Hacker News new | comments | ask | show | jobs | submit login
Show HN: A new annotation tool for information extraction with Mechanical Turk (github.com)
72 points by varal7 30 days ago | hide | past | web | favorite | 10 comments



This looks really really nice, do you have any plan to add relational annotation support as well? Asking since a student is about to embark on a project that would require it. In hindsight, I regret that we did not get around to implementing Mechanical Turk support back when we made brat. [1]

[1]: http://brat.nlplab.org/


Thanks for brat, its quite useful and is also the default annotation visualizer for CoreNLP.

Edit: can’t find github but the JavaScript made its way to annodoc project too. Have tried using both and I am sure there’s lots of use cases for brat.


Thanks for the BRAT software - I used it for my PhD and it was very helpful. From the looks of things things though BRAT is now abandonware? Despite that, I actually used it this semester to have students do annotations and it worked out quite well.

I also agree the tool looks nice but relation support is definitely something that I also would need to see supported.


Thank you! I had no plans to add this feature in the immediate future (I also need to work on the project I built this for). I might consider doing it if people want it, in which case I will help brat for inspiration.


I will use* brat for inspiration


Can also you Dataturks do a similar thing, hooking up with Mturks/your own team..also allows overlapping selection and NER+Classification etc. Adding extra data and much more.

Here is an online demo:

https://dataturks.com/projects/Dataturks/Demo%20Document%20A...


This is a nice product! I appreciate that you made a free version: very valuable for students


Could someone explain to me what this is used for ?

I know what Mechanical Turk is. I know what "annotation" is (and I'm proud of it) but beyond that I don't understand what this is useful for.

Thanks.


Nice. One confusing thing about the demo is: isn't SMTP the clear answer for a protocol mentioned in the description? And does it accommodate multiple answers for the same category?


Good catch. Yes, it should be SMTP. Yes, you can extract multiple spans from the text.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: