Text Classification of Conversational Implicatures Based on Lexical Features

Figures & data

Table 1. Features and their coefficients in the Optimal Feature Set F₁.

Feature	The Explanation of Feature	Coefficient
WRDFRQmc	CELEX Log minimum frequency for content words, mean	0.23078
WRDHYPv	Hypernymy for verbs, mean	0.158253
WRDFRQc	CELEX word frequency for content words, mean	0.1365
DESWLltd	Word length, number of letters, standard deviation	−0.11829
DESWLlt	Word length, number of letters, mean	0.097787
WRDHYPnv	Hypernymy for nouns and verbs, mean	−0.06214
DESWLsyd	Word length, number of syllables, standard deviation	−0.04945
WRDPOLc	Polysemy for content words, mean	0.047379
WRDHYPn	Hypernymy for nouns, mean	−0.03848
DESWLsy	Word length, number of syllables, mean	−0.03355
WRDFRQa	CELEX Log frequency for all words, mean	0.025816
LDTTRc	Lexical diversity, type-token ratio, content word lemmas	0.009295
LDMTLD	Lexical diversity, MTLD, all words	−0.0049
LDTTRa	Lexical diversity, type-token ratio, all words	−0.00348
WRDFAMc	Familiarity for content words, mean	−0.0023
WRDMEAc	Meaningfulness, Colorado norms, content words, mean	0.002255
WRDIMGc	Imagability for content words, mean	−0.00211

Table 2. Results of the performance measures, T-test and Goodness-of-fit test for the text classification of conversational implicatures with context utterance.

Folds	Performance measures					T-test	Goodness of fit test
	TP	sensitivity	TN	specificity	accuracy	Sig.	${\chi }^2$	Sig.
fold 1	50	0.667	46	0.613	0.640	0.500	11.760	0.001
fold 2	45	0.600	42	0.560	0.580	0.622	3.840	0.050
fold 3	42	0.560	48	0.640	0.600	0.321	6.000	0.014
fold 4	41	0.547	40	0.533	0.540	0.871	0.960	0.327
total	178	0.593	176	0.587	0.590	0.868	19.440	0.000

Table 3. Features and their coefficients in the Optimal Feature Set F₂.

Feature	The Explanation of Feature	Coefficient
WRDFRQmc	CELEX Log minimum frequency for content words, mean	0.290257
DESWLltd	Word length, number of letters, standard deviation	−0.255635
DESWLlt	Word length, number of letters, mean	0.182417
WRDFRQc	CELEX word frequency for content words, mean	0.136391
DESWLsyd	Word length, number of syllables, standard deviation	−0.108100

Table 4. Results of the performance measures, T-test and Goodness-of-fit test for the text classification of conversational implicatures without context utterance.

Folds	Performance measures					T-test	Goodness of fit test
	TP	sensitivity	TN	specificity	accuracy	Sig.	${\chi }^2$	Sig.
fold 1	42	0.560	44	0.587	0.573	0.743	3.227	0.072
fold 2	46	0.613	53	0.707	0.660	0.230	15.360	0.000
fold 3	36	0.480	51	0.680	0.580	0.013	3.840	0.050
fold 4	49	0.653	43	0.573	0.613	0.318	7.707	0.006
total	173	0.577	191	0.637	0.607	0.133	27.307	0.000

Table 5. Comparison between with and without context utterance.

Folds	Performance measures		Goodness of fit test		Effect comparison	Matched T-test
	accuracy		${\chi }^2$
	with C.	without C.	with C.	without C.		Sig.
fold 1	0.640	0.573	11.760	3.227	1 > 2	0.105
fold 2	0.580	0.660	3.840	15.360	1 < 2	0.064
fold 3	0.600	0.580	6.000	3.840	1 > 2	0.687
fold 4	0.540	0.613	0.960	7.707	1 < 2	0.153
total	0.590	0.607	19.440	27.307	1 < 2	0.473

Table 6. Results of frequency statistics and contingency analysis for context and implicature.

Folds	1			-1			Contingency analysis
	yes	no	total	yes	no	total	value	sig.
fold 1	14	10	24	6	8	14	0.849	0.357
fold 2	11	4	15	12	15	27	3.249	0.071
fold 3	17	12	29	11	15	26	1.460	0.227
fold 4	12	12	24	20	15	35	0.293	0.589
total	54	38	92	49	53	102	2.206	0.138

Data Availability Statement

The authors confirm that the data supporting the findings of this study are available within the article and its supplementary materials. The data named “LexicalData.rar” is also available at https://github.com/Albert-Dillon/CompImplicature.