Overview
This work aims to shed light on bias in BBC reporting on Palestine in a way that is both transparent and reproducible. We analyzed a total of 600 articles and 4000 livefeed posts on the BBC website between October 7, 2023 and December 2, 2023 in an attempt to surface the systematic disparity in how Palestinian and Israeli deaths are treated in the media.
The pipeline of the study is as follows:
We obtained source articles and livefeed posts from the BBC website by selecting relevant topics (see below for full list) and
We parsed the individual sentences using the Stanford CoreNLP natural language processing
Using the results from step 2, we identified sentences with mentions of death and manually tagged each one of them as referring to Palestinians, Israelis, neither or both. None of the tagging was performed automatically.
I think we should apply the same scrutiny to other websites and reveal their bias as well. But yes, it is interesting to note that killed was used 1066 times for Israel’s 1200 victims and only 1630 times for more than 22K killed by Israel’s strikes, that’s to some extent because every time BBC was reporting on some civilian casualties on the Palestine’s side they were feeling obliged to reiterate the reason of those killings.
Whether killing 22K, many of whom are kids and women in retaliation to 1200 on the other side is justified is a completely different story though
There was another NLP study done three months ago, which this study is based on. That study focused on three other major US newspapers and their bias. Those three were:
The New York Times (seemingly the biggest offender)
The Washington Post
The Wall Street Journal
It would be great to have some kind of unified framework that could ingest the link to a news site and use NLP to fully analyze their language, though I suspect that it’s difficult to fully automate, as the existing studies needed to do some data cleaning by hand.