stefan-it commited on
Commit
890c6b7
1 Parent(s): bec3c8d

readme: add initial version

Browse files
Files changed (1) hide show
  1. README.md +44 -0
README.md ADDED
@@ -0,0 +1,44 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - flair
4
+ - token-classification
5
+ - sequence-tagger-model
6
+ language: de
7
+ widget:
8
+ - text: "Herr Oberst Brunner ist nämlich Hauptagent für den Kanton Zürich."
9
+ license: mit
10
+ ---
11
+
12
+ # Triple E - Effective Ensembling of Embeddings and Language Models for NER of Historical German
13
+
14
+ Based on [our paper](http://ceur-ws.org/Vol-2696/paper_173.pdf) we release a new baseline model for the German
15
+ [CLEF-HIPE shared task](https://impresso.github.io/CLEF-HIPE-2020/).
16
+
17
+ In contrast to the models used in the paper, we manually sentence-segmented and normalize hyphenations and
18
+ trained a NER model using the German Europeana BERT model.
19
+
20
+ Additionally, we perform experiments with different context sizes. This approach is described in
21
+ more detail in [this paper](https://arxiv.org/abs/2011.06993).
22
+
23
+ # Results
24
+
25
+ The results with different context sizes can be seen in the following table:
26
+ | Model | Run 1 | Run 2 | Run 3 | Run 4 | Run 5 | Avg.
27
+ | -------------------------- | --------------- | --------------- | --------------- | ------------------- | --------------- | ---------------
28
+ | German Europeana BERT | (81.45) / 76.92 | (**81.53**) / 77.03 | (80.49) / 77.83 | (80.88) / 77.19 | (81.39) / 77.00 | (81.15 ± 0.45) / 77.19 ± 0.34
29
+ | German Europeana BERT (16) | (**82.56**) / 77.38 | (81.19) / 77.76 | (80.99) / 76.34 | (81.27) / 77.70 | (81.28) / 77.22 | (81.46 ± 0.63) / 77.28 ± 0.57
30
+ | German Europeana BERT (32) | (**82.04**) / 78.50 | (81.14) / 76.56 | (81.81) / 78.28 | (81.50) / 76.90 | (81.64) / 77.94 | (81.63 ± 0.34) / 77.64 ± 0.86
31
+ | German Europeana BERT (64) | (81.21) / 78.39 | (81.27) / 75.98 | (**81.88**) / 78.40 | (81.66) / 77.35 | (81.29) / 76.70 | (81.46 ± 0.29) / 77.36 ± 1.06
32
+ | German Europeana BERT (80) | (82.13) / 77.77 | (81.31) / 76.81 | (82.09) / 78.69 | (**82.30**) / 76.79 | (80.65) / 77.10 | (81.70 ± 0.70) / 77.43 ± 0.81
33
+
34
+ ## Comparisons
35
+
36
+ The following figure shows the results with different context sized (on development dataset):
37
+
38
+ <link>
39
+
40
+ We perform "Almost Stochastic Order" tests as proposed in the
41
+ ["Deep Dominance - How to Properly Compare Deep Neural Models"](https://www.aclweb.org/anthology/P19-1266/) paper.
42
+ The heatmap figure is heavily inspired by the ["CharacterBERT"](https://arxiv.org/abs/2010.10392) paper.
43
+
44
+ <link>