85% of Pak’s tweets on 2nd wave crisis were in support of India | India News
3 min read
[ad_1]
When India was mired in the Covid-19 crisis towards the end of April, #IndiaNeedsOxygen and #PakistanStandsWithIndia started trending in the neighbouring country. It was a strong show of support from the other side of a seven-decade-old adversarial equation.
But just then, #EendiaSaySorryToKashmir started trending alongside. In the end that divisiveness dwindled, shows the first publicly available data set analysing geopolitical relations between the two countries amid a raging pandemic; 85% of the tweets from Pakistan about the Covid-19 crisis in our country used hashtags supportive of India.
The researchers, from Carnegie Mellon University, Maulana Abul Kalam Azad University, and blockchain and AI company Onai, went over more than 3 lakh tweets by 1.5 lakh users between April 21 and May 4 this year. “Majority of the tweets are supportive, with supportive tweets receiving nearly 1.5 times more likes and retweets than non-supportive ones. (Of the tweets in the data set), 85% use supportive hashtags,” lead author Dr Ashiqur Khudabukhsh from Carnegie Mellon University’s Language Technology Institute told TOI.
In the two weeks of Twitter activity in India and Pakistan they studied, they found about 21,600 tweets with the #PakistanStandsWithIndia hashtag and about 3,800 with the #PakistanStandWithIndia hashtag.
#IndiaNeedsOxygen had nearly 20,000 tweets from Pakistan and #IndiaNeedOxygen, 2,400. In comparison, the discordant trending hashtag #EendiaSaySorryToKashmir had just 8,000 tweets from Pakistan and #IndiaSay Sorry ToKashmir , just about 170. “We further notice that tweets containing supportive hashtags that originated in Pakistan received substantially more likes than those from India,” the paper, accepted by the Association for Computational Linguistics Workshop on Natural Language Processing for Positive Impact, said.
With this, Khudabukhsh and his team have taken the conversation from one centred on hate speech to one on hope speech, a project he began in 2019, right after the Pulwama crisis. “Along with the late Prof Jaime Carbonell, Shriphani (Palakodety, co-author) and I have extensively worked on the web-manifestation of political tensions between India and Pakistan in the context of the 2019 India-Pakistan conflict,” Khudabukhsh said. At the time, they had analysed comments on YouTube videos to train a hope speech classifier — a machine learning technique to identify “hostility-diffusing, peace content”.
For this study, they used the hope speech classifier they had created in 2019 along with an empathy-distress classifier trained on responses to news story snippets to identify tweet patterns. Because they had to be cautious — hashtags can be hijacked. “Many of the (tweets) would not contain any text or just an image or a link, or can post unrelated content (e.g. product promotion) or can express content that are not supportive,” Khudabukhsh said. It goes both ways. For instance, someone tweeted, “India deserves this. You are facing what you did to kashmir and fool pakistani supporting india on this you are just slaves to british thats all” with the #IndiaNeedsOxygen hashtag. Likewise, another tweet said, “Political differences have their place but the prayers of us Pakistanis are without Indian brothers and sisters. May Allah give health to all,” with the otherwise negative hastag #EendiaSaySorryTo Kashmir. But their classifiers detect positive tweets 83% of the time, much higher than existing methods.
“While we were working on this project, we had friends and family back in India suffering from the crisis. We are proud to work on this humanitarian project that focuses on national well-being under a healthcare crisis,” Khudabukhsh said. “In this and our earlier study, we found that both countries (India and Pakistan) have web users who behave kindly to each other. In a world fraught with divisive content, I think focusing on positive content to combat hate speech could be an effective strategy.”
But just then, #EendiaSaySorryToKashmir started trending alongside. In the end that divisiveness dwindled, shows the first publicly available data set analysing geopolitical relations between the two countries amid a raging pandemic; 85% of the tweets from Pakistan about the Covid-19 crisis in our country used hashtags supportive of India.
The researchers, from Carnegie Mellon University, Maulana Abul Kalam Azad University, and blockchain and AI company Onai, went over more than 3 lakh tweets by 1.5 lakh users between April 21 and May 4 this year. “Majority of the tweets are supportive, with supportive tweets receiving nearly 1.5 times more likes and retweets than non-supportive ones. (Of the tweets in the data set), 85% use supportive hashtags,” lead author Dr Ashiqur Khudabukhsh from Carnegie Mellon University’s Language Technology Institute told TOI.
In the two weeks of Twitter activity in India and Pakistan they studied, they found about 21,600 tweets with the #PakistanStandsWithIndia hashtag and about 3,800 with the #PakistanStandWithIndia hashtag.
#IndiaNeedsOxygen had nearly 20,000 tweets from Pakistan and #IndiaNeedOxygen, 2,400. In comparison, the discordant trending hashtag #EendiaSaySorryToKashmir had just 8,000 tweets from Pakistan and #IndiaSay Sorry ToKashmir , just about 170. “We further notice that tweets containing supportive hashtags that originated in Pakistan received substantially more likes than those from India,” the paper, accepted by the Association for Computational Linguistics Workshop on Natural Language Processing for Positive Impact, said.
With this, Khudabukhsh and his team have taken the conversation from one centred on hate speech to one on hope speech, a project he began in 2019, right after the Pulwama crisis. “Along with the late Prof Jaime Carbonell, Shriphani (Palakodety, co-author) and I have extensively worked on the web-manifestation of political tensions between India and Pakistan in the context of the 2019 India-Pakistan conflict,” Khudabukhsh said. At the time, they had analysed comments on YouTube videos to train a hope speech classifier — a machine learning technique to identify “hostility-diffusing, peace content”.
For this study, they used the hope speech classifier they had created in 2019 along with an empathy-distress classifier trained on responses to news story snippets to identify tweet patterns. Because they had to be cautious — hashtags can be hijacked. “Many of the (tweets) would not contain any text or just an image or a link, or can post unrelated content (e.g. product promotion) or can express content that are not supportive,” Khudabukhsh said. It goes both ways. For instance, someone tweeted, “India deserves this. You are facing what you did to kashmir and fool pakistani supporting india on this you are just slaves to british thats all” with the #IndiaNeedsOxygen hashtag. Likewise, another tweet said, “Political differences have their place but the prayers of us Pakistanis are without Indian brothers and sisters. May Allah give health to all,” with the otherwise negative hastag #EendiaSaySorryTo Kashmir. But their classifiers detect positive tweets 83% of the time, much higher than existing methods.
“While we were working on this project, we had friends and family back in India suffering from the crisis. We are proud to work on this humanitarian project that focuses on national well-being under a healthcare crisis,” Khudabukhsh said. “In this and our earlier study, we found that both countries (India and Pakistan) have web users who behave kindly to each other. In a world fraught with divisive content, I think focusing on positive content to combat hate speech could be an effective strategy.”
[ad_2]
Source
