Machine learning for improving conversation analysis and optimizing social listening

In a previous Rebel Fridays, Isaac Hernández from Google for Work reminded us of the time a computer beat one of the world’s greatest players in a game of Go. This instance illustrates how we need to understand what machines can give us: if in the past, we thought it was noteworthy that a computer could beat us at chess, how is it possible that a computer can beat us at an even more complicated game just a few years later?

We find out the answer in machine learning: the machine that won playing Go analyzed millions of games to find patterns and use them to play very successful games. How can we use machine learning in activities affecting us?

Improving the categorization or sentiment of our analyses

Machine learning is a subcategory in artificial intelligence (AI) that deals with algorithms that let machines. These algorithms, using data packets, deduce information regarding the properties of those data sets and the derived information lets it make predictions about future information.
This is possible since nearly all non-random data has patterns, and those patterns give the machine the power to generalize. With the aim of concluding, it trains itself using a model that will let it determine the most important parts of that information.

We often find ourselves with social listening or monitoring tools that classify content by sentiments, tags, or categories. According to the application, they can do this manually, through the resources they have by default or with rules that determine certain operations and terms to look up. The manual way is nearly 100% precise but very slow and time-consuming while the automated way is a lot less accurate despite being faster.

How can we improve both? Using machine learning to combine human knowledge and serve it to the machine so it, bit by bit, establishes its procedure and can automatically categorize with a much better precision. MonkeyLearn is an example of a tool we tend to use to do this.

Draw conclusions impacting the business in real time

Natural Language Processing lets the machine “understand” text or voice messages that later go to text channels that are susceptible to analysis. To understand it, consider a simple example in the complex task of filtering email. Let’s suppose that you receive a lot of spam that contains the words “online pharmacy.” As a human being, you can recognize patterns and quickly determine that any message containing “online pharmacy” is spam and should go directly to Trash. This is the generalization that a mental model uses to find out what spam is.
After marking several of these emails as undesired messages, a learning algorithm designed for spam filtering should be capable of making the same generalization that you just did.

Translation: AutoMadrid Dealer: “They treated me perfectly, what I liked the most were the test drive and the way the salesman dealt with me.”

Thanks to Natural Language Processing, besides identifying patterns in casual text or video messages that get sent out spontaneously. For example, a customer tweets about how good or bad they were treated in a car dealership and why they ended up having that experience. Thanks to tools with NLP and machine learning, we can draw concepts and identities (using text mining) that serve as a way of identifying which parts of the experience were positive or negative: the way the salesman treated the customer, the test drive, suggested ticket price, or the atmosphere. Like this one, other cases in different contexts mark the customer journey. For example, how we draw conclusions that impact the business from customer reviews left on hotel comparison sites.

Saving lives

Suicide is the second leading cause of death for people between the ages of 15 and 29, a severe problem impacting daily life. With this in mind, the Universidad de Alicante’s Natural Language Processing and Information Systems Group (Grupo de Procesamiento de Lenguaje Natural y Sistemas de Información) decided to take action: they’ve launched the Life! Crowdfunding Project to develop a tool that can detect feelings or emotions in posts in social media and try to prevent suicides by getting in touch with suicide prevention organizations anonymously.

Looking at the future

Thanks to the application of techniques such as machine learning and Natural Language Processing combined with advances in software and hardware, we can apply the precision of labor done by a person using automation. That means a machine more and more advanced at identifying patterns, saves us time at work, and gives us learning opportunities that let us improve the process analysis that we try to do every day merely using a database with little information that also results in us spending too much time and money to be worth the cost.

By David García Navas.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.