Studying lithic microdebitage with a dynamic image particle analyzer

Eberl, Markus; Johnson, Phyllis S.; Aguila, Rebecca Estrada

doi:10.1177/01976931221109301

Cited by 4 publications

(11 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Most of the variables measured are dependent on one another, and some even use the exact same measurements. Our investigations have revealed that transparency is overwhelmingly the most significant variable for differentiating between these two particle types (Eberl et al, 2022, this issue; Estrada Aguila et al, 2022, this issue; Johnson et al, 2021).…”

Section: Methods: Present Studymentioning

confidence: 83%

“…DIA was conducted at Vanderbilt University using the PartAn 3D particle analyzer (see Eberl et al, 2022, this issue) (Figure 6). Each sample was individually poured into a funnel at the top of this analyzer and then passed through a vibrating tray until each particle was dropped individually into the measurement area (Figure 7).…”

Section: Methods: Present Studymentioning

confidence: 99%

“…The results indicate that the random forest model has the greatest accuracy of the four algorithms applied, at 83.5 percent of microdebitage correctly identified, with only 0.9 percent of microdebitage misclassified. Following Random Forest, XGBoost and Glmnet performed best, with Naïve Bayes having the least accurate models (Eberl et al, 2022). Though we expected to see the same results when applied to actual archaeological soil samples, in the present study, we applied all four algorithms to 50 soil samples taken at Group 5PS-d at Tamarindito to test the accuracy of these models on archaeological contexts.…”

Section: Methods: Present Studymentioning

confidence: 99%

“…The tray gently shakes so that individual particles are moved to the edge of the tray, where they fall one-by-one between a lighted backlight and a high-speed camera. This camera takes several photographs of each particle as it falls, and up to 40 measurements are calculated from these photographs (Eberl et al, 2022; Johnson et al, 2021). Because this analysis is nondestructive, soils can then be stored for further analysis.…”

Section: Pilot Studymentioning

confidence: 99%

“…Because traditional methods for analyzing microdebitage are quite tedious and require significant time and monetary commitments, most archaeologists devote little to no resources to microdebitage analysis, even if stone tool production is an important aspect of their research agenda. To address this issue, my colleagues and I have recently developed a novel method that combines dynamic image analysis and machine learning for analyzing microdebitage in archaeological soil samples that significantly reduces the time needed for analysis while increasing the reproducibility and robusticity of the data (Eberl et al, 2022, n.d.; Estrada Aguila et al, 2022; Johnson et al, 2021). Using experimental microdebitage, we have a demonstrated accuracy of up to 83.5 percent.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Using tiny artifacts to answer big questions: Machine learning, microdebitage, and household spaces at Tamarindito

Johnson

Eberl

Aguila

et al. 2022

North American Archaeologist

Self Cite

View full text Add to dashboard Cite

The spatial analysis of microdebitage (measuring less than 6.3 mm) can identify areas where stone tools were knapped at archaeological sites. These tiny artifacts tend to become embedded in the locations where they were first deposited and are less vulnerable to post-depositional movement, making microdebitage an important artifact class for identifying primary areas of stone tool production. Traditional microdebitage analysis, however, can take multiple hours spread over several days to complete. Because of this, microdebitage analysis is typically completed in very small areas of sites due to the intensive time and labor commitment required. Recently, however, my colleagues and I have developed a novel, interdisciplinary method that combines dynamic image analysis and machine learning to analyze microdebitage taken from soil samples at archaeological sites. Analyses of experimental microdebitage demonstrated that microdebitage could be accurately and efficiently identified within archaeological soil samples using this method. In the present study, we apply these methods to soil samples taken from the Maya Capital of Tamarindito in Guatemala to verify whether these methods remain accurate when applied to archaeological contexts.

show abstract

Section: Methods: Present Studymentioning

confidence: 83%

Section: Methods: Present Studymentioning

confidence: 99%

Section: Methods: Present Studymentioning

confidence: 99%

Section: Pilot Studymentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Using tiny artifacts to answer big questions: Machine learning, microdebitage, and household spaces at Tamarindito

Johnson

Eberl

Aguila

et al. 2022

North American Archaeologist

Self Cite

View full text Add to dashboard Cite

show abstract

Redefining lithic microdebitage with experimental archaeology

Eberl,

Johnson,

Estrada Aguila

et al. 2023

Archaeol Anthropol Sci

View full text Add to dashboard Cite

Machine Learning–Based Identification of Lithic Microdebitage

Eberl

Bell

Spencer-Smith

et al. 2023

Adv. archaeol. pract.

Self Cite

View full text Add to dashboard Cite

Archaeologists tend to produce slow data that is contextually rich but often difficult to generalize. An example is the analysis of lithic microdebitage, or knapping debris, that is smaller than 6.3 mm (0.25 in.). So far, scholars have relied on manual approaches that are prone to intra- and interobserver errors. In the following, we present a machine learning–based alternative together with experimental archaeology and dynamic image analysis. We use a dynamic image particle analyzer to measure each particle in experimentally produced lithic microdebitage (N = 5,299) as well as an archaeological soil sample (N = 73,313). We have developed four machine learning models based on Naïve Bayes, glmnet (generalized linear regression), random forest, and XGBoost (“Extreme Gradient Boost[ing]”) algorithms. Hyperparameter tuning optimized each model. A random forest model performed best with a sensitivity of 83.5%. It misclassified only 28 or 0.9% of lithic microdebitage. XGBoost models reached a sensitivity of 67.3%, whereas Naïve Bayes and glmnet models stayed below 50%. Except for glmnet models, transparency proved to be the most critical variable to distinguish microdebitage. Our approach objectifies and standardizes microdebitage analysis. Machine learning allows studying much larger sample sizes. Algorithms differ, though, and a random forest model offers the best performance so far.

show abstract

Studying lithic microdebitage with a dynamic image particle analyzer

Cited by 4 publications

References 26 publications

Using tiny artifacts to answer big questions: Machine learning, microdebitage, and household spaces at Tamarindito

Using tiny artifacts to answer big questions: Machine learning, microdebitage, and household spaces at Tamarindito

Redefining lithic microdebitage with experimental archaeology

Machine Learning–Based Identification of Lithic Microdebitage

Contact Info

Product

Resources

About