Motivation

Frozenpaths.org aims to benefit Alaskan indigenous language preservation. Different web tools are included within the web application to support users in data visualization. Tools include a chatbot to help readers summarize the collected data, also allowing users to query the expert system database. Knowledgeable users have the option to get involved through the option for users to suggest words that may be at risk of expiration. Gathered findings are later collected within our optimal AI aggregation model, which helps preserve Artic information for future research, strengthening knowledge of related fields.

It has been reported that at the end of the 15th century over 300 American Indian and Alaska Native languages were spoken [1]. Thus, there is a demand for a language preservation tool that can process local language data efficiently. Optimal language preservation requires high quality data storage and data acquisition techniques to operate properly.

Languages can help represent culture, heritage and identity. The Frozenpaths.org mission is to raise awareness about endangered Alaskan indigenous languages and support efforts to document, digitize, and preserve Alaskan indigenous languages for future generations.

Introduction

• Definitions of the Arctic region may differ between communities.
• The Arctic region contains a mixture of cultures, values, and beliefs.
• Small differences inspire different communications: i.e, rare words, dialects, slang, etc.
• Supporting native speaker communities and helping preserve native culture.
• Collecting knowledge, language and preserving this data digitally can help strengthen a knowledgebase for Arctic resilience.
• The diminishment of Arctic sea ice has led to increased human activities in the Arctic, and has heightened interest in, and concerns about, the region’s future [2].
• AI applications may be increasingly powerful for prediction of operational decision making in the Arctic [3].

Methods

• Rare languages found in Alaska are at risk of going extinct. A data processing approach that is flexible and scalable for languages is needed to preserve languages in Alaska.
• Language preservation can be challenging for reasons including:
  1. different dialects motivating different language representations
  2. resource requirements to translate, teach, and engage the community to participate in a language data processing application
  3. older speakers' language may not connect thoroughly with younger speakers, causing a disconnect in meanings.
• The overlap of different language elements found in rare languages could benefit from cross examination in a historical context.
• An application that can operate in all areas of the Arctic such as areas with lower levels of connectivity due to remoteness is essential.
• Voice-to-text translation is a method that can assist in the preservation of language data without inducing translation challenges on the user.
• Digitally preserving this language data can improve the understanding of cultures in Alaska, such as experiences with harsh Arctic climates.

Pseudocode

(1) Users download the local model, database and functions packaged within the application. (2) Users create a local account and get assigned a unique user identification (UUID) (3) Users can add existing language data into local language records; or users can run voice-to-text and have language data recorded using English phonics. (4) Users can add English translations to collected phonics as well as update any misspellings in the column separate for input language. The language data records can be represented in 3 columns: [[Language Text][English][User Identification]] (5) Users can view the updated language records. (6) Users can explore built-in developmental tools such as the exploratory data analysis tool which applies principle component analysis (PCA). (7) Users can export the local language records database and export the database.

Results

1. Accents can be difficult to record within voice-to-text translation tools
2. The Google Voice API was the translation technology chosen due to way it would segment speech character by character in a reasonable manner. For example, if a user recorded the words ‘the aakaaluk’ into the voice-to-text, the result would be approximately be: [t][h][e][] [a][k][a][ ][l][u][k][e], although in the worse case the voice-to-text result was [t][h][e][ ][a][c][t][o][r][ ][l][u][k][e]. The English translation for ‘the aakaluk’ would be ‘the grandmother’.
3. Further consideration can be placed into how much data is required for optimal Alaskan AI and machine learning applications.

Conclusions

In conclusion, we propose an Arctic language data processing application. We approach language data preservation with flexibility, scalability and accessibility. We examine how Arctic language data processing can benefit Arctic resilience through the preservation of knowledge. Lastly, we research modern challenges for language data.

References:

1. United States Department of the Interior Indian Affairs. FAQs. (2017). Do all American Indians and Alaska Natives speak a single traditional language?
2. Congressional Research Service Report. (2022). Changes in the Arctic: Background and Issues for Congress.
3. Frazier, K. (2024). Arctic Climate Data Science: The Role of Artificial Intelligence in Supporting Operational Decision Making. Ted Stevens Center for Arctic Security Studies Special Report.