5 Ways Sentiment Analysis is Ramping Up in 2009
The New York Times has an article today about sentiment analysis, a trend which has been accelerating on the back of the Real-time Web - and Twitter in particular. Sentiment analysis is no short-term hot trend. It will eventually become a key feature of search engines, which will integrate the aggregate sentiment of the crowd into search results.
The NY Times article looked at 3 sentiment analysis tools: Scout Labs, The Financial Times' Newssift, and Jodange. It also mentioned 3 Twitter apps: Tweetfeel, Twendz and Twitrratr. In our post we take a look at five other examples of how sentiment analysis is starting to ramp up on the Web. We invite you to add more examples in the comments.
1. Social Media Monitoring and Analysis: Sysomos
Sysomos launched its two flagship products in June: MAP ('Media Analysis Platform') and Heartbeat. As Frederic Lardinois wrote: both products are powerful (but costly) social media monitoring and analysis tools.
MAP is a powerful and flexible analysis tool. It gives its users the ability to research any topic on blogs, social media sites, and in traditional news media reports. MAP can, for example, tell you that the largest number of Twitter users who wrote about the Palm Pre come from California and Great Britain. It can also quickly give you a history of when and how often a Wikipedia article was edited, or what the most popular forum posts and YouTube videos about the any given topic were in the last 30 days. Heartbeat provides a subset of MAPs features, with a focus on making it easy for companies to track social media metrics.
See also: This Machine Eats Tweets: The System Behind @Comcast and Others
2. Conversation Monitoring: Backtype
As Marshall Kirkpatrick noted in a post from April, a whole class of technologies are emerging to help companies keep track of the conversations exploding online. For example, Backtype is an online tool that lets you search for and monitor keywords across the Web in an effort to put an end to 'comment fragmentation.' It's also used under the hood at Radian6.
3. Mood Analysis: LiveJournal & MoodViews
In his recent analysis Could Real Time Information Be An Unfair Advantage?, Marshall Kirkpatrick explored the connection between communication on social networks and real-world events. As an example, he pointed to a tool called MoodViews which can correlate mood messages on LiveJournal with world events. Some of the trends noted using this tool:
- Mass increase in the level of worriedness around major weather phenomena, such as hurricane Katrina on August 29, 2005
- Excitedness around global media and culture events, such as the release of a new Harry Potter book on July 15, 2005
- Mass increase in the level of distress and sadness after terror attacks, as witnessed by the response to the London bombings on July 7, 2005.
MoodViews updates every 10 minutes based on LiveJournal posts. As of writing, these moods were "hot": irritated, stressed, thoughtful, cheerful, creative.
Hope was big in November 2008. Image: omnicam
See also: Feel Good: Top 10 Mood Apps and Visualizers
4. Semantic Search: Evri's New Sentiment API
Earlier this month semantic search engine Evri released a new sentiment web API that claims to understand how the web feels. As Sarah Perez wrote: while busy scouring the net for people, places, and things and determining the relationships between them, the search engine is now able to understand the feelings associated with these entities, too, be them positive or negative. Using the API, developers can build applications for things like market intelligence, market research, sports and entertainment, brand management, product reviews and more.
The sentiment API does much more than most Twitter sentiment trackers - it allows you deeper insight into the "who's," and "what's," and "why's" associated with the particular expression or feeling.
5. Twitter as Zeitgeist
How can we do a round-up of sentiment analysis tools without mentioning Twitter? Marshall Kirkpatrick wrote in March about the Twitter platform turning 3:
"Twitter already 'breaks news' faster than traditional media outlets on a regular basis and monitoring the ebb and flow of conversation is helping media, marketers and academics put their finger on the pulse of a significant number of people.
We've seen how breaking large numbers of Twitter followers into topical groups can help make the service much, much more useful. We expect to see services launched soon that will take the pulse of topical groups. Bubbling up hot topics early in the world of physicists on Twitter, real estate agents, stock traders, etc. is a powerful tactic that more than one company will cash in on. We'll all benefit when that happens, too."
See also: 7 Top Twitter Topic Trackers
Top image: .michael.newman.
Share
Facebook
Comments
Subscribe to comments for this post OR Subscribe to comments for all ReadWriteWeb posts
Here's another Sentiment Analysis app for Twitter: http://tweetsentiments.com
PostRank (formerly AideRSS) announced Sentiment Analysis as part of our data streaming API's earlier this month.
Details here: http://blog.postrank.com/2009/08/13/postrank%E2%84%A2-releases-real-time-engagement-data-apis-and-adds-sentiment-analysis-to-real-time-content-api/
I have been using www.tweetfeel.com to get an idea on how people feel about movies and music before I go watch or listen - so far it has proven pretty reliable.
