Knowledge Exploration is commonly described as the process of obtaining insights within a new domain and is often directed towards a complex open-ended goal.
Effective Knowledge Exploration requires a network structure of interconnected information that makes it easier for the user to sample a large information space, quickly selecting what information to skip and what information to lookup in detail.
While Wikipedia provides a knowledge base of interconnected pages through hyperlinks, these connections are created by hand and generally connect complete articles that each cover large amounts of details about a topic. This makes it hard to find relevant information quickly and often leads to information overload. Furthermore, the manual nature of connecting articles by hand makes this approach not applicable to other large textual datasets.
This work explores multiple methods for automatically connecting text segments from Wikipedia using measures of semantic textual similarity and aspect-based similarity. We define aspect-based similarity as a concept for describing relationships between texts that are not captured by mere similarity measures, such as shared latent attributes or additional relevant content. The results are evaluated in a study that compares the effectiveness of navigating within Wikipedia using Wikipedia links with the effectiveness of textual similarity-based and aspect-based navigation on a complex open-ended task.
Research Questions
Q1: Wha is aspect-based similarity?
Q2: What is the state of the art for semantic text similarity and aspect-based similarity?
Q3: How to create embedding representations that capture the similarity for specific aspects?
Q4: How to evaluate the improvements in knowledge exploration on Wikipedia?
Important References
SBERT: https://arxiv.org/abs/1908.10084
Evolution of Semantic Similarity—A Survey: https://arxiv.org/abs/2004.13820
Exploratory Search - Beyond the Query–Response Paradigm: http://www.iro.umontreal.ca/~nie/IFT6255/Books/ExploratorySearch.pdf
Name | Type | Size | Last Modification | Last Editor |
---|---|---|---|---|
Gerber Antrittspräsentation MA.pdf | 753 KB | 13.10.2022 | ||
Gerber Final Presentation MA.pdf | 2,43 MB | 13.10.2022 | ||
Gerber Final Thesis.pdf | 5,18 MB | 13.10.2022 |