The Effort to Save an Endangered Language
An Interview with Professor Sylvia Schreiner about her work on St. Lawrence Island Yupik.
From GMU Faculty Website: Sylvia Schreiner is a linguist specializing in syntax, morphology, semantics, and language documentation. Much of her work to date has focused on the morphosyntax and semantics of tense, aspect, mood, and modality. She has worked on documenting Scottish Gaelic since 2007 and is now co-leading an NSF-funded project documenting St. Lawrence Island Yupik.
A few months ago, I had the opportunity to speak with Professor Schreiner about her work on St. Lawrence island. Speaking with her gave me a new perspective on the importance of language to a culture and all of the efforts that go into language documentation and the revitalization process. Here is a slightly edited transcript of the interview:
Z. Could you talk about how you got personally involved in linguistic fieldwork and documentation?
S: When I was a graduate student at the University of Arizona, I was a student of Professor Andrew Carnie who is a syntactician. He had recently gotten an NSF grant to work on documenting Scottish Gaelic and he needed a research assistant. So I started working with a Gaelic speaker who lives in Arizona for part of the year doing “fieldwork” except it wasn’t in the field, it is in an office in a university. I started with just the one-on-one interview kind of thing.
After I left grad school, I kept working on Gaelic. In 2016, I was working at the University of Illinois and a colleague of mine came to me and asked me if I wanted to work on this native Alaskan language with him. That got me into the Yupik project.
“..they’re feeling the loss, they’re feeling the reduction in the past 20–30 years…It’s kind of scary to a lot of people.”
Z. So what is the cultural significance of a language like Yupik on St. Lawrence Island?
S. To the speakers?
Z. Yeah to the speakers.
S: My colleague lived there for a few years when he was a kid. His parents were teachers at the school for a few years. And when he was a kid there in the mid-80s, his little friends came to school basically not speaking English. They had Yupik in the home and they came to school and learned English at school. In the mid-90s, there was a dramatic shift that kind of coincided with more people having television and satellite television. You know, just having more media and outside pressure. Such that nowadays, most kids are learning English at home and coming and essentially getting a Yupik class in school like you’d get a Spanish class in second grade like once a week. I think maybe they do it every day, but it’s like a subject.
Basically, there’s a big split between people who are about 40 and older and people who are younger than that where, for the people who are 40 and older, almost everyone grew up speaking Yupik and didn’t learn English until they were in school. And people under that, it’s not everyone grew up speaking English, but it’s much different
So something that people talk about, like in terms of culture, is that they feel really connected, older generations especially, to the language and feel that it’s part of who they are and how they live. One thing about this group of people is that, unlike a lot of native groups in the continental US, and even in continental Alaska, did not ever lose their land. They’ve been on St. Lawrence island for many generations. In kind of, not prehistory, but a couple thousand years ago, it wasn’t ever a permanent settlement, the native people who were either from the Russian side or the Alaskan side, used it as a hunting ground, kind of. My understanding is like families — maybe a particular family — may have lived there for their entire lives, but there weren’t any permanent settlements there until just a few hundred years ago. But it was always the same groups of people who would have their little settlements around the island. The point is, they’ve never been moved off their land by the American government. They own the island in fact. The native corporation owns the island. So it’s a privately held island that’s part of the US. Their culture and their way of life was very strong and still is, but even more so up until very recently.
This group of people still does traditional hunting, so they’re allowed to hunt whales; they’re one of the few groups. There are a couple of native groups in Japan and a couple of native groups in the United States who are allowed to harvest whales and a certain number they can. They hunt walruses and seals, and they gather bird eggs, and there are berries and greens on the mountain they go pick. Some of the houses still don’t have running water. But then, everybody’s got a cellphone. So it’s a really interesting combination of things where they’re still very much connected to their culture but they’re feeling the loss, they’re feeling the reduction in the past 20–30 years. One person said for instance that on the whaling boats, they’ll go out and it’s a big operation cause these are huge animals and it takes several boats and a person. [They] said, up until recently, everyone spoke Yupik on the boats because you knew all the words in Yupik for all the things you were doing. Just recently, within the past 5 or 10 years, a lot of boats had to switch to English because there were younger people coming in who don’t know the words in Yupik. That’s a life-or-death kind of thing. You can’t be saying stuff that everybody’s not going to understand. So it’s very literally part of how they live. And that’s shifting, and people have been starting to realize that recently, and it’s uncomfortable. It’s kind of scary to a lot of people.