Timely post. Twitter offers individuals and brands alike a depth and breath of information (I think) never seen before. Buddy Media is experiencing phenomenal traction with a Twitter product in conjunction with custom facebook pages, and custom applications. Understanding the conversations that are occurring (in real time) about a brand is very powerful information.
Very interesting article on 5 Ways Sentiment Analysis is Ramping Up in 2009
Very informative.. looks like I missed a lot of these great tools, thanks for sharing.
I use Twazzup as part of my daily work. It's a Twitter search tool that provides context by including relevant blog posts. I get a better feel for a topic with Twazzup than I do with Twitter search.
Twazzup: http://twazzup.com
Hi
I've recently published the e-Reputation Market Map with a lot of tools using sentiment analysis : http://bit.ly/uezl6 Spread the word :-) Thanks
http://www.demainlaveille.fr/2009/08/24/e-reputation-market-map-carte-du-marche-de-la-e-reputation/
This cas be also added to the e-reputation market map :
http://www.demainlaveille.fr/2009/08/24/e-reputation-market-map-carte-du-marche-de-la-e-reputation/
Samepoint.com provides social media search and RSS feeds for monitoring online brand reputation and includes positive/negative word analysis. BTW, it's free.
I was surprised that no mention was made of OpenAmplify, a web service that provides sentiment analysis (among other things) of a given text.
I've used it to power TweetyPants.com (analyzes a particular user's sentiments on recent topics), and have also written a module for Drupal that provides the tools necessary to automatically integrate sentiment analysis into the Drupal CMS.
Very well-done. Iimely and useful. Thanks!
Thank you for sharing, I thought.
Replica breitling
Replica TAG Heuer
Replica Watches
Richard,
Thanks for mentioning Sysomos as another company offering ways to measure sentiment within social media. For companies that want to do social media monitoring and measurement, Heartbeat is a cost-effective service with a lot of flexibility such as the ability to do unlimited queries.
cheers, Mark
I will be checking out Sysomos and twazzup as anything making blogging easier is a welcome thing!
looks like I missed a lot of these fantastic tools, thank you very much :)
moodspin --> http://www.moodspin.com also allows you to easily share your moods in to check out your friends status.
Also check www.amplifiedanalytics.com. We decided to focus translation of customer reviews into actionable metrics and go at least a level deeper in sentiment analysis, to measure not only the direction (positive vs. negative), but also a reference of the sentiment to a root cause. We are currently testing our first product - PRD (Product Reputation Detective" and hope to be included into the next review of companies in this space, when a time comes.
Interesting article. Congrats to Scout Labs for getting some press in the NY Times and re-igniting this conversation about sentiment analysis. We've been in the space a while along with Cymfony & BuzzMetrics and one of the key areas to discern actionable insight in order to glean consumer insight analysis and engagement from social data is to have sentiment analysis. Earlier this year Forrester research analyst, Suresh Vittal, published an updated WAVE on the largest players in the space and a key criteria was sentiment measurement capabilities. Feel free to download a free copy here: http://bit.ly/14DXC
@bcahill
svp of marketing
@visible_tech
The huge weakness of this whole approach is how easily it is gamed. These are nothing more than sophomoric word filters, that are so easy to spam, any script kiddy can do it.
Excellent article, excellent comments! That's why I like RWW so much :-)
Richard,
Great look at some of the sentiment stuff coming down the pike. We're particularly excited about some of the possibilities that the semantic web is showcasing; bridging the distance between human judgment and mechanization will always be a fascinating and hugely valuable discussion. I think we'll always need a human element in order to properly draw conclusions from the analysis, but the more we can help take the first pass and eliminate some of the heavy lifting, the better off we'll be.
Stay tuned for more movement on the sentiment stuff from us in the near future.
Best,
Amber Naslund
Director of Community, Radian6
@ambercadabra
Here you go and thanks for all!
glad tools are helping measure sentiment - but the real challenge is for companies to then act about it.
as I wrote this morning
My concern with Social CRM is we will build better antennae and pick up even more customer signals. But unless we have passionate (and empowered) employees who can follow up and do something about it, we will gradually turn off our advocates. And go back to traditional CRM – hope our marketing and PR dollars drown out the non-advocates.
at http://dealarchitect.typepad.com/deal_architect/2009/08/customer-advocacy-and-abdication.html
It's so interesting. I believe human better than machine.
Brilliant article. Understanding what your market is doing now is far more important than what it was doing last week.
Interesting article and it´s nice to see that the NYTimes pay attention to the field of sentiment analysis.
Sentiment Analysis is a field that is growing (and will continue to grow), not only in the US.
