Adaptions to Database - Githubissues

DeNederlandscheBank / nqm

A Transformer-based Machine for answering questions on insurance companies

MIT License

0 stars 0 forks source link

Adaptions to Database #7

Closed jm-glowienke closed 3 years ago

jm-glowienke commented 3 years ago

There are some things, which might be handy to change in the underlying database.

[x] Check if it is possible to add predicate "legalNameContains" to prevent that precise name has to be used to find subject
[x] assimilate names, e.g. only use small letters

jm-glowienke commented 3 years ago

If the above is not possible, a workaround for the names could be "Fuzzy Name Matching" (compare Open Source Lunch Presentation by Tim Haarmann, 15-Oct-2020). However, first we have to see, whether the name matching will actually be a problem for the translator to learn.

jm-glowienke commented 3 years ago

Use excel file with alternative names is likely the best option!

This should be the approach to be taken.

jm-glowienke commented 3 years ago

predicate "eiopaBase:isIdentifiyingName" has been added where ID and names in small letters are used.

For every, subject several options are given and only one has to be matched