This is our blog. We write about cool experiments, interesting case studies, open talks and technicalities.
Grab the RSS feed
Back

2016-04-04

All your topics are belong to us

Jodel is undoubtedly one of the most interesting online communities at the moment. For those of you who have not already become addicted to the Jodel app, Jodel is an anonymous and localized community where users post and react to real-time updates (both text and photos) about anything and everything that is going on in the local community right now. Although the intended target audience is primarily university students, the appeal of anonymous real-time localized information has spread wide outside the student community.

We at Gavagai love interesting text data, and Jodel seems like a veritable goldmine of information to us. Since the good folks at Jodel seem to like analytics (and cool technologies in general), they were kind enough to let us play around with one week’s worth of Swedish Jodel content from november 2015. Here are some of the things we found out.

The perhaps most obvious question to ask when encountering a community such as Jodel is what do people talk about on Jodel? As it happens, figuring out what people write about in text content is one of the things we have built solutions for at Gavagai; in the following example, we use the Gavagai Explorer to extract the top ten topics per day for the Stockholm and the Umeå Jodels.

Topic flow for the Stockholm Jodel

Topic flow for the Stockholm Jodel

The Stockholm topic flow shows a number of interesting things, the most striking of which are two events that took place that week in November: the terror shootings in Paris on Friday, November 13, and the fact that the Swedish soccer team actually made it to the European Championship playoff on Tuesday, November 17. The former is manifested by the topics Paris and IS, while the latter is manifested by the topic Zlatan Ibrahimovic. Other interesting things to note are the following: the topic Ikväll (tonight), is present on weekends, but not on weekdays; and the only two days of the week that are special enough to become topics of their own are Måndag (monday), and Fredag (friday). There are also a number of recurring topics; Känna (to feel), Folk (people), Tjejen (the girl), Killen (the boy), Ligga (to lie down, but more commonly used as slang for getting laid). It seems that people talk a lot about other people, and things you can do with other people. Also, when on Jodel, people talk about Jodel.

Topic flow for the Umeå Jodel

Topic flow for the Umeå Jodel

Although similar to the topic flow of the Stockholm Jodel community, the Umeå Jodel differs in a few distinct ways for the week in question. The Paris events were echoed by means of four different topics, instead of just two: Paris, IS, Muslim, and Islam. Zlatan Ibrahimovic also made an impact in Umeå, as did a host of other topics common with the Stockholm Jodel community. Two noticeable differences between Stockholm and Umeå is that the dating service Tinder reached as high as top 10 in Umeå, as well as the fact that the first snow (storm) of the season took place on Saturday, November 20, which made a strong entry that day (Snö, meaning snow). We refrain from speculating about whether the cold weather in Umeå is a contributing factor to the use of the Tinder app.

An essential feature of the topic analysis done using the Gavagai Explorer is the possibility to expand keywords into concepts (i.e. groups of semantically similar terms). The topics mentioned above are merely labels for sets of near synonyms, as suggested by the Explorer. For instance, the most common topic in the Stockholm Jodel, Fan (damn), denotes a concept consisting of several semantically similar terms (“fan”, “faen”, “satan”, “fy fan”, “jävlar”, “fyfan”, “jäkla”, “jävla”, “jävligt”, “jäääävla”, “jäkligt”, “fyfan va”, “sjuukt”, “jävlas” – i.e. a large number profanity terms that are most commonly used as intensifiers). Another example is the topic Zlatan Ibrahimovic, which consists of the synonyms “Zlatan Ibrahimovic”, “Zlatan”, “Slatan”, “Zlatans”, “Zlaaatan”, and “Ibra”. Without the possibility to (semi-)automatically group these terms into a concept, any analysis of term occurrences would just be a simple frequency list. The ability to recognize semantically similar terms is due to the use of unsupervised semantic memories, which learn semantic relations between words based on how the words have been used. The Gavagai Living Lexicon is a good example of how this technology works, and is the backbone of the near synonym suggestion functionality in the Gavagai Explorer.

One of the most appealing and striking aspects of Jodel is the generally good vibe of the Jodel communities. Since sentiment (or mood, attitude or opinion) is one of the things we regularly measure, this is something we can quantify using Gavagai technology. The following plot shows the hour-by-hour sentiment for the Stockholm Jodel throughout the week of Saturday, November 14 through Sunday, November 20, 2015. The sentiment is normalized with respect to frequency so that the number of Jodels at a particular point in time does not affect the relationship between the individual sentiment values. The analysis was made using the Gavagai API, using our standard set of sentiments, i.e. positivity, negativity, skepticism, fear, love, violence, desire, and hate.

Sentiment analysis of the Stockholm Jodel

Sentiment analysis of the Stockholm Jodel

Note that negative expressions come in strong as the week begins just as the terror events in Paris are broadcast around the world. This, of course, is reflected by the Jodel community in Stockholm. Expressions of negativity, fear, violence and hate are unusually strong, at the expense of other sentiments, as Saturday, November 14 unfolds. On the following morning, the positivity, desire, and love increase in relative value, to reach what seems to be the standard level of the various sentiments throughout the week. Note that positivity, desire and love seems to increase at the end of the period, at the same time as negativity and skepticism decreases. People obviously seem to get a bit more positive and amorous towards the weekends. We can thus confirm that the Jodel community essentially embraces a good vibe, and that weekends are happier times than weekdays (at least for the Jodlers).

Category: case studies, Gavagai Explorer, Gavagai Monitor, sentiment analysis