Laurent Pacalin
http://www.laurentonmarketing.blogspot.com

but in all seriousness, I like the idea here

Aydin.

We scour the internet for all kinds of opinions data including social media such as blogs, message forums and consumer reviews. We have reached current count of 3.6M reviews alone (not including the blog feeds). We have developed domain agnostic automated technology that has taken a opinion search engine route, so we don’t use any pre-meditated approach to find what you want in the opinions. We visualize and quantify sentiment and buzz but underlying it is a opinion search engine that lets you as customer slice and dice in any which way you want.

For instance. lets take example of shampoos, we let you evaluate products based on their benefit type such as dandruff, dry, oily hair etc, or maybe based on price, or maybe based on size etc. Again everything gets defined automatically through crawling. You can do unfiltered,unadulterated drill down search of opinions in middle of night based on some hypothesis with absolutely no intervention. From our perspective, Consumer is one application and not the “THE” application for our engine.

I did recent interview with SoCalTech for those who are interested.
http://socaltech.com/interview.....12555.html

You can also visit http://www.wisewindow.com or shoot me email at rajiv@wisewindow.com.

We have seen lot of interest generated but rather than bickering and undermining competition, I would love partnership. UNITED WE STAND, DIVIDED WE FALL :-))

Scout’s interface, especially, seems really cool to me. I’m less in awe of their sentiment analysis – looking through some of their examples (specifically for Hillary Clinton), I saw a couple of dramatically miscategorized posts.

I realize that there is a margin of error with all sentiment analysis/NLP programs, though. I can’t tell how their program actually works. Is their program relying primarily only on word co-occurrence?

I’m slightly biased against that approach, probably because I work for a company that relies on a rules-based NLP approach. We use a grammatical tagger to establish tone. You don’t have to have sentence-level or paragraph level tone; companies can just identify what’s being said specifically about them.

If we can help it, we will not compete for creating the best automated sentiment analysis system. There is simply too much good work to be done, whether it be application development or language processing research, to create a compelling application for marketers to be more successful in the changing landscape on the web. As Jenny mentioned in #26 above, we feel that it is much more important WHAT we do with the data, whether we create it ourselves or receive it from others.

I worry about the Demon bride of Twitter. I worry about the industry as a whole the way I worry about family members. Maybe I need to worry more about my consulting practice? Check out this exchange in a presentation I was making at a large, establishe…

Although I believe that they get most of their cool widgets and analysis from a company called Lexalytics (www.lexalytics.com).

You should also check out their political trends website (www.politicaltrends.info) as well. Their CEO Jeff Catlin was interviewed about it on Fox News yesterday. This is truly cutting edge stuff.

Mike, thanks for the interesting article and good luck to scoutlabs

Some companies are putting smart technology in the hands of consultants to deliver insight for companies. Scout Labs has chosen to put smart, easy to use analysis tools the hands of teams of users (people at companies and their agency partners). Scout Labs offers real-time, automated analysis, insight and trend-spotting — moments after hitting “Go Scout”. But then users can append insight and make connections for their fellow team members to see, which not only adds another layer of insight, but it actually improves our technology, as our algorithms learn from and improve based on explicit and implicit user actions.

If you go to our site, you will see that one of the things “we believe” is: Let technology do what it does best, and people do what we do best. Together, we’re a pretty good team.

We’ve architected our entire system with this in mind.

But Arnaud is absolutely right. The glory days of controlling the conversations are over and there is a near-universal need for tuning in to the voice of the people. So, there is plenty of room for different approaches (and price points). In any case, looks like Scout Labs is going to shake things up a bit

Point #2: this is a nice interface with some good usability, but these are just the first layer of metrics. Probably OK for a small company but for a big brands the post volume will quickly overwhelm the user’s ability to really understand the gist of the conversation. These data tell you what is going on, but to be actionable you must dig in on the why behind this. This takes a lot more time that most marketers simply don’t have.

Point #3: the first stage of any social media strategy should be to listen and understand what consumers are saying, so the more that can be done to get marketers into listening mode, the better for the entire industry.

Michael: thanks for bringing this to the TechCrunch audience’s attention!

The Computational scoring of the wild corpus is a science that is truly in its infancy; it just so happens that the first services out of the gate mimic the brand agency offerings of the old guard. There are implications for any number of forecasting models using stochastic means, and new metrics are, as we speak, being developed in the world’s research kitchens.

My redacted report on these agency offerings is available at: http://bizcast.typepad.com/cli.....ntsect.pdf

I haven’t seen the chart and haven’t seen everybody’s capabilities and UI. This said, the market is just emerging and there is clearly room for lots of applications and models. Buzzlogic, Andiamo Systems, Visible Technologies, nStein, Collective Intellect, Umbria, Nielsen BuzzMetrics, Sentimetrix, Cymfony, Biz360, Clarabridge and a few others are still tweaking there models, pitches, technology, target markets, applications, messaging … still early!

Spam, i.e. content seeded by social marketers (pay-per-post) is a challenge. Scalability is also a challenge as Peter mentioned; having human reviews of all these comments for scoring and semantic categorization won’t scale given the explosion and diversity of user generated content. This said, there is no reason for sarcasm, irony and double negatives not to be solved and properlyinterpreted. Algorithms have done harder things in the past and every campus seems to have a computational linguistic program these days. Sentiment Analysis draws on computational linguistic to transform unstructured online dialog into marketing insights. Brand monitoring is only one application of sentiment analysis; there will be many more, maybe all the way to predicting financial markets if data mining can start making sense of emotions.

Progress around tone polarity scoring (beyond just +/-), entity extraction, semantic categorization broken down by whatever … reach, influential segments, geo-mashups, temporal trends, … will turn branding into a more predictable business, accelerating the shift of budgets online; not unlike paid search, contextual placement, demo and behavioral targeting have over the past years.

-arnaud