This week’s readings brought me back to my time teaching in Egypt, during the election of Morsi and the coup the following summer that put him out of power. Though my Internet access was limited both summers–largely relegated to communal computers and one dodgy PC in the prep room–I often tried to check in with Twitter.
My second summer, the day of the military takeover, a few tweets entered the stream about tanks in Cairo and the Presidential Palace. I saw journalists and activists positing frantically, while others were trying to get confirmation. No one knew what was happening. For a few days, protestors for Tamarod had taken to the streets against Morsi. Meetings both with and without Morsi went on amid these protests. For my part, the seminary where I was teaching was on lock down, preventing anyone from coming or going without approval. So beyond the nightly sounds of protestors gathering for nearby hot spots, Twitter was my only window–or “stream”–on the action.
I felt surreal during the take sightings. Seeing the news pour in on “real time.” None of the networks had anything, but across Twitter, people were mobilized and locked in.
The memory was significant but largely on the back burner, with grad school taking first seat. This changed somewhat last semester, when I went to a conference and heard Zizi Papacharissi discuss her work with Egypt and Twitter and the “affective publics” she saw influencing the January 25 Revolution. By current standards, the sample-size was small–which she admitted–but the insight about the role of affect in public discourse proved significant.
Around the same time, in a digital humanities class, we were discussing sentiment analysis alongside other big data analysis. Primarily discussing DICTION–as that was the tool in the readings–we all were kind of skeptical about the sort of qualitative claims made from somewhat quantitative methods–like counting the number of “happy” phrases. Moreover, we were also skeptical about the “bag of words” approach that reduces the syntax of the material.
But we were also excited. As Papacharissi–and others–have argued and shown, sentiment, affect, and other emotional qualities do seem significant in social media platforms. And since these platforms exhibit big data, traditional “by hand” coding may not do enough. So I think the key now, in the language of the class, is how one can operationalize sentiment analysis in these spaces, in light of the issues.
For this operationalizing, the readings were helpful for me. The Gaffney and Puschmann piece was broke down the materiality of Twitter. For example, I had never hear of the differences between the usual “spritzer” of Twitter and the different “hoses.” Similar to this, I learned a lot about the different APIs, the sample/filter distinctions, and the breakdown of tools.
Thelwall helped me consider the timetable of affect and sentiment, how one can sample tweets for a given length of time to measure trends. This presented a helpful reason for wanting to do this work. In a similar way, Dang-Xuan et al presents a different approach to use sentiment that still involves time, but is more concerned with the connections of sentiment to other phenomenon–here, retweets in political Twitter conversations.
These articles present contexts and examples of the when and why for this research, but did not salve my concern for the limitations in mass sentiment analysis. Fink et al. helped with this. Though I admit I was lost in the details, I appreciated the organized, focused way that he broke down and worked through the details of different subjectivities, audiences, sentiment, etc., while still acknowledging the limits. And though I did feel a bit lost, I feel like I would need to follow his approach–or at least acknowledge it–if I were to go in this direction.
Increasingly, then, I’m getting better at trying to productively think through and negotiate the scales and layers of research–the excitement of seeing Twitter alive during a coup, the skepticism with general concepts of methods, and the often complex and mundane realities of needing to operationalize. This “research” is its own way of thinking. Of “working” and writing–of being even, do to its breadth.