figur 0.0.2

Figurenerkennung for German literary texts.

Figurenerkennung for German literary texts

An important step in the quantitative analysis of narrative texts is the automatic recognition of references to figures, a special case of the generic NLP problem of Named Entity Recognition (NER).

Usually NER models are not designed for literary texts resulting in poor recall. This easy-to-use package is the continuation of the work of Jannidis et al., using state-of-the-art techniques from the field of Deep Learning reaching a Micro F1-Score of 95.89 and a Macro F1-Score of 67.74.

Figurenerkennung statistics

This is based on the test set of 55,868 tokens.

	TP	FP	FN	TN
AppA	2	2	288	2
AppTdfW	418	131	383	418
Core	626	114	211	626
pron	2284	604	766	2284
_	52538	1546	749	52538

Installation

$ pip install figur

Example

>>> import figur
>>> text = "Der Gärtner entfernte sich eilig, und Eduard folgte bald."
>>> figur.tag(text)
    SentenceId      Token      Tag
0            0        Der        _
1            0    Gärtner  AppTdfW
2            0  entfernte        _
3            0       sich     pron
4            0      eilig        _
5            0          ,        _
6            0        und        _
7            0     Eduard     Core
8            0     folgte        _
9            0       bald        _
10           0          .        _