Analytics Powered by ML — Watson Explorer Community Edition

Welcome to the first in a series of posts about how we’re embedding machine learning capabilities across our analytics offerings. IBM is at the leading edge of this trend, which comes from a recognition that any product that stores, manages, or governs data will multiply in value for customers if it can also perform machine learning analytics on the data locally and in real time.

That said, the machine learning capabilities can’t just be bolted on the side; they have to be carefully woven into the architecture. Let’s start with Watson Explorer Community Edition, our text mining and content analytic solution.

One of its strengths is an ability to help organizations to uncover and anticipate the underlying reasons for costly, reputation-crippling issues. For example, an auto manufacturer could apply the machine learning capabilities of Watson Explorer Community Edition to mitigate recalls, poor market fit, and angry customers. (By the way those aren’t just nice-to-haves. Currently, the auto industry spends $45 billion every year on recalls and warranty claims. Just for recalls, the average expense per vehicle is close to $600 — a huge hit to their balance sheets.)

Other sectors like retail, finance, and healthcare, might use the capabilities in different ways, but the common thread is uncovering insights into processes at their deepest level, in order to find the inefficiencies and the opportunities.

Watson Explorer Community Edition enables content administration and analytics with the support of three different personas. A system administrator persona uses the admin console to create a data set and to set up a crawler with the data sources. In this new edition, which is available for free download, users can set up their local file systems and directories to crawl while installing the product or after it is installed and configured.

Automotive Accidents: A Use Case

The business analyst/data scientist persona can create collections, analyze content, and save his analysis for further use and sharing. A line-of-business user can consume these insights using the dashboard. Let’s consider the use case where Chad, a data analyst, needs to analyze car accident reports for early quality warning. The first productive use cases are built into the product with guided popups, but I’ll outline the steps here.

1. Chad downloads Watson Explorer Community Edition and installs the product on his laptop (Mac or Windows) in less than 15 minutes.
2. He launches the Content Miner web app from the landing page and logs in. He’s presented with a popup with options to run the Market Analysis tour or the Early Quality warning tour.
3. He chooses Early Quality Warning tour and clicks on the collection. A UI shows multiple facets generated by the tool.
4. Chad picks the facets, applies date range filter and clicks apply.
5. He follows the guided tutorial and selects a car model called “Hill Walker”. The search returns 51 documents.
6. Chad now clicks Analyze trends and anomaly. Not only does the tool show that this particular model of car had higher numbers of accidents in December and January, but the tool also answers “why” by showing Topics and Trends.

Figure 1: Watson Explorer answers not just “what” but “why”.

Now that Chad knows December is the month accidents started to increase, he selects December and clicks Analyze more. Doing so retrieves characteristic words for the model and the accidents. To see more, Chad can select words such as “rust” and “severely” and click the documents that appear to dive deeper into the quality warnings.

Figure 2: Continuing the investigation

Larger trends

The use case just described is typical of larger trends. Like so many other industries, auto manufacturers are embracing tools like Watson Explorer Community Edition hoping to cut costs — and they’ve already had great success. By one estimate, data mining and other track-and-trace software solutions have already cut warranty costs by 20 to 30 percent, administration costs by 25 to 45 percent, and compliance costs by 30 percent.

Those numbers are a testament to the kind of transformation that’s possible with data analytics. As you’ve seen, Watson Explorer delivers insights by combining search and content analytics to help users find and understand the unstructured data that can tell the true story of the business. The current release uses machine learning to do Natural Language Processing (NLP), to cluster documents, and to expand on keyword queries to build out full pictures of what’s at play.

This edition expands potential target users of Watson Explorer by delivering easy-to-use functions and new Content Miner that offers guided navigation to improve the experience for novice business users.

Stay tuned for the next in this series on our efforts to embed machine learning, where we’ll look at a use case for Data classifications and automated Metadata discovery in our Unified Governance portfolio. In the mean time, feel free to download Watson Explorer Community Edition and try it yourself.