Version 1 (modified by 4 years ago) (diff) | ,
---|
Stylometry
IA161 Advanced NLP Course?, Course Guarantee: Aleš Horák
Prepared by: Honza Rygl
State of the Art
The analysis of author's characteristic writing style and vocabulary has been used to uncover author's traits such as authorship, age, or gender documents by both manual linguistic approaches and automatic algorithmic methods.
The most common approach to stylometry problems is to combine stylistic analysis with machine learning techniques:
- specific style markers are extracted,
- a classification procedure is applied to extracted markers
References
- Stamatatos, E. (2009), A Survey of Modern Authorship Attribution Methods (2009), Journal of the American Society for Information Science and Technology, 60(3), 538-556. pdf
- Kestemont, M. (2014), Function Words in Authorship Attribution From Black Magic to Theory? Proceedings of the 3rd Workshop on Computational Linguistics for Literature, EACL 2014, 59–66 pdf
- Walter, D. Explanation in Computational Stylometry
Practical Session
Student will get to know a *Style & Identity Recognition* tool. They will test this tool on prepared data. Their goal will be to implement a small function to extract style markers from a text.
- go to
asteria04.fi.muni.cz
server:ssh asteria04.fi.muni.cz
- Download a python package with the assignment
wget https://nlp.fi.muni.cz/trac/research/chrome/site/bigdata/stylometry-assignment.zip
- Unzip the downloaded file
unzip stylometry-assignment.zip
- Go to the unziped folder
cd sir-assignment
- Test the prepared program that analyses data from on-line dating services to distinguish gender (masculine/feminine) by text style features
./run.sh
run.sh
can have two optional parameters:
./run.sh [number_of_testing_cycles] [show_first_N_erroneously_predicted_documents]
The default values, i.e. running ./run.sh
without parameters, are 10
cycles and no documents
(./run.sh 10 0
). With longer feature testing ./run.sh 100
could provide better results (but not necessarily).
Example with document output:
[xrygl@asteria04:~/temp/sir-assignment]$ ./run.sh 10 1 author: on text: Ahoj, (nejen) pro výlety do víru podivnězimního velkoměsta, či divočiny venkova, hledá se partnerka přiměřených rozměrů, tvarů a úrovně. Slečny veselé povahy preferovány; ona je to nejspíš nutnost :-) morphology: ((u'<s>', u'<s>'), (u'Ahoj', u'N.N.I.S.1.-.-.-.-.-.A.-.-.-.-'), (u',', u'Z.:.-.-.-.-.-.-.-.-.-.-.-.-.-'), (u'(', u'Z.:.-.-.-.-.-.-.-.-.-.-.-.-.-'), (u'nejen', u'T.T.-.-.-.-.-.-.-.-.-.-.-.-.-'), (u')', u'Z.:.-.-.-.-.-.-.-.-.-.-.-.-.-'), (u'pro', u'R.R.-.-.4.-.-.-.-.-.-.-.-.-.-'), (u'v\xfdlety', u'N.N.I.P.4.-.-.-.-.-.A.-.-.-.-'), (u'do', u'R.R.-.-.2.-.-.-.-.-.-.-.-.-.-'), (u'v\xedru', u'N.N.I.S.2.-.-.-.-.-.A.-.-.-.-'), (u'podivn\u011bzimn\xedho', u'A.A.N.S.2.-.-.-.-.1.A.-.-.-.-'), (u'velkom\u011bsta', u'N.N.N.S.2.-.-.-.-.-.A.-.-.-.-'), (u',', u'Z.:.-.-.-.-.-.-.-.-.-.-.-.-.-'), (u'\u010di', u'J.^.-.-.-.-.-.-.-.-.-.-.-.-.-'), (u'divo\u010diny', u'N.N.F.S.2.-.-.-.-.-.A.-.-.-.-'), (u'venkova', u'N.N.I.S.2.-.-.-.-.-.A.-.-.-.-'), (u',', u'Z.:.-.-.-.-.-.-.-.-.-.-.-.-.-'), (u'hled\xe1', u'V.B.-.S.-.-.-.3.P.-.A.A.-.-.-'), (u'se', u'P.7.-.X.4.-.-.-.-.-.-.-.-.-.-'), (u'partnerka', u'N.N.F.S.1.-.-.-.-.-.A.-.-.-.-'), (u'p\u0159im\u011b\u0159en\xfdch', u'A.A.I.P.2.-.-.-.-.1.A.-.-.-.-'), (u'rozm\u011br\u016f', u'N.N.I.P.2.-.-.-.-.-.A.-.-.-.-'), (u',', u'Z.:.-.-.-.-.-.-.-.-.-.-.-.-.-'), (u'tvar\u016f', u'N.N.I.P.2.-.-.-.-.-.A.-.-.-.-'), (u'a', u'J.^.-.-.-.-.-.-.-.-.-.-.-.-.-'), (u'\xfarovn\u011b', u'N.N.F.S.2.-.-.-.-.-.A.-.-.-.-'), (u'.', u'Z.:.-.-.-.-.-.-.-.-.-.-.-.-.-'), (u'<s>', u'<s>'), (u'Sle\u010dny', u'N.N.F.P.1.-.-.-.-.-.A.-.-.-.-'), (u'vesel\xe9', u'A.A.N.S.1.-.-.-.-.1.A.-.-.-.-'), (u'povahy', u'N.N.F.S.2.-.-.-.-.-.A.-.-.-.-'), (u'preferov\xe1ny', u'V.s.T.P.-.-.-.X.X.-.A.P.-.-.-'), (u';', u'Z.:.-.-.-.-.-.-.-.-.-.-.-.-.-'), (u'ona', u'P.P.F.S.1.-.-.3.-.-.-.-.-.-.-'), (u'je', u'V.B.-.S.-.-.-.3.P.-.A.A.-.-.-'), (u'to', u'P.D.N.S.1.-.-.-.-.-.-.-.-.-.-'), (u'nejsp\xed\u0161', u'D.g.-.-.-.-.-.-.-.3.A.-.-.-.-'), (u'nutnost', u'N.N.F.S.1.-.-.-.-.-.A.-.-.-.-'), (u':', u'Z.:.-.-.-.-.-.-.-.-.-.-.-.-.-'), (u'-', u'Z.:.-.-.-.-.-.-.-.-.-.-.-.-.-'), (u')', u'Z.:.-.-.-.-.-.-.-.-.-.-.-.-.-')) Acc: 73.7 +- 2.2% (baseline 50.0%, 10 iterations)
Task
Examine files in stylometry_features
folder.
Modify the assignment.py
file to increase the accuracy of methods (use vi
or nano
command for editing).
You can create another classes inside, change their names and class names to improve the accuracy score. Don't forget to add new classes into the assignments
list in this file.
Suggestions for your inspiration include:
- diacritics usage (yes/no), a regular expression will be needed
- sentence endings (number of sentences, or typical endings)
- repetitions of words in sentences/in the text
- usage of uppercase letters
- length of sentences/text
- POS tags (n-grams)
- word n-grams
- character n-grams
Each modification can be tested by running ./run.sh
again.
The first call of run.sh
can be slower, because documents are morphologically analysed during the first run.
Write the resulting Acc:
line into the top comment of the assignment.py
file.
Submit your assignment.py
file.