Z: Yupik has a relatively small population so did environmental factors have a greater impact on them in comparison to if a language had a larger population of speakers? Are smaller populations more susceptible to outside forces?
S: That’s a good question. It gives you a smaller base of speakers in the first place. So there’s kind of a sister language — Central Alaskan Yup’ik — that’s spoken in mainland Alaska. (they put an apostrophe after the p which is most commonly worked on in comparison to St. Lawrence Island Yupik). So one thing that happened in the past was at the end of the 19th century, so in like 1890 something, there was a famine. There are arguments about what caused it; some people said that the western traders had come in and brought in alcohol and they didn’t put enough food away and some people said maybe some disease. Maybe a confluence of things. But essentially it decimated the population. It cut it by 70% or something. There were like 4 or 5 thousand people on the island and it cut it to 400. The population today of like 1400 has come back from that but, people are also going to the mainland. When you have more people, you can withstand more changes. And to the extreme, if you have a world language like English or Mandarin or something, you have the political power, you have the person power, and everything to be influential. Whereas when you have a small group, it’s harder to have a say in certain ways.
“..there will have to be, if not textbooks with a hardcover, texts and materials created, worksheets, all that stuff, and you have to think about things like how are we gonna assess them? Who has the language ability to assess them?”
Z. You mentioned that there was one resource, a pretty big one, the Yupik Grammar of Jacobson, but it was targeted mostly toward fluent college-level speakers. Were there any challenges in creating an educational system that was geared towards younger people who didn’t have such a great understanding of the language already?
S: Yes and it hasn’t been accomplished yet. Now, they’re doing a great job, but the challenge is that they’re essentially teaching it as a second language. Somebody we worked with quite a bit, she only speaks with her daughter in Yupik. So her daughter is completely fluent in understanding but she’ll only respond in English. Which is something you hear from anybody with a heritage language. Kids will speak how their friends talk.
In the 70s, they made a big curriculum called bilingual bicultural. It was English and Yupik and teaching like “this is how we gather these greens”; “this is how we hunt whales” because not everybody was getting this information at home anymore. No one was using it so it was just in binders at the school. One of the things we’ve done in the project is we’ve digitized all that. We’ve hand-scanned everything, so we’ve given that stuff back but, it’s not easy to completely rehaul a curriculum. Basically, the main goal that the community has at this point is to start an immersive curriculum in the school. Maybe you’ve heard of Spanish immersion or Mandarin immersion around here, it’s demonstrably the best way. The challenge of that is creating the curriculum because, if you don’t have any textbooks in the language, you have to create them, you have to translate the stuff, so it’s a whole thing. Right now, we’re working with the revitalization committee and the tribal council to try and maybe try and get some grants written so that they can get the curriculum going and try to get that going in the school.
One challenge in changing the curriculum is the school board. Because they’re part of the US, Alaska has certain standards. They have to take standardized tests. When we started working on this in 2016, we were told that the school board had not responded favorably when people had asked, but just within that time, the school board has changed. So last we’ve heard, the current school was much more amenable to that kind of possibility, so now would be, you know, “strike when the iron is hot”. Essentially there will have to be, if not textbooks with a hardcover, texts and materials created, worksheets, all that stuff, and you have to think about things like how are we gonna assess them? Who has the language ability to assess them? It’s this huge project.
“..if they’re playing with it, it’s gonna be a different Yupik than it was 200 years ago, but it’s not gonna be English, it’s still very different.”
Z. When documenting and working on this language that’s spoken by such a small population, how can you be sure that this language is as accurate as possible? Especially since a lot of the learners of this language will not be L1 speakers, is the Yupik that’s going to be taught in schools as close to as it was 100 years ago, or is it going to be slightly different now that there are new language influences like English or Russian?
S: Yeah so that’s a great question and you run into some attitudes that have to be worked with. Even in English, it’s not even old people, like anyone over 20, going “kids these days are changing the language”. And that’s every language of all time. Everybody gets upset but language changes constantly and we can’t stop it.
