Below is a recording of a presentation made to AMEC’s individual consultant’s group on the topic of Google Chrome extensions and which ones work best for media evaluation. This features Grammarly, Multi-highlight and Color-Zilla.
While browsing these extensions I noticed a couple that purported to evaluate the sentiment of a clip. This facility would provide a useful proxy of favorability within the bounds that these auto-sentiment tools can be a bit hit and miss. That said, I could not get any of these sentiment extensions to work which was a shame.
It is, however, possible to create an extension, but involves a degree of coding skills (Python mainly) which may take some work. Google’s own notes make this actually seem fairly straightforward though the challenge is pairing it with a sentiment engine. AWS has a version though users have to purchase ‘units’ to trial while Googles offering (Google Cloud Natural Language) allows you to use a number of hundred ‘units’ using the API before it charges.
This is very much a work-in-progress and I have not got much further than playing with the query screen ‘Try the API’ using a number of news clip texts. It has actually proved to be really interesting as it splits out the individual entities, be it people, organisations, etc. It then reports a Salience score on how central to the story the entity is and a Sentiment score for the particular entity. This final point is interesting as with all the auto-sentiment checkers I have used over the years, this is the first one that picks up on an individual sentiment score for an individual entity, rather than just providing an overall sentiment for the entire clip.
Leave a Reply