Identification of Code-Switched Sentences and Words Using Language Modeling Approaches

Table 3

Results of code-switched word identification in Mandarin-Taiwanese code-switched sentences.

Methods		Recall	Precision	𝐹-measure
Random	Top 1	17.65%	18.32%	17.98%
	Top 2	31.62%	16.41%	21.61%
	Top 3	49.26%	17.05%	25.33%
Word bigram	Top 1	40.46%	41.73%	41.09%
	Top 2	60.31%	31.85%	41.69%
	Top 3	74.81%	27.53%	40.25%
Word trigram	Top 1	14.50%	15.83%	15.14%
	Top 2	35.88%	20.35%	25.97%
	Top 3	55.73%	22.53%	32.09%
POS bigram	Top 1	42.75%	43.08%	42.91%
	Top 2	67.94%	35.18%	46.35%
	Top 3	82.44%	29.35%	43.29%
POS trigram	Top 1	52.67%	53.49%	53.08%
	Top 2	73.28%	39.34%	51.20%
	Top 3	83.97%	31.98%	46.32%