Stuart Towns Automated text analysis
- Login to post comments
Wed, 06/18/2014 - 14:03
Mon, 03/05/2018 - 20:11
#1
Customer exchanges on the
Customer exchanges on the web, item audits, and advanced chronicles of news articles and official statements are only a couple of potential hotspots for experiences about purchaser demeanors, communication, and culture. it is a helpful Help With Assignment Writing device for inspecting designs in content that neither analysts nor customers can recognize unaided.
Fri, 01/05/2018 - 15:10
#2
online
You join the most fun to use the latest app here the youtube to mp3 converter it is the amazing app for the music lovere you can use this to convert youtube videos to mp4 and download the most songs forever thanks for the visit here.
Fri, 01/05/2018 - 13:49
#3
sah
have
Mon, 06/30/2014 - 23:44
#4
Hi Stuart,
Thank you for the
Hi Stuart,
Thank you for the interesting issues you have made about using automated tools and your examples make me aware of how much I can trust the tools I use.
In the final stage in your example, you have got 108 values. By the limitations of analysis using Coh-Metrix you have presented, how much you are confident about those numbers? (and in case that your confidence is not very high, are there any additional steps that you would take before interpreting the numbers?) Thank you.
Wed, 07/16/2014 - 10:05
#5
Coh-Metrix Confidence
Hello Punjaporn. Thank you for a very interesting question! I think that determining confidence levels for automated tools is a great topic for future research. With many automated tools, such as concordances, frequency lists, range, etc. the confidence is very high if not 100%. The researcher can have full faith that the tool is giving them the right answer. But Coh-Metrix is a little bit different because they are trying to push the technology to do things that it can't quite do at 100% accuracy.
This month I am working on a conference paper that takes a look at one small part of this equation with Reference Cohesion in Coh-Metrix. Preliminary findings show that the results are roughly 80-90% accurate on what it is advertised to do.
Someone at my talk asked something like, "but isn't 'close enough' ok?" Yes, I think 80-90% is probably pretty good, but I think that it would be very helpful if the research includes a small section on limitations or at least is careful in the way that the results are presented.
Mon, 06/23/2014 - 13:27
#6
I look forward to hearing from you all!
Hello everyone.
First of all, thank you to KMUTT for a great conference and for setting up this post-conference discussion forum. I hope that people will use it and that we can learn a lot from each other.
The Powerpoint from my DRAL presentation is attached above. The title was "Blueprints or Conduits: Using an automated tool for text analysis". I would be very happy to hear any comments, questions, or thoughts from any of you. I'll be sure to check in every now and then to see if anyone has anything to say.
Thanks!
Stuart Towns