Academia.eduAcademia.edu

Figure 5 – uploaded by Vittorio Loreto

See full PDFdownloadDownload figure
Fig. 5. Cross-linguistic Pearson’s correlation for the different terms. The results are presented as sliding window averages with a window size of 250. We use the data obtained from WCSj, for producing these results.

Figure 5 Cross-linguistic Pearson’s correlation for the different terms. The results are presented as sliding window averages with a window size of 250. We use the data obtained from WCSj, for producing these results.

Related Figures (7)

Fig. 1. Cross-linguistic rank-frequency distribution of the terms. The inset shows the distribution of color inventory sizes for the 110 languages of WCS. We use the data obtained from WCS, for producing these results. A widely accepted method among linguists to quantify markedness is through the statistical frequency of occurrence of the terms across the languages. In the following, we study this cross-linguistic frequency distribution of the unique color terms* documented in the WCS. In particular, we define the cross-linguistic fre- quency as the total number of times a particular term has been used to name different color chips for all the speakers across all the languages. Figure 1 shows the rank versus frequency distribution. Clearly, one can observe that there are a few terms with a very high frequency that in principle correspond to the names that are “cross-linguistically” most basic followed by the rest of the low frequency color words. In the inset we plot the number of unique color terms present in a language (color inventory size) for all the 110 languages documented in WCS. Note that while most of the languages have around 9-10 color words, there are languages that have as low as 3 unique terms (Yacouba) and as high as 79 unique words (Mampruli).
Fig. 1. Cross-linguistic rank-frequency distribution of the terms. The inset shows the distribution of color inventory sizes for the 110 languages of WCS. We use the data obtained from WCS, for producing these results. A widely accepted method among linguists to quantify markedness is through the statistical frequency of occurrence of the terms across the languages. In the following, we study this cross-linguistic frequency distribution of the unique color terms* documented in the WCS. In particular, we define the cross-linguistic fre- quency as the total number of times a particular term has been used to name different color chips for all the speakers across all the languages. Figure 1 shows the rank versus frequency distribution. Clearly, one can observe that there are a few terms with a very high frequency that in principle correspond to the names that are “cross-linguistically” most basic followed by the rest of the low frequency color words. In the inset we plot the number of unique color terms present in a language (color inventory size) for all the 110 languages documented in WCS. Note that while most of the languages have around 9-10 color words, there are languages that have as low as 3 unique terms (Yacouba) and as high as 79 unique words (Mampruli).
Fig. 2. Rank versus frequency of the color terms within a language for nine different languages. We use the data obtained from WCS, for producing these results.
Fig. 2. Rank versus frequency of the color terms within a language for nine different languages. We use the data obtained from WCS, for producing these results.
Fig. 3. A real example taken from WCS; illustrating the process of calculating (S's,5.)’ and (ORs, s9)* for the snippet ci01,¢102,.--,¢130 of the binary vector. The language considered is Abidji and the term t considered is “lobu” (term abbreviation: LB). where (A refers to the binary AND operation. Figure 3 shows a real example taken from WCS, illustrating the process of calculating (S5,5,)' for the snippet C101; €102;---;€139 Of the binary vector. If there are N speakers in / then we define the overall similarity on the term t as the average S,5, across all the N(N — 1)/2 pairs of speakers.
Fig. 3. A real example taken from WCS; illustrating the process of calculating (S's,5.)’ and (ORs, s9)* for the snippet ci01,¢102,.--,¢130 of the binary vector. The language considered is Abidji and the term t considered is “lobu” (term abbreviation: LB). where (A refers to the binary AND operation. Figure 3 shows a real example taken from WCS, illustrating the process of calculating (S5,5,)' for the snippet C101; €102;---;€139 Of the binary vector. If there are N speakers in / then we define the overall similarity on the term t as the average S,5, across all the N(N — 1)/2 pairs of speakers.
Fig. 4. Average cross-linguistic similarity for the different terms. (a) Ratio of the real to randomly generated average cross-linguistic similarity (Scheme I). (b) Ratio of the real to randomly generated average cross-linguistic similarity (Scheme II). In both cases, the inset shows the real and the random similarity values separately. The results are presented as sliding window averages with a window size of 250. We use the data obtained from WCSj, for producing these results.
Fig. 4. Average cross-linguistic similarity for the different terms. (a) Ratio of the real to randomly generated average cross-linguistic similarity (Scheme I). (b) Ratio of the real to randomly generated average cross-linguistic similarity (Scheme II). In both cases, the inset shows the real and the random similarity values separately. The results are presented as sliding window averages with a window size of 250. We use the data obtained from WCSj, for producing these results.
Fig. 6. The proportion of color chips occupied by the terms corresponding to the basic color names. Each color on the pie-graph represents the percentage area it covers on the Munsell chart.
Fig. 6. The proportion of color chips occupied by the terms corresponding to the basic color names. Each color on the pie-graph represents the percentage area it covers on the Munsell chart.
Fig. 7. A hypothetical example illustrating the process of average cross-linguistic similarity based inverse rank sum calculation. r© denotes the rank of a particular chip c. The chips “G1” and “G2” actually belong to the “RED” area of the Munsell chart according to [12, 14].
Fig. 7. A hypothetical example illustrating the process of average cross-linguistic similarity based inverse rank sum calculation. r© denotes the rank of a particular chip c. The chips “G1” and “G2” actually belong to the “RED” area of the Munsell chart according to [12, 14].
Fig. 8. Average cross-linguistic similarity based inverse rank sum for the chips corresponding to basic colors. Each color on the pie-graph represents the percentage of the total inverse rank sum corresponding to the chip(s) used to name the color. In order to produce these results, the rank list is obtained from WCS, and the “color chip-to-rank” correspondence is obtained using WCS2.
Fig. 8. Average cross-linguistic similarity based inverse rank sum for the chips corresponding to basic colors. Each color on the pie-graph represents the percentage of the total inverse rank sum corresponding to the chip(s) used to name the color. In order to produce these results, the rank list is obtained from WCS, and the “color chip-to-rank” correspondence is obtained using WCS2.
Related topics:
Applied MathematicsComputer ScienceStatistical PhysicsComputational Cognitive Science
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved