Table 5.

Expert-generated marijuana terms.

Rank	Term	Cosine similarity	Drug relevancy
			Annotator 1	Annotator 2	Average relevancy, %
1	ediblesa	0.58	180/200	185/200	91.3
2	weeda	0.71	186/200	177/200	90.8
3	ganjaa	1.00	190/200	169/200	89.8
4	kusha	0.77	192/200	165/200	89.3
5	sativaa	0.68	176/200	177/200	88.3
6	sour diesela	0.68	91/112	98/112	84.4
7	indicaa	0.54	132/200	170/200	75.5
8	devil’s lettucea	0.67	145/200	150/200	73.8
9	droa	0.72	136/200	134/200	67.5
10	dabsa	0.54	123/200	142/200	66.3
11	purpa	0.67	75/200	59/200	33.5
12	pota	0.49	64/200	69/200	33.3
13	danka	0.54	56/200	43/200	24.8
14	herba	0.48	37/200	54/200	22.8
15	reggiea	0.47	41/200	46/200	21.8
16	waxb	0.45	42/200	43/200	21.3
17	nugb	0.36	16/200	55/200	17.8
18	mary janeb	0.31	24/200	36/200	15.0
19	pineapple expressb	0.38	29/200	27/200	14.0
20	chronicb	0.26	35/200	18/200	13.3
21	shatterb	0.16	18/200	32/200	12.5
22	budb	0.36	20/200	21/200	10.3
23	skunkb	0.31	24/200	14/200	9.5
24	hazeb	0.36	13/200	17/200	7.5
25	exoticsb	0.19	9/143	12/143	7.3
26	hydrob	0.32	15/200	12/200	6.8
27	shardb	0.14	5/138	4/138	3.3
28	flowerb	0.29	1/200	9/200	2.5
29	fireb	0.30	0/200	10/200	2.5
30	mudb	0.31	1/200	4/200	1.3
31	flameb	0.40	0/200	5/200	1.3
32	greenb	0.25	1/200	3/200	1.0

^aTerms provided by author-affiliated experts included in the model-derived list of 200 candidate terms.

^bExpert-provided terms that were not included in the candidate term list returned by our model queries.