Overall induction design plan

PedroArvela / sentido

SENse Through InDuctiOn

1 stars 0 forks source link

Closed PedroArvela closed 6 years ago

PedroArvela commented 6 years ago

In gitlab by @PedroArvela on May 23, 2017, 23:49

For word w1:

[x] Find all (wx, wy) co-occurrences in all contexts of w1.
- [x] Apply restrictions here.
[x] Run graph clustering algorithm on (wx, wy) pairs.
[x] Reconstruct words from their IDs

PedroArvela commented 6 years ago

In gitlab by @PedroArvela on May 30, 2017, 23:46

Depends on #8.

PedroArvela commented 6 years ago

In gitlab by @PedroArvela on Jul 9, 2017, 17:39

Induction of a single word on the complete database takes about 30 minutes to run.

PedroArvela commented 6 years ago

In gitlab by @PedroArvela on Jul 9, 2017, 17:39

It is advisable to probably create a new storage place for the induced clusters.

PedroArvela commented 6 years ago

In gitlab by @PedroArvela on Aug 21, 2017, 18:24

Only thing missing is to reconstruct the words from their IDs, to export to the CSV and use on the disambiguator.