Table of Contents
Current Query-Focused Summarization (QFS) models consider a single input query. Given an arbitrary corpus of text, the query's formulation is burdened with addressing it at various language registers and degrees of specificity.
The Compound Bias-Focused Summarization (CBFS) framework combines the effects of multiple biases, thus partitionning the linguistic and informational constraints. This is analoguous to humans reformulating questions from multiple perspectives or through various language registers for a wider coverage of their audience. Additionally, non-query biases such as sentiment strength can be factored in for Explicative Sentiment Summarization (ESS).
Here, we extend Kazemi et al. (2020)’s Biased TextRank model to Multi-Bias TextRank to demonstrate the CBFS framework.
curl -sSL https://install.python-poetry.org | python3 -git clone https://github.com/croesuslab/lab30-summarization-ahmedm/tree/main/models/MultiBiasTextRank
cd MultiBiasTextRank/
poetry installSee test_mbtr.py and test_ess_mbtr.py.