So concrete example: in the culture before, saying thanks was not a super common thing. There’s a way to say “I am grateful to you” but it’s not a super common thing like “thanks”. So the way you say “I am thankful to you”, it’s kind of a mouthful. Younger people, even people who speak Yupik fluently, have started abbreviating it. [They’re] playing with it. It’s shortened. Some of the elders have heard them say this and are like “that’s not Yupik”. You understand, because it’s changed and it’s scary. But basically, the elders see this as evidence that the language is going downhill. From my perspective, I think that’s great. It shows that the younger speakers are thinking about it as their own in a way. If they’re only copying the elders, if they’re keeping the frozen forms, that’s a bad sign. But if they’re playing with it, it’s gonna be a different Yupik than it was 200 years ago, but it’s not gonna be English, it’s still very different.
A very difficult question when you’re documenting, is “whose Yupik do you document?”. Some of our work has been like “hey, could you put this base with this ending?” (Yupik is a polysynthetic language, so it’s got a base with different morphemes or endings). It’s like how I remember people saying “oh yeah yesterday I… texted… someone?” You have to figure out how to add the endings. Some people are just earlier adopters than others. Some people are like “texted, sure whatever” and some people are like “no, I’m not gonna say that”. It’s the same thing where you have some [Yupik speakers] who are like “oh I could imagine saying that” and another person who is only like five years older [going] “no who gave you that?”. Then you go ask the elder and they’re like “oh yeah, it’s fine.” So there’s just so much variation and in fact, there is maybe more variation in this area for some reason than even usual.
Essentially, we’ve had to figure out who are the people in the community that other people are gonna expect that you talk to. We talk to various people, but you’ve got to consider the opinions of respected elders for instance.
One of my research assistants has digitized the dictionary and has put the dictionary into a website and wants to record people saying the words so learners or people who are less fluent can go listen to them. So the question is, who do you record? Who has the “good Yupik”? This is a town of 700 people, so everybody knows each other. Everybody’s at least a third cousin or less. So it’s so much politics, interpersonal politics. It’s not gonna be the same Yupik even in a generation, but in my opinion, and the opinion of most of the people we’ve talked to, it’s better to have some Yupik than no Yupik, even if it’s different from what your grandparents used. (It would be different anyway, but people don’t really get that until they start talking to linguists) So yeah, it’s definitely an issue, but [we’re] trying to get a variety of speakers and then talking to people who are kind of respected so that you have the consensus.
Z. So due to the advancements in computational linguistics in the last few decades, would this level of language restoration be possible like 50 years ago?
S: There have been a lot of improvements in even the last 10 years in people working on a low-resource language (languages that don’t have a lot of corpora, different corpus materials, different writings, even an alphabet…) So with those languages it’s difficult to create modern tools because modern computation tools depend essentially on large data sets. Basically, part of how you program [things like autocomplete] is making a beginning program and throwing tons and tons of English sentences at it and it can learn if you teach it that way. My autocomplete looks different from your autocomplete. It picks things up. But you can’t even get to that level until you get to what the basic structure is. And it can’t do that until you teach it, and you can’t teach it until you know, and you can’t know until you have enough stuff written down about it.
So one of the things that people are working on in computational linguistics is the technology for low-resourced language ways of bootstrapping from lower amounts of data. It’s like, if you have a basic data set, how can you fill in the blanks with tools rather than having to necessarily talk to a person each time? Basically, programs that can learn from small data sets.
So the question of “is this helping?” It’s difficult to say since it’s all happening so fast, but there are a lot more resources for endangered and low-resourced languages now than say 10, or 15 years ago. The dictionary, for instance, that’s not gonna help anybody become a speaker of Yupik instantaneously. But one thing we ran into was people in their 20s or 30s, who have some Yupik but it’s rusty, are anxious about asking elders how to say stuff or speak to them because the elders chastise them. It’s the same thing as trying to speak your home language and you’re mom’s like “you said that wrong” and you’re like … “well I guess I’m speaking English now.”
