Friday, January 15, 2016
Everyone has impressions of their favorite characters from big screen movies and certainly things that they always remember. For example, most Star Wars fans remember Darth Vader’s quote: “Luke, I am your father.” When people recall Yoda, they likely recall someone who is small, green and very, very wise and speaks with a different grammar. If you can communicate like Yoda well, I would assert this would make you wise too. Well, I recently challenged Watson Analytics to understand Yoda speak. As you may recall Yoda turns most sentences into quizzical speech, for example, “Do or Do not, there is no Try. ” Can the cognitive powers of Watson Analytics handle and understand the way Master Yoda speaks? Why is this important? As you know, not everyone communicates the same way. If Watson Analytics is able to understand the user despite the way the user may communicate, the user will have a better experience and be more proficient. If you can do this while making suggestions, this solution adds to the user’s productivity.
In testing this out with Watson Analytics I will use a Social Media Star Wars data set captured by Watson Analytics for Social Media. The dataset includes information around sentiment, geography and author followers among other data. The dataset also includes topics which contain values of “The Force Awakens” and “Star Wars” as shown below.
The dataset additionally includes themes which are a means to breakdown the topics. The Themes for this project are the characters within the movie: Yoda, R2-D2, Darth Vader etc. (Click on the image to enlarge)
Yoda, being as wise as he is, will want to use measures around authors, mentions, documents and followers. Yoda would also be interested in trending or breaking things down by topic, theme, sentiment and geography.
Now here comes the hard part for any software trying to understand the user. Yoda has tough
questions! Can Watson Analytics make sense of it based on what it knows of the data and of what Yoda asks?
If Yoda asks for a mention level breakdown of Topics and Themes, he may inquire with: “The breakdown of the number of mentions by theme by topic, what is”. Below you will see that Watson Analytics assesses Yoda style questions and provides suggestions as to what Yoda really means and perhaps what the data provides as alternate lines of inquiry. The top two suggestions circled nail down what Yoda really wanted. This is a great start for Yoda and for Watson Analytics. Yoda is also getting new ideas of what else to ask with these other suggestions presented. (Click on the image to enlarge)
In this case, Watson Analytics is up to the task with Yoda's last inquiry. However, Master Yoda is yet not done with his inquisition. He would like to know who are the top authors. Naturally he asks, “Top 20 author names by document count, compare? “ And again Watson Analytics provides suggestions. In this case, the first suggestion is what Yoda really wanted. Kudos to Watson Analytics! (Click on the image to enlarge)
The second suggestion from Watson Analytics above leads Yoda to think about mapping authors (spatially speaking). He also wants to learn more about social media influencers as the sixth suggestion hints around the number of followers for authors. The suggestions causes Yoda to ask “The authors by follower count from, where are?” Watson Analytics responds with the following:
(Click on the image to enlarge)
Note that the different suggestions provide an image of what the chart could look like. In the suggestion above, Yoda would expect to see a map heated by author followers. If Yoda clicks on the first suggestion panel, you will see the visualization below. As you can see below the suggestions provided you exactly what to expect. Further there are additional findings to the data around author followers to help Master Yoda ask more questions and get more insights. For example, Yoda sees the top cities with high follower counts on the top left. These highlights are stemming from Watson Analytics understanding of the inquiry as well as the data. (Click on the image to enlarge)
Now Yoda wants to know how people feel about movie The Force Awakens. Yoda asks the question, ”The authors feel sentiment for topic is the force awakens, do?” This may be a confusing question but Watson Analytics is up for the task as shown below. (Click on the image to enlarge)
In the panel above you see how the Topic: The Force Awakens is shown which identifies Watson Analytics is suggesting a filter on Topic for “The Force Awakens” which matches Yoda’s request. This suggestion will provide a bar chart as depicted in the graphic in the suggestion.
Finally, Yoda wants to feel hip and trendy. He wants to understand the trends in the data around the different data sources. Naturally or not, he asks, “The trend over days of the number of mentions by data surce, what is”? Below you will see that the suggestion uses “by Source” rather than data source. Watson Analytics uses the best matching category to match what Yoda meant since there is no data source category in the data but rather “Source”. This can accommodate at points for misspelling, which by the way it did as you can see Yoda made a “typo” in “surce” and still Watson Analytics accommodated for this mishap. Yoda smiles as he again is well understood by Watson Analytics. (Click on the image to enlarge)
Being so old, Yoda’s search for knowledge and truth will take a rest for the day, which concludes this article. With these cognitive results, Yoda points out “The Force is strong with Watson Analytics”. But the key point of this article was to challenge Watson Analytics with some different styles of inquiry to see if it can handle a diverse user base including those from a different planet. Watson Analytics met the challenge. As shown above, Watson Analytics cognitive capabilities can be robust enough to handle Yoda speak. Furthermore the suggestions provide additional lines of inquiry conducive to increased insights. Users will ask more questions and find actionable insights! Try it yourself and May the Force be with you!