A Grammar-Based Semantic Similarity Algorithm for Natural Language Sentences

Table 6

Results of the grammar-based and competitive methods on the Microsoft Research Paraphrase Corpus.

Category	Metric	Accuracy	Precision	Recall	𝐹-measure
Corpus-based	PMI-IR	69.90	70.20	95.20	81.00
	LSA	68.40	69.70	95.20	80.50
	STS Meth.	72.64	74.65	89.13	81.25
	SyMSS_JCN	70.87	74.70	84.17	79.02
	SyMSS_Vector	70.82	74.15	90.32	81.44
	Omiotis	69.97	70.78	93.40	80.52
Lexicon-based	JC	69.30	72.20	87.10	79.00
	LC	69.50	72.40	87.00	79.00
	Lesk	69.30	72.40	86.60	78.90
	L	69.30	71.60	88.70	79.20
	W&P	69.00	70.20	92.10	80.00
	R	69.00	69.00	96.40	80.40
	M	70.30	69.60	97.70	81.30
Machine learning-based	Wan et al. [58]	75.00	77.00	90.00	83.00
	Z&P	71.90	74.30	88.20	80.70
	Qiu et al. [59]	72.00	72.50	93.40	81.60
Baselines	Random	51.30	68.30	50.00	57.80
	VSM	65.40	71.60	79.50	75.30
	LG	71.02	73.90	91.07	81.59