moreo
/
dante-verification
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Merge branch 'master' of https://github.com/AlexMoreo/dante-verification

This commit is contained in:
Alejandro Moreo Fernandez 2020-04-03 11:21:34 +02:00
parent 843cfbe8fe 6471250118
commit dc810272b2
1 changed files with 35 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

38
README.md
View File

@ -5,6 +5,15 @@ Code to reproduce the experiments reported in the papers
and
["L’Epistola a Cangrande al vaglio della Computational Authorship Verification: Risultati preliminari (con una postilla sulla cosiddetta XIV Epistola di Dante Alighieri)"](https://www.academia.edu/42297516/L_Epistola_a_Cangrande_al_vaglio_della_Computational_Authorship_Verification_risultati_preliminari_con_una_postilla_sulla_cosiddetta_XIV_Epistola_di_Dante_Alighieri_in_Nuove_inchieste_sull_Epistola_a_Cangrande_a_c._di_A._Casadei_Pisa_Pisa_University_Press_pp._153-192)
## Requirements:
The experiments have been run using the following packages (older versions might work as well):
* joblib==0.11
* nltk==3.4.5
* numpy==1.18.2
* scikit-learn==0.22.2.post1
* scipy==1.4.1
## Disclaimer:
The dataset is not distributed in this version. We have asked the Editors for permission to publish the corpus.
We are waiting for some of these responses to arrive.
@ -13,18 +22,23 @@ We are waiting for some of these responses to arrive.
The script in __./src/author_identification.py__ executes the experiments. This is the script syntax (--help):
```
usage: author_identification.py [-h] [--loo] [--unknown PATH] [--log PATH]
CORPUSPATH AUTHOR
Authorship verification for Epistola XIII
positional arguments:
PATH Path to the directory containing the corpus (documents
must be named <author>_<texname>.txt)
positive Positive author for the hypothesis (default "Dante"); set
to "ALL" to check every author
CORPUSPATH Path to the directory containing the corpus (documents must
be named <author>_<texname>.txt)
AUTHOR Positive author for the hypothesis (default "Dante"); set to
"ALL" to check every author
optional arguments:
-h, --help show this help message and exit
--loo submit each binary classifier to leave-one-out validation
--unknown PATH path to the file of unknown paternity (default None)
--log PATH path to the log file where to write the results (default
./results.txt)
```
The following command line:
@ -42,6 +56,18 @@ to the positive class.
Similarly, the command line:
```
cd src
python author_identification.py ../Corpora/CorpusI Dante --loo
python author_identification.py ../Corpora/CorpusI ALL --loo
```
will perform a cross-validation of the binary classifier for Dante using all training documents in a leave-one-out (LOO) fashion.
will perform a cross-validation of the binary classifier for all authors using all training documents in a leave-one-out (LOO) fashion.
The script will report the results both in the standard output (more elaborated) and in a log file. For example, the last command will produce a log file containing:
```
F1 for ClaraAssisiensis = 0.400
F1 for Dante = 0.957
F1 for GiovanniBoccaccio = 1.000
F1 for GuidoFaba = 0.974
F1 for PierDellaVigna = 0.993
LOO Macro-F1 = 0.865
LOO Micro-F1 = 0.981
```
(Note that small numerical variations with respect to the original papers might occur due to different software versions and as a result from any stochastic underlying process. Those changes should anyway not alter the conclusions derived from the published results.)