My own company in Sweden Saplo (www.saplo.com) have shown astounding results when it comes to semantically understand the meaning of a corpus (text). In comparison with an analyst at a leading media monitor company our algorithm is about 80-90% accurate, which is better results then the human analyst (80%).
While awaiting the launch of Saplo´s Sentiment API we offer online media companies to automatically tag their articles with key words and also match them with other interesting articles.
There are several other companies in Sweden worth mentioning like Aitellu (www.aitellu.com) who mostly rely on AI to understand text and Twingly (www.twingly.com) Scandinavias (the worlds?) leading blog search engine.
Cheers
Peter Larsson
@saplo.com
Hi,
I agree that this type of information is important. It also seems to miss a large part of the picture. If you use the following set of terms to track a particular guys. It will make you search much more effective Shaq, Shaquille O'Neal, Cavs center, the cavalier center. I have built a tool which helps create a search for sets of terms.
Users can add a Newssift sentiment widget to their iGoogle page at http://www.google.com/ig/directory?type=gadgets&url=widgets.newssift.com/widgets/igoogle/sentiment.xml
I was also pleased to see the NYTimes coverage of sentiment analysis. I've been tracking the growth of the sentiment analysis and natural language processing field since last year and the coverage generally has been exploding of late.
My own company, Adaptive Semantics, offers a sentiment analysis product called JuLiA (Just a Linguistic Algorithm) for publishers of online content. We automate comment moderation based upon sentiment (abusiveness), provide a number of community reports, and have several recommendations tools on the way.
Sentiment analysis...my bread and butter (and the wave of the future)
http://adaptivesemantics.com
Great article, i use http://www.whorules.nl a great sentiment analysis tool and free. Introduced by the company Teezir in The Netherlands
The best gauge of sentiment is to walk around and look at people's long faces. Economy terrible, government scrambling, corporations hurting for credit and sales, citizens trying to stay employed of looking for decent paying jobs. Can you gauge my sentiment? Do you need an algorithm, a Gallop poll?
http://www.NumisEX.com
It has been a great week for topics related to sentiment analysis. Nice to read your thoughts, as well.
Naturally, the most talked about topic lately is how to determine sentiment from content shared on social networks and from other online sources. There have been great advances in that area from all the sentiment vendors.
We've seen lots of movement in sentiment analysis used in trading systems and other financial services products. Customer satisfaction is always a great example of how to successfully integrate the automated process with human analysis.
Looking forward to exploring new ways to apply the technology.
Cheers,
Christine
Lexalytics, Inc
Thanks for your coverage of these sentiment analysis tools. As the comments suggest there are many consumer and enterprise search applications which can utilize these complex relationships between topics, experiences, opinions and the people behind them.
One service, TipTop delivers sentiment-based search results from content published on Twitter. For example, do a search for 'iphone':
http://feeltiptop.com/iphone/
>Individual sentiment is presented in the form of a Tip or Pit-type messages in the search results columns. Messages with neutral sentiment are found by clicking the Related Messages results button.
>Aggregate sentiment about the search topic is presented as a percentage of Tips and Pits in the search results column title bar.
>Related concepts around the search topic are found below "Filter results by".
Sentimentally speaking, whatever topic you are looking for, TipTop helps you find the best advice, opinions, and answers as well as people to engage and share experiences with.
Greg Martin
For companies that want to do social media monitoring and measurement, Heartbeat is a cost-effective service with a lot of flexibility such as the ability to do unlimited queries.
Punsons Flora | Send Flowers to Agra | Send Flowers to Ahmedabad | Send Flowers to Amritsar | Send Flowers to Bangalore | Send Flowers to Baroda
Social Media Monitoring tools vendors are all full of shite.
It reminds me of Google Alerts where half of what you get is irrelevant or very old. Unless you invest countless amounts of time and money, you'll never have clean data, and most of the demo's given do not match the reality of pulling data for real companies.
DIY can not work in the pr and marketing world when all users want are automated answers, and where clients can't afford to pay the costs for people to search for needles in the haystack.
Think about any data element that gets pulled - it was most likely tagged by the original user, and these tags drive many of the hits people receive. Users tagging don't care that you need a taxonomy to see what they provide, they're using their own taxonomy. This is why spider index matching systems aren't working right.
Think about any data element that gets pulled - it was most likely tagged by the original user, and these tags drive many of the hits people receive. Users tagging don't care that
It reminds me of Google Alerts where half of what you get is irrelevant or very old. Unless you invest countless amounts of time and money, you'll never have clean data, and most of the demo's given do not match the reality of pulling data for real companies.
Heartbeat is a cost-effective service with a lot of flexibility such as the ability to do unlimited queries.
Very interesting article on how sentiment can be calculated. I am still thinking on how best these systems work and how easily can they may be fooled. Do they just work based on some word filters that define the mood of positive and negative