Mordecai Geoparser User Survey

134 responses

How do you use Mordecai?

120 out of 134 answered

Academic research or education
69 resp.
57.5%
Personal/hobby
38 resp.
31.7%
Industry
30 resp.
25%
Government
4 resp.
3.3%
Other
0 resp.
0%

How would you rate Mordecai's accuracy?

121 out of 134 answered

6.6
Average rating

How would you rate its speed?

122 out of 134 answered

6.1
Average rating

How would you rate its documentation?

120 out of 134 answered

5.8
Average rating

How would you rate its reliability/stability?

118 out of 134 answered

6.6
Average rating

What are the main obstacles you face in using Mordecai?

115 out of 134 answered

it's difficult to install
41 resp.
35.7%
the results aren't accurate
41 resp.
35.7%
it doesn't support non-English text
40 resp.
34.8%
lack of documentation or tutorials
39 resp.
33.9%
it's slow
32 resp.
27.8%
the code is buggy
10 resp.
8.7%
Other
2 resp.
1.7%
the results could be more accurate (more more more)
It requires a lot of memory with the default geonames index.

Which new features would you like to see?

112 out of 134 answered

training scripts to update the models with new data
60 resp.
53.6%
non-English support
52 resp.
46.4%
a model for geolocating events in text
47 resp.
42%
a whitepaper describing the models
34 resp.
30.4%
a hosted service
26 resp.
23.2%
Other
8 resp.
7.1%
simple override system where common mistakes can be whitelisted/blacklisted
A REST Api
longer text input (right now I get an error message if it's more than a short paragraph)
The installation and initial-use is quite bumpy.
step-by-step tutorials including setup and troubleshooting
Argument for ElasticSearch timeout variable. Default of 10 is too low.
ability to run on M1 mac using updated elasticsearch
n-gram tokenization seems to be limited? Also, in my case, a lot of obvious places (e.g. cities such as Paris or London) aren't always identified

Thanks for participating!
Final question:

Given a tradeoff between speed and accuracy, which would you prefer?

114 out of 134 answered

Accuracy: I want the most accurate results possible and I'd pay a cost in speed
92 resp.
80.7%
Speed: a slower model would be less useful, even if more accurate.
22 resp.
19.3%
Powered by Typeform