I
have been interested in both the broad array of tools for Web monitoring and
the emergence of practical semantic technology tools. NetBase addresses both of
these interests, so I was very pleased to recently speak with Jonathan Spier,
NetBase CEO, and their CMO, Lisa Joy Rosner. Jonathan said that NetBase now reads
four languages: British, American, Canadian, and Austrailian. Having lived in
three of these countries I can understand the need to treat them separately.
So
I asked what it means to for the software to read a language. Jonathan showed
me an excellent slide reproduced below. A traditional search tool such as
Google treats all words the same. Traditional text analytics uses a dictionary
for sentiment but still treats words in isolation so mistakes can be plentiful.
NetBase understands grammar and picks the pivot words that determine the
context for meaning.
NetBase
was taught English by a team of computational linguistic PhDs. They have now
created several targeted versions. One of these is ConsumerBase that looks at
brands on the Web. It is described as an insight discovery tool that is used for “social
media understanding”. It was
co-developed with several of their Fortune 500 clients, including five of the
top 10 CPG companies in the world, with the goal of making it accessible to the
business user for market research.
ConsumerBase goes beyond
simply monitoring for mentions to providing a software generated understanding
of what is being said. With monitoring tools you need to know what you are
looking for. A tool like ConsumerBase allows you to discover things you did not
know to look for about your brand or other content of interest. It reads over
50,000 sentences a minute, 10 billion documents a month, along with 400,000
social media feeds.
We first looked at what
was being said about the Wii from Nintendo. Below you can see as screen
providing the likes and dislikes connected with the Wii. You can also see sound
bytes. Drilling down on any of the items in either tag cloud will allow you to
go deeper into the content. These
results were generated in a few seconds while we talked. It does not use
pre-categorization but does the categories on the fly to help you find the
unanticipated. It can look at the
same word from a positive or negative sense as it understands the context. For
example, the Wii is connected with the term injury in both positive and
negatives ways. It can be seen as a way to help recovery from injuries (e.g.
regaining balance after traumatic brain injury) and it can be seen as a source
of injury (repetitive strain injury).
In the screen below you
can see a pie chart of the top likes around a product. In this case, it is
Listerne. The text is small to see in this image but the pies in the chart are
labeled with the green area representing the 51% who like it because it kills
germs and smallest slice in blue mentions the 4% who feel it can be a mosquito
repellent. I will have to try this.
You also can break the results
into other categories such as emotions and actions and then have positive and
negatives within these categroies. In the example below we see the mixed
emotions and behaviors connected with the Prius. There seems to be strong love
and hate around this product and these are the top two emotions. Likewise buy
and not buy are the top two actions discovered. Since I have a Prius and a Jeep perhaps I am in a way part of both populations but I love my Prius, as well as my basic Jeep which still gets over 25MPG.
I like the flexibility in data
visualizations. ConsumerBase also takes out potentially irrelevant data. For
example, if you type Comcast into Google you get a lot of returns connected
with the word love. It turns out that much of this is from the presence of
Randy Love a Comcast employee who blogs a lot. ConsumerBase will recognize this
as potentially irrelevant and not include it.
To help with analysis they
have developed a Brand Passion Index that measures the
intensity of consumer passion for brands expressed in social media. In the
image below ConsumerBase used the index to look at grocery stores. We can see the strong passion of love
for Whole Foods and Costco and the strong passion of hate around Walmart.
I like both the technology
and the data visualizations. It is nice to see both aspects done well in the
same product. Jonathan mentioned that clients tell him that the tools can be
addictive and I can certainly see this.
ConsumerBase certainly takes brand monitoring to greater heights and
depths at the same time and it could understand that both attributes were
positive in this sentence.
Comments