tools, are indicated by no. The features labeled partially denote those ofered by the respective tool but enhanced to some extent in SemTUI. Features marked yes indicate functionalities that are comparable to those provided by SemTUI.
For more details, we refer to [
Annotation and reconciliation services. End-to-end STI annotations (computed by clicking on the Automatic Annotation button shown on the top-right corner in Figure 1) are currently served by a selected service (Alligator). Other reconciliation services interoperate following the
7Another limitation of the current version of SemTUI w.r.t. ASIA is the lack of support for manual editing of schema
annotations; this feature is planned for future integration
specification of W3C Reconciliation Service API v0.2 8 and include, among others: geocoding
services based on GeoNames and Here APIs, linking services based on GeoNames APIs and
Wikidata (via Alligator [
Extension services. Extension services are accessed by clicking on a reconciled column that forms the input for the extension service and by specifying in a widget other input conditions (more optional columns) and the features to add in new columns. Extension services currently supported include, among others: Wikidata properties, geographical data (GeoNames attributes, and route information from Here APIs), and meteorological data (from OpenMeteo10).
Visual exploration and edit functionalities. Entities identified by shared systems of identifiers can be inspected by the users. The users can look at the best candidates for each cell, modify the link, or enter a new identifier. Visual codes (colors and shapes) are introduced to communicate the status of reconciliation for each cell (confidence, manual revision, etc.).
Pilot applications. We summarize various pilot applications of SemTUI for data enrichment
tested so far. 1) Enrichment from annotations computed by STI approaches, where additional
data are collected from an annotated table using found links (see the example in Figure 1 or a
short online demo. 2) Enrichment of data from digital marketing campaigns, where historical
performance data is enriched using API services to obtain coordinates and weather data; this
approach streamlines an enrichment use case discussed in prior work [
Demonstration. In the demonstration, we plan to use data from the first and the second scenarios. In the first scenario, we use tables from the SemTab challenges and show an example of end-to-end annotation process with the Alligator STI framework and show examples of entity links found with other reconciliation, e.g., the Wikidata lookup service; we will show how to explore the results and change links that are not correct; finally, we will show examples of enrichment using Wikidata-based extension services. In the second scenario, we use digital marketing data reconciled and enriched with third-party services to demonstrate the servicebased generalization of the link & extend paradigm using APIs. In both scenarios, we will walk the audience through diferent enrichment steps, discussing the features of SemTUI.
This work has been partially funded by the European innovation action enRichMyData (HE 101070284) and the Italian PRIN project Discount Quality for Responsible Data Science: Human-inthe-Loop for Quality Data (202248FWFS) funded by the European Community - Next Generation EU.