iSentenizer-𝝁: Multilingual Sentence Boundary Detection Model

Table 6

Performance of systems on different languages of Europarl corpus.

Corpus	Candidates	Recall	Precision	𝐹-Score
Danish	iSentenizer	98.84%	92.88%	95.77%
	Punkt	97.69%	79.37%	87.59%
	MxTerminator	35.48%	94.13%	51.54%
German	iSentenizer	97.61%	95.77%	97.61%
	Punkt	97.87%	87.53%	92.41%
	MxTerminator	81.00%	93.69%	86.89%
English	iSentenizer	98.98%	95.79%	97.36%
	Punkt	97.95%	93.34%	95.59%
	MxTerminator	96.09%	93.97%	95.02%
Spanish	iSentenizer	99.40%	94.21%	96.74%
	Punkt	98.11%	89.80%	93.77%
	MxTerminator	96.67%	90.09%	93.26%
Dutch	iSentenizer	99.34%	96.24%	97.77%
	Punkt	97.79%	92.34%	94.99%
	MxTerminator	91.95%	95.32%	93.61%
French	iSentenizer	98.82%	95.77%	97.28%
	Punkt	97.84%	91.37%	94.49%
	MxTerminator	95.04%	91.88%	93.44%
Italian	iSentenizer	98.90%	95.99%	97.42%
	Punkt	98.25%	93.69%	95.92%
	MxTerminator	94.96%	94.43%	94.70%
Portuguese	iSentenizer	99.58%	96.60%	98.07%
	Punkt	98.50%	95.76%	97.11%
	MxTerminator	94.88%	96.50%	95.68%
Greek	iSentenizer	97.83%	96.44%	97.13%
	Punkt	96.98%	95.36%	96.16%
	MxTerminator	97.24%	93.97%	95.58%
Finnish	iSentenizer	98.98%	95.76%	97.34%
	Punkt	98.33%	92.34%	95.24%
	MxTerminator	92.46%	95.32%	93.87%
Swedish	iSentenizer	95.91%	94.30%	95.10%
	Punkt	98.94%	89.45%	93.95%
	MxTerminator	99.49%	88.33%	93.57%