Bots are going to be big in the near future, period. With Microsoft and Facebook already having stated that chatbots would be the next thing on their agenda for their developments this year, we’re heading into a world where applications will be replaced by bots and mini assistants – thanks to the power of artificial intelligence and complex machine learning algorithms.
Now one thing that was made clear by both Facebook and Microsoft was that bots would pervade the messenger platforms – be it Messenger (in Facebook’s case) or Skype (for Microsoft). While bots would be able to interact with each other and with the user, there’s still a long way to go before their use cases become more practical and viable.
Having said that, let’s look at things from a QS perspective. Maybe bots won’t be able to track your fitness – trackers already do an excellent job and data aggregators and crunching algorithms do the job of getting insight from the numbers. But what about mood tracking?
And that shall be the crux of this article. How bots can make use of complex sentiment analysis and natural language processing to come up with effective mood tracking.
EASING INTO THE BASICS: NLP AND SENTIMENT ANALYSIS
Bots have actually been around for quite some time, with chatbots, chatterbots and SmarterChild doing the rounds of the internet way before the inception of modern assistants such as Siri and Google Now. The way they make sense of words is quite interesting, and most of them rely on sophisticated forms of natural language processing.
Natural language processing, in essence, refers to the way a computer software can derive the meaning out of a human sentence. This process is carried out with intricate tasks in the interim, such as coreference resolution, machine translation and morphological segmentation. In simpler terms, NLP involves near contextual levels of correlation between keywords and referencing them either with a database or experience gathered through machine learning algorithms, in order to come up with human-like responses.
Sentiment analysis is the next step for natural language processing to evolve into effective mood tracking. The onus is to understand the polarity of the user’s commands, and to do this the system often makes use of a scaling mechanism which rates words based on the happy to sad scale. Advanced methodologies include subjectivity / objectivity differentiation, and feature based sentiment analysis.
WHERE DO BOTS COME IN?
Bots have already been used commercially for usage in dialogue systems, building conversational toys, smartphone applications and even providing customer service online. In these use cases, the area of mood tracking is quite limited, since that isn’t the primary objective.
However, bots can actually help in mood tracking in two methods – one would be to actually track the mood, and the other would be to provide valuable feedback and analytics.
For this aspect to actually work effectively, two things must be of absolute importance – either the user must chat with the bot as honestly as possible, or the bot should be capable of monitoring the phone / messenger based conversations that the user has. The latter case would of course have to be done in accordance with the security side of things for QS.
So, how can bots track mood? It all boils down to language processing and sentiment analysis. Long replies, with enthusiastic answers that are given in short time frames would suggest a positive grade on the emotion scale. Short, curt replies would project the user as being in an irritable state, so that would be on the negative side of the scale.
Real time tracking would help the system become better. And more than the tracking part, if corrective measures (as simple as a basic “Calm down”) are taken on the fly, then the system would actually pave the path for more quantifiable mood tracking. Which brings us to the next part of the use case.
DESIGNING THE FEEDBACK MECHANISMS
The Quantified Self has always considered daily reports and real time feedback in the form of graphs and charts to be of essence, and the same would also apply to the modus of mood tracking by bots.
What if the bot could calculate the time you’ve spent online chatting, monitor those chats and effectively deduce for what percentage of the time you were happy, or annoyed. Let’s take things a step further, wherein bots could deduce with whom you are most comfortable in chatting, and with whom you aren’t that eager to talk to.
Okay, so feedback mechanisms could be of a graphical format. But in ideal cases, a bot would be more intelligent than it seems. So, social news feeds could be filtered depending upon the mood of the individual. The next thing in mood tracking would be dynamic content filtering – a sad mood would throw up motivational quotes, a happy mood would produce a funny cat video, a serious mood would produce some interesting news and articles. The scope is endless.
Finally, something as simple as happiness tips at intervals during the day could effectively shape the mood for later. For example, Humblebot is a bot for Slack which delivers simple tips to enable the user to feel better.
And the benefits would be observed almost instantly. Bots making their presence in mood tracking would help in more effective resolution of psychological problems, mood swings, emotional trauma and a healthier lifestyle.
Mood tracking is static for the most part – it records your behaviour over a particular time frame and comes up with report cards. I believe that the addition of bots to the equation would make mood tracking more dynamic, more sentient, and bring forth a more quantifiable lifelogging mechanism.
The screenshots used in this article come from Mood It – a mood tracking mechanism which integrates moods with conversations on social networks. Read more here.
Sign up for our monthly mailing list