The dictionary was made in the 80s or 90s or something but it’s two volumes. It’s this huge thing that costs like 150 dollars if you buy it from the place that publishes it. There are like two copies in the whole town; there’s one in the library and an elder has a copy because we gave her one. No one has it. They made this great resource, but because of a lack of money and tech, people couldn’t access it. It’s published by the University of Alaska so we said “hey, can we make this into an online resource that people can share?” And they said yeah so now people on their smartphones can look up a word and try to use it and not feel like they have to go ask an elder and get chastised. There are also people who live off the island and some of them have kids and speak Yupik and they want to teach [them]. So I’m not sure that we can yet say that language technologies are solving the problem of language endangerment but I think they’re moving closer to at least being really useful for learning in a way that wasn’t possible before.
Q. So you mentioned that to analyze the different morphemes of words, you had a morphological analyzer and it had a success rate of 75% but the rest had to be verified through elicitation and fieldwork, so how important is the relationship between technology and linguists and is the goal that eventually linguistics won’t have to be there all the time and that this system will just work itself?
S: So the analyzer is basically two things. One use is just to help us do our research, but the other use is being the baseline technology for things like predictive text or spellcheck. So you’re saying can we get rid of linguists? I don’t want to stop doing this because I find it fun, but my ideal would be that they wouldn’t need us anymore. I mean they don’t need us now, but they would have all the tools they need for their revitalization work and plans without us needing to be there. Part of that would not need to involve the morphological analyzer but that’s part of our vision. For instance, my research assistant is working on his dissertation right now and his project is basically building an interface that brings together access to the corpus with all the texts that we’ve digitized: access to the dictionary, a user-friendly version of the analyzer. If someone’s first language is English, it can build the words for you so you can use them to practice. Building those things into an interface that’s simple for an everyday person to use could help with that effort.
There is one kind of emerging technology. The first analyzer is called a finite state analyzer which is basically a complicated computer program where [my colleagues] took the information from the Jackson Grammar [and] wrote a really long program that is just regular expressions based on the grammar. (if x, then y) . The grammar is not complete because no grammar is fully complete, but it’s also kind of a starting point. So you give it a word and if it [doesn’t understand], then we can go ask a speaker. The emerging technology, instead of a finite state analyzer, is called a neural net-analyzer and that’s a learning kind of thing. I don’t understand the details but essentially you say “hey, notice patterns” very generally.
Basically, with a finite state analyzer, it’s like teaching a kid to walk on different surfaces by saying “on grass, you pick up your foot like this and put it down like this and on concrete, you have to put it up a little less”. In contrast to the kid learning to walk by “okay, I pick up my foot enough to not trip and then put it down.” It’s generalizing. The neural analyzer, essentially you’re programming it in a way that goes like “notice when there’s a pattern and applies that pattern to the next iteration of your understanding.” It’s learning as it’s iteratively adding to its knowledge. This is in its infancy, but that is kind of a way to get at moving past having to interview speakers every time.
“They text each other in Yupik, they want spellcheck, they want a dictionary, they want language learning apps, they want a little bird that tells us how to pronounce things like most of us have.”
Q. So are you hopeful for the revitalization of this language and maybe the use of these programs in other endangered languages?
S: Yeah I’m concerned right now because I think we’re at a tipping point, we’re at a key point. Right now, if we don’t do anything, the language will be gone in a generation because the kids are not learning enough. But luckily, the community is coming to be aware of this issue as we came to be aware of it and they are very positive about our work with them and we’re grateful to them. It’s a big happy family most of the time. [Their revitalization group] did not meet during covid times, so it’s picking back up. I had less hope like a year ago because I was still under the assumption that the school board was gonna disapprove everything. From what I heard last time, now is the time to do stuff. I think we have a set of people that can get things done. So yes, I am hopeful, I think that the technologies that we’re working on will be useful. I mean, everybody wants stuff on their phone. They don’t want to read through a 400-page book. They want apps and programs. They text each other in Yupik, they want spellcheck, they want a dictionary, they want language learning apps, they want a little bird that tells us how to pronounce things like most of us have. Essentially right now it hinges on this group of people taking action but also in engaging the younger generations in a way that makes them interested in the language. We can leverage it to anybody and if we can get that going, I think that people will see it as a viable thing for the future. If it’s part of the modern age, it will keep going. Maybe in a different form than it used to be, but it can survive and thrive, rather than peter off into the pages of history.
Thank you to Professor Schreiner for talking with me! You can learn more about her and her work here:
