Introduction

Trinidad and Tobago doesn’t have many political debates or neutral discussions among candidates vying for political office (as opposed to those on political platforms), so when anything that smells like a debate is discussed, noses perk up. The word ‘debate’ apparently sparks fear in the minds and hearts of many a political candidate, so they are often avoided, despite the existence of a Debates Commission. Nevertheless, Tobago, led by Civil Society and Media, had a series of “Town-hall Styled” Forums for the public to ask questions directly to the 2020 candidates putting themselves up for parliamentary elections and this was presented in a local news article.

The present article is more technical in nature. The goals here are, using the data collected on the Facebook page where the events were live streamed:

1. explore the engagement with the public
2. capture questions of interest
3. identify themes in the questions and discussion in preparation for other similar events like the local government election in Tobago/Tobago House of Assembly/THA.

Event Format

The Media House that hosted the series (called ‘Conversations with the Candidates’) was the local multimedia company Tobago Channel 5 (TC5). TC5 live-streamed the three nights of interviews on their Facebook page here, here and here and they also showed it to their cable customers. These events were from 6–9pm on Sunday 2 August (with 3 candidates), 7.30–8.30p on Monday 3 August (1 candidate), and 7.30- 10p on Tuesday 4 August 2020 (2 candidates) (the general election was 10 August, 2020). Each candidate was interviewed alone by 2 moderators for 1 hour and was asked questions from the public sent in via Facebook comments as well as Zoom call and phone call. For the two constituency seats of Tobago, 7 candidates were putting themselves for Election: 2 each from the major parties in Tobago: PNM and PDP, and 3 independent candidates. Of the 7 candidates, only 6 allowed themselves to be interviewed.

Overall Facebook Data and Statistics

The 3 live streamed sessions were viewed almost 24,000 times as of the writing of this article, though this double counts the same person watching multiple videos (for comparison, Tobago has around 65,000 residents). This stream count excludes those viewers who saw the discussions via cable TV (TC5 is a cable company). Collectively 1,467 comments, including responses to comments, were left on those three videos by viewers (several questions were sent via phone and Zoom video, but we have collected no data on these). In collaboration with TC5, these comments were extracted from the webpage, using freely available webtools such as CommentExporter.com and processed using freely available python software by members of the Civil Society team and TC5. Data extracted included comment text, the comment to which each comment was a response if any, commenter Facebook account name, number of likes for the comment, and the timestamp of the comment.

Exploratory Data Analysis

Here we set about examining what insights we might gain from the data focussing on visual outputs. We start with a plot of the number of comments by each commenter.

Ordered count of the number of comments made by each commenter. The most active commenter posted 78 comments, and most comments were made on Day 1: Sunday.

We see above the number of persons that actually commented is only around 250 despite the 24,000 streams (this 24,000 would include some repeat persons though since Facebook would count a single person viewing each of the three different streams as 3 views. However since the first video had 11,000 views, and the others fewer, we can expect that there were at least 11,000 persons viewing online). The plot above shows some key points including that there was a small number of very engaged persons, and a lot of more passive consumers of the event. This is likely to be the case in the real world political discussion where a few persons make more noise about issues than the vast majority of the electorate. Additionally, there was much more engagement on Sunday than the other days. This could be simply that Sunday had the largest number of candidates (3 compared to 1 and 2 on Monday and Tuesday respectively). It could also be due to the public’s curiosity about the new event.

Each dot represents a single comment made at that time stamp (in local time). Along the vertical axis, the dots are spread only to make them easier to see. Blue lines show the approximate start and end times that each candidate’s interview. Note that while on Sunday candidates were consecutively interviewed, on Tuesday, there was a break of around 30 minutes between the two candidates.

The second figure above shows far more activity during interviews of some candidates than others, and it also shows that comments continued long after the live streams ended which indicates much interest in the event series. It could also be due to expatriate viewers in perhaps Europe watching the record of the stream during their daylight hours. Note for instance the green dots around 2pm on Tuesday (to right side of figure) that were made for the Sunday live stream.

Wordcloud of the comments with filler words and local slang etc removed: larger words were used more frequently.

A wordcloud was also prepared from the words of the comments posted (more frequently used words are larger) using a python addon. Many of the comments were brief phrases of support for the candidate onscreen at that time, or in support for one political party or the others. Only a few of them were questioning or discussing issues. Additionally there was much use of slang (both Trini and American). So before this word cloud was developed or further analysis performed on the 1,467 comments, several words were filtered out including filler words (“the”, “an” etc), local slang words (“yuh”, “de” etc) and other words that mainly reflect the event context rather than the question content (‘Tobago”, “candidate” etc). This was performed with the free natural language tool kit for python (nltk) updated with a selection of local words.

The word cloud therefore gives a qualitative sense of the most frequently raised ideas based on quantitative analysis. The big ideas come down to the wants and needs of the people for development based on issues, with a heavy reliance on good government actions. The topic of the new airport currently being developed generated some discussion among the commenters.

Finding Meaning

That’s all great, now we can see the most frequently used words, but that doesn’t give us deep meaning of the ideas expressed. For that we need some more complex tools that can extract information, but first we have a big problem: people misspell a lot. So we need to clean up the misspelling throughout the comments to ensure that we don’t miss any useful commentary (the computer may not know that “platform” and “platforn” should be the same word). For this, we use an automatic spell correction tool (symspellpy) which has large dictionaries of real English words. Since many local words, person names and placenames did not exist in that American-leaning dictionary (like “Roxborough” or the names of the candidates), these were added and the spell check run on the comments.

With a larger comment dataset, we may have been able to statistically extract topics discussed in the comments using what’s called topic modelling, but this didn’t seem to provide any meaningful topics over the 1,467 comments having fewer than 14,000 words, so we will examine instead the conversations that took place by extracting only the comments that were made in response to other comments along with the initiating comment. This revealed some interesting discussions.

Conversations with the Commenters

191 comments with at least 1 response were found. The largest number of responses on a single comment was 11. We’ll call each of these (comment, responses) groups a conversation. After examining the largest or longest conversations, we’ve identified the following topics that generated the most discussion:

1. Tobagonian mentality regarding development
2. Management of the relocation of residents from an area carded for a new airport terminal
3. Tobago Internal self-government bill
4. Developing tertiary educational institutions in Tobago
5. Economic opportunity
6. Agriculture and Fisheries

Other topics discussed included improving the health care system, general evaluation of the incumbent politicians, and the relevance of the challengers. These topics are not an exhaustive list of those discussed by the candidates or the commenters since some topics were discussed on a smaller number of comments dependent on the flow of the live stream interviews in progress. For instance, Tobago is relatively dependent on tourism, so it’s surprising that Tourism wasn’t mentioned above. But note in the Word Cloud that Tourism is shown, but with a relatively small size indicating that it was not discussed much in the comments.

We have extracted the conversation around the these above and have them available for the next event (Tobago local government/ THA elections), to be used as feeders to the question development process.

Civil Discourse

The internet, especially on discussion forums, has become a place where civility is unwelcome, and where mutual respect is a foreign language. It’s where racism, misogyny, homophobia, and general intolerant disrespect are commonplace. The comment area of these live streams was no different. There was a large amount of non-civil discourse with personal criticisms of the candidates and of other commenters along with some racist and sexist comments. Clearly improved moderation in the online forum is needed for future events. Some comments specifically asked other commenters to focus on the issues and keep the commentary clean.

Future Conversations

Nevertheless, it was enlightening to see voters able to interact with their electoral candidates in a relaxed manner. Several comments commended the organizers and celebrated the ability to engage in such an event. There were also already calls for a similar event series for the THA local government elections. With some better management of the behaviour on the online forums, this can become a key and anticipated part of the political landscape in Tobago and in Trinidad.