After the amazing new Rules feature made available in our last product update, the list of additions to the Brandwatch app can seem a little slim this time round. This is more a testament to our commitment to a fortnightly update cycle and the benefits that it affords than it is to any stinginess in our offerings.
By maintaining a highly regular update schedule we are able to implement fixes and urgently required features at no longer than a few days after they are first highlighted.
By the same token, we deliver a consistently robust product that is always balanced on the cutting edge of development. Due to the browser-based nature of the Brandwatch app, this also means that our clients aren’t bombarded with compulsory downloads or installs every fortnight.
This round of updates is extensive, but mostly addresses known minor issues within the app and focuses on small tweaks rather than major feature implementations.
One of the aspects of Brandwatch that we have worked on to improve is the range of language classifiers used to determine the sentiment of mentions across the internet.
We use a number of techniques to deliver our sentiment analysis, yet so far for many languages we have been using machine-learning to determine sentiment. This is where the classifiers learn different rules as the data comes in over time, and continually adapt accordingly.
This week we have created a bespoke rules-based classifier for Norwegian, which will complement the machine-learning approach. This classifier has pre-stated rules to adhere to, helping it to establish sentiment more accurately.
As well as Norwegian, we already have multiple rules-based classifiers for languages such as English and German, and hope to add even more in the coming weeks.
This weekend we have also created more stopwords for Japanese, a language that is still in the beta stage of development for use within Brandwatch.
Another element we are slowly trying to improve is our collection of Google Plus material, and we have taken measures to boost our Google Plus coverage with this update.
Perhaps some of the most significant continuous updates we implement are the improvements to forum data collection. We have improved our analysis of forums that use IPBoard and for the popular automobile forum, Pistonheads.
We have accurate date-specification for these forums, and now Brandwatch is able to see each forum post as an individual mention, rather than simply a solitary page with numerous instances of what would have been essentially the same mention.
This now enables users to use Brandwatch exclusively for one of these types of forum. For example, we could set up a query that collects data from Pistonheads by using either the ‘site’ or ‘URL’ operator. Our query would look something like ‘site:www.pistonheads.com’.
We are now able to search just that site itself, and can perform an analysis of the data within it. Using the categories function, we can determine what users of the Pistonheads forum are talking about.
In this case we have isolated the four main German car manufacturers to see which one gets discussed most on the forum.
You could even take this a step further and isolate a single thread, so that data could be pulled from responses to a single question, for example “what is your favourite car from 2011?” and then analyse the data from there.
We never stop improving Brandwatch and are always looking at ways to better the product. We listen, discuss, note, test and create new ways to make Brandwatch more accessible and more powerful, and we aim to bring you an improved iteration every fortnight.
These new feaures are nothing short of awesome. Now we can reliably produce domain-level reports for clients considering establishing relationships (PR, paid media, etc) with some of the most robust social media sites out there: forums.
Product Update: Now With Added Norwegian
December 12th, 2011. Posted by Joel WindelsAfter the amazing new Rules feature made available in our last product update, the list of additions to the Brandwatch app can seem a little slim this time round. This is more a testament to our commitment to a fortnightly update cycle and the benefits that it affords than it is to any stinginess in our offerings.
By maintaining a highly regular update schedule we are able to implement fixes and urgently required features at no longer than a few days after they are first highlighted.
By the same token, we deliver a consistently robust product that is always balanced on the cutting edge of development. Due to the browser-based nature of the Brandwatch app, this also means that our clients aren’t bombarded with compulsory downloads or installs every fortnight.
This round of updates is extensive, but mostly addresses known minor issues within the app and focuses on small tweaks rather than major feature implementations.
One of the aspects of Brandwatch that we have worked on to improve is the range of language classifiers used to determine the sentiment of mentions across the internet.
We use a number of techniques to deliver our sentiment analysis, yet so far for many languages we have been using machine-learning to determine sentiment. This is where the classifiers learn different rules as the data comes in over time, and continually adapt accordingly.
This week we have created a bespoke rules-based classifier for Norwegian, which will complement the machine-learning approach. This classifier has pre-stated rules to adhere to, helping it to establish sentiment more accurately.
As well as Norwegian, we already have multiple rules-based classifiers for languages such as English and German, and hope to add even more in the coming weeks.
This weekend we have also created more stopwords for Japanese, a language that is still in the beta stage of development for use within Brandwatch.
Another element we are slowly trying to improve is our collection of Google Plus material, and we have taken measures to boost our Google Plus coverage with this update.
Perhaps some of the most significant continuous updates we implement are the improvements to forum data collection. We have improved our analysis of forums that use IPBoard and for the popular automobile forum, Pistonheads.
We have accurate date-specification for these forums, and now Brandwatch is able to see each forum post as an individual mention, rather than simply a solitary page with numerous instances of what would have been essentially the same mention.
This now enables users to use Brandwatch exclusively for one of these types of forum. For example, we could set up a query that collects data from Pistonheads by using either the ‘site’ or ‘URL’ operator. Our query would look something like ‘site:www.pistonheads.com’.
We are now able to search just that site itself, and can perform an analysis of the data within it. Using the categories function, we can determine what users of the Pistonheads forum are talking about.
In this case we have isolated the four main German car manufacturers to see which one gets discussed most on the forum.
You could even take this a step further and isolate a single thread, so that data could be pulled from responses to a single question, for example “what is your favourite car from 2011?” and then analyse the data from there.
We never stop improving Brandwatch and are always looking at ways to better the product. We listen, discuss, note, test and create new ways to make Brandwatch more accessible and more powerful, and we aim to bring you an improved iteration every fortnight.