Towards a corpus for credibility assessment in software practitioner blog articles

TL;DR

This paper introduces a new annotated corpus of software practitioner blog articles, focusing on credibility criteria like argumentation and evidence, to support research in credibility assessment of grey literature.

Contribution

It creates and publicly shares a labeled corpus with high inter-annotator agreement, and demonstrates initial system performance for identifying claim sentences.

Findings

Achieved an F1 score of 0.64 with InferSent and SVM.

Corpus has high inter-annotator agreement of 0.82 (Fleiss' Kappa).

Preliminary results show promise for credibility detection in grey literature.

Abstract

Blogs are a source of grey literature which are widely adopted by software practitioners for disseminating opinion and experience. Analysing such articles can provide useful insights into the state-of-practice for software engineering research. However, there are challenges in identifying higher quality content from the large quantity of articles available. Credibility assessment can help in identifying quality content, though there is a lack of existing corpora. Credibility is typically measured through a series of conceptual criteria, with 'argumentation' and 'evidence' being two important criteria. We create a corpus labelled for argumentation and evidence that can aid the credibility community. The corpus consists of articles from the blog of a single software practitioner and is publicly available. Three annotators label the corpus with a series of conceptual credibility criteria, reaching an agreement of 0.82 (Fleiss' Kappa). We present preliminary analysis of the corpus by using it to investigate the identification of claim sentences (one of our ten labels). We train four systems (Bert, KNN, Decision Tree and SVM) using three feature sets (Bag of Words, Topic Modelling and InferSent), achieving an F1 score of 0.64 using InferSent and a Linear SVM. Our preliminary results are promising, indicating that the corpus can help future studies in detecting the credibility of grey literature. Future research will investigate the degree to which the sentence level annotations can infer the credibility of the overall document.