Language-independent Profile-based Tag Recommendation for Community Question Answering Systems

Community question-answering (CQA) systems are helpful for knowledge sharing. However, they can become difficult to manage as the number of questions and answers increases. Effective tag recommendation facilitates the discovery of relevant material, yet prevailing methods typically depend on language-specific resources or necessitate sophisticated Natural Language Processing (NLP) tools, rendering them unsuitable for less-resourced languages. This paper introduces a novel profile-based tag recommendation strategy that transcends language and structural barriers.  The approach leverages raw text data without the need for complex text mining tools. By constructing distinct profiles for each tag from key terms in associated questions, the method enables a nuanced content association. An adaptation of the Term Frequency-Inverse Document Frequency (TF-IDF) metric is proposed to calculate similarity and recommend tags aligned with these profiles.  The efficacy of this approach is validated across datasets in both English and Persian, showcasing comparable to or superior recall rates against baseline models and contemporary advanced systems.  This methodology is straightforward to implement, offering a valuable tool for enhancing content accessibility in CQA platforms, particularly for low-resource languages.