README.md

CLICS³

The third installment of CLICS - the database of Cross-Linguistic Colexifications.

Cite as

Rzymski, Tresoldi et al. 2019. The Database of Cross-Linguistic Colexifications, reproducible analysis of cross- linguistic polysemies. DOI: doi.org/10.17613/5awv-6w15

This repository contains

• specifications of the source data as human readable table and as requirements file to install with pip
• a map showing the geographic distribution of languages in the CLICS database, encoded in GeoJSON
• instructions to compile the CLICS database (see below) and
• the CLICS data:
  • the (zipped) SQLite database created in step 4 below, zipped running

    zip -9 clics3.sqlite.zip clics.sqlite

  • the full, GML encoded network created in step 5 below, zipped running

    zip -9 clics3-network.gml.zip graphs/network-3-families.gml

• the exact set of pip requirements used to create the artefacts above.

This data is licensed under CC BY 4.0.

Creating the CLICS database

0. Analyses should ideally be carried within virtual environments, so that guarantee that the necessary libraries don't interact with other pipelines. This is particularly true if you have multiple Python versions installed in your system. Please confirm that you are running clics3 in a virtual environment if you run into any issue. There are many different solutions on how to use virtual environments, with details depending on the system you use (for example, if you are using Conda). A good overview is provided here but, for most systems, it should be enough to create an environment with

   python -m venv env

   And activate it with

   source env/bin/activate

   Once all the work had been carried, you can leave the virtual environment by closing your shell or issueing the deactivate command.

1. Install version 3 of the Python package pyclics :

   pip install "pyclics>=3.0"

2. Download and install the Lexibank datasets from which to aggregate colexifications:

   curl -O https://raw.githubusercontent.com/clics/clics3/master/datasets.txt
   pip install -r datasets.txt

3. Download data for the reference catalogs Glottolog and Concepticon:

   cldfbench catconfig

4. Create the SQLite database clics.sqlite:

   clics load --glottolog-version v4.0 --concepticon-version v2.2.0

   and confirm all 30 datasets have been loaded

   clics datasets

#	Dataset	Parameters	Concepticon	Varieties	Glottocodes	Families
1	abrahammonpa	304	304	30	16	2
2	allenbai	499	499	9	9	1
3	bantubvd	420	415	10	10	1
4	beidasinitic	736	735	18	18	1
5	bodtkhobwa	553	536	8	8	1
6	bowernpny	338	338	175	172	1
7	castrosui	510	508	16	3	1
8	chenhmongmien	793	793	22	20	1
9	diacl	537	537	371	351	25
10	halenepal	699	662	13	13	2
11	hantganbangime	299	299	22	22	5
12	hubercolumbian	346	345	69	65	16
13	ids	1310	1308	320	275	60
14	kraftchadic	433	428	66	59	2
15	lexirumah	604	602	357	231	12
16	logos	707	707	5	5	1
17	marrisonnaga	580	572	40	39	1
18	mitterhoferbena	342	335	13	13	1
19	naganorgyalrongic	969	877	10	8	1
20	northeuralex	952	951	107	107	21
21	robinsonap	391	391	13	13	1
22	satterthwaitetb	418	418	18	18	1
23	sohartmannchin	279	279	8	7	1
24	suntb	929	929	49	49	1
25	tls	1140	811	126	107	1
26	transnewguineaorg	904	865	1004	760	106
27	tryonsolomon	317	314	111	96	5
28	wold	1459	1458	41	41	24
29	yanglalo	875	869	7	7	1
30	zgraggenmadang	311	310	98	98	1
	TOTAL	0	2906	3156	2271	200

5. Create the colexification network (encoded as GML graph):

   clics -t 3 -f families colexification --show 20 --format pipe

   This will create the graph at graphs/network-3-families.gml and show the 20 most common colexifications:

ID A	Concept A	ID B	Concept B	Families	Languages	Words
906	TREE	1803	WOOD	59	348	361
1313	MOON	1370	MONTH	57	324	327
72	CLAW	1258	FINGERNAIL	55	236	243
1297	LEG	1301	FOOT	52	349	358
1352	KNIFE	3210	KNIFE (FOR EATING)	51	268	282
2267	SON-IN-LAW (OF MAN)	2266	SON-IN-LAW (OF WOMAN)	49	261	280
763	SKIN	1204	BARK	49	209	213
1307	LANGUAGE	1599	WORD	49	148	149
1277	HAND	1673	ARM	48	294	300
1408	HEAR	1608	LISTEN	48	107	109
634	MEAT	2259	FLESH	47	252	262
2265	DAUGHTER-IN-LAW (OF MAN)	2264	DAUGHTER-IN-LAW (OF WOMAN)	47	234	256
763	SKIN	629	LEATHER	46	236	258
837	BLUE	1425	GREEN	46	195	204
2261	MALE (OF PERSON)	2263	MALE (OF ANIMAL)	45	145	163
1199	WIFE	962	WOMAN	44	289	301
480	PLATE	481	DISH	44	155	170
2260	FEMALE (OF PERSON)	2262	FEMALE (OF ANIMAL)	44	146	154
1228	EARTH (SOIL)	626	LAND	43	159	167
667	ROAD	2252	PATH	43	133	153
705	GO UP (ASCEND)	1102	CLIMB	43	132	146
683	PERSON	1554	MAN	41	199	205
2255	FATHER-IN-LAW (OF MAN)	2254	FATHER-IN-LAW (OF WOMAN)	41	187	204
133	WEAVE	3294	BRAID (VERB) OR WEAVE (BASKET)	41	122	133
2261	MALE (OF PERSON)	1554	MAN	41	104	115
1474	SEA	645	OCEAN	41	101	110
215	LIE DOWN	1585	SLEEP	40	191	197
1265	HIGH	711	TALL	40	168	182
256	FOOD	1526	MEAL	40	124	136
1732	SKY	1565	HEAVEN	40	117	120
1443	WALK	695	GO	39	288	320
2257	MOTHER-IN-LAW (OF MAN)	2256	MOTHER-IN-LAW (OF WOMAN)	39	181	203
1618	GRANDSON	1619	GRANDDAUGHTER	39	133	151
1203	LONG	711	TALL	39	113	121
2260	FEMALE (OF PERSON)	962	WOMAN	39	109	119
948	WATER	666	RIVER	38	197	200
1229	OLD	406	OLD MAN	38	103	107
531	HOW MUCH	3450	HOW MANY PIECES	37	184	203
706	DARK	163	BLACK	37	95	97
855	SEIZE	702	CATCH	36	150	161

6. Run subgraph and infomap cluster algorithms:

   clics --seed 42 -t 3 -f families makeapp

   The clustered networks will be written to GML graphs and exported in a way suitable for exploring with the CLICS javascript app. We can get some summary statistics running

   clics -t 3 -f families --graphname infomap graph_stats
   -----------  ----
   nodes        1647
   edges        2967
   components     92
   communities   249
   -----------  ----

   Note that clustering may be non-deterministic, i.e. you may compute slightly different clusters than the ones distributed in the GML files in this repository.
7. Finally, we can explore the clusters in the CLICS javascript app:

   clics runapp