The Inevitable Wolfram|Alpha Problem: Semantics

Now that Wolfram|Alpha has launched, I thought I’d go take a look. After failing to find out the average happiness of France vs US (not a supported database yet, apparently), I tried out some of the examples and then checked out the “community” area.

Inevitably, Wolfram|Alpha suffers from the same problems that have plagued every attempt to provide answers to human questions using technology since the dawn of computers: semantics, and specifically meaning and syntax. These are nicely illustrated in some of the questions posted in the forum.

Meaning:

• Is Scotland a country, or a town in Connecticut?
• Does “Sun” mean a planet, or is it short for “Sunday”?

Syntax:

• circumference earth / distance london auckland
• Cross referencing doesn’t work

There are technology vendors in both these areas that could bolster Wolfram|Alpha’s capabilities. For example the problem of syntax, although considerable, is now largely solved at least for the types of instructions that Wolfram|Alpha is interested in. And I’m sure it won’t take long before a better database helps the company discover Scotland.

The long-term problem of understanding metadata remains challenging, however – esp…

Now that Wolfram|Alpha has launched, I thought I’d go take a look. After failing to find out the average happiness of France vs US (not a supported database yet, apparently), I tried out some of the examples and then checked out the “community” area.

Inevitably, Wolfram|Alpha suffers from the same problems that have plagued every attempt to provide answers to human questions using technology since the dawn of computers: semantics, and specifically meaning and syntax. These are nicely illustrated in some of the questions posted in the forum.

Meaning:

• Is Scotland a country, or a town in Connecticut?
• Does “Sun” mean a planet, or is it short for “Sunday”?

Syntax:

• circumference earth / distance london auckland
• Cross referencing doesn’t work

There are technology vendors in both these areas that could bolster Wolfram|Alpha’s capabilities. For example the problem of syntax, although considerable, is now largely solved at least for the types of instructions that Wolfram|Alpha is interested in. And I’m sure it won’t take long before a better database helps the company discover Scotland.

The long-term problem of understanding metadata remains challenging, however – especially within organizations. Indeed, most of the effort of implementing business intelligence projects often goes into trying to determining what people are trying to measure – i.e. which data sources need to be connected to each other, and how common business terms should be calculated. It’s one of those areas that exasperate business users: “how hard can it be to give me sales revenue by product?!” – but the IT department understands that the devil is in the details.

For example, at the SAPPHIRE user conference last week, I attended a session where a presenter from Kroger explained that during their BI implementation they had found six different definitions of “sales” and four of “gross profit” – all of them legitimate in certain circumstances. Trying to figure out which one to use when was one of the key challenges of the project.

Once this hard work is done, though, it becomes a very valuable resource that can then be leveraged throughout the organization. For example, a new prototype is available that automatically suggests corporate charts as people type in a PowerPoint slide. It uses the metadata from the existing BI installation, and combines it with SAP BusinessObjects’ Text Analytics technology to provide something that looks a little like magic to the average business user – see the demo below.

Was this interesting? Share with others on Twitter with automatic URL shortening!