Language and Linguistics in OSINT

Speaker 1: Welcome to the World of Intelligence. A podcast for you to discover the latest analysis of global military and security trends within the open- source defense intelligence community. Now onto the episode with your host, Harry Kemsley.

Harry Kemsley: Hello and welcome to this edition of World of Intelligence by JANES, your host Harry Kemsley, and as always, my co- host, Sean Corbett. Hello, Sean.

Speaker 3: Hello, Harry, all the way from the US.

Harry Kemsley: A remote co- host. Hi Sean. So Sean, we have talked about so many things now in this podcast around the utility, the power potential of open- source and one of the things that we've never touched on to my recollection anyway, is the power of language within the open- source or indeed other parts of intelligence work. So I thought today we might try and combine the world of open- source intelligence as well as language and linguistics, can't even speak, my language has failed me. The guest we have today is a language expert and also an open- source analyst here at JANES, Claire. Hello, Claire.

Claire: Hi, good to be here. Thanks for having me.

Harry Kemsley: So Claire, just get us started for the audience to understand just how many languages do you speak?

Claire: I speak nine, but I always say that the little asterisks, they're all off various levels, right? So I speak Belgian French because of my mother. And then I think my next top few languages are Spanish, Arabic and Mandarin from undergrad, Persian from working with Afghan refugees, and then Russian, Somalian, German are all kind of on the back burner, but Russian I've been working on a lot more recently and I've also studied Hebrew, Greek, Pashto, and a little bit of Urdu, all to varying levels.

Harry Kemsley: All right, so I think that puts you in the bracket of a language expert and given your background in open- source derived intelligence, I think you're the perfect guest for this conversation. Now, to get us started and a little unusually for this podcast, I'm going to start us with a few quickfire questions, not many, just a few to set the scene and for these questions, which I will pose to you, Claire, initially, but Sean, stand by, buckle up, some are coming your way as well. You're allowed to answer yes or no or at a push given it's the first time we've tried this, a single sentence answer will do. So let's try this out. So first of all Claire, do you think language skills are vital for your work in the open- source environment?

Claire: Absolutely, yes.

Harry Kemsley: That's a two word answer, but I'll allow that for the first time. Do you think language training should be part of any analyst's preparation for work in the intelligence community, whether it's open- source or not?

Claire: Yes.

Harry Kemsley: Do you think translating tools, including the ones that are coming onto the market now through artificial intelligence, are useful?

Claire: Yes.

Harry Kemsley: And do you think they are sufficiently accurate? I suspect the answer is yes given the last one, but do you think they're sufficiently accurate for use of the intelligence world?

Claire: Most of the time with an asterisk and we'll get into what those cases look like.

Harry Kemsley: All right, thank you. Sean, as a man who, like me is linguistically limited, although for the record, I am speaking in my second language, my first language is Spanish. Sean, do you feel that language is a vital component of the intelligence trade craft that we've spoken of so many times?

Speaker 3: It is.

Harry Kemsley: Two- word answer, not bad for your first go at quickfire, one- word answer. Do you think language should become part of the onboarding of new analysts, the training of analysts as they prepare themselves to be intelligence analysts?

Speaker 3: I can't answer that in one word, but the answer is no, but with an asterisk as well and I'll say why later.

Harry Kemsley: All right, let's unpack the asterisks and the caveats a little bit then in this conversation because fundamentally, human beings, we communicate by many means, but the one that we are talking about today is the various languages that we speak and one of the things that I know from my own experience of translation tools and indeed speaking two languages is the amount that can get lost when you move from one language to another, be that physical languages like sign languages even into the written word or indigenous verbal languages. So given that we're trying to derive intelligence, insight, useful things out of the things that we're reading and seeing in the open- source environment, language skills must surely be important. So my first question to you, Claire, is given your propensity for speaking so many languages and that you work in the open- source environment to derive intelligence insights, how do you use language? What's your language skills primary utility to you as you're working in the open- source environment, Claire?

Claire: I think the first thing I would say is of course to conduct research, but of course when we're finding information, and I work on the GITI team, which is the Geoeconomic Influence and Threat Intelligence team. So we track Russian and Chinese business activity around the world. So a lot of times we'll find information on a Chinese or Russian company operating in a third country. So whether that's in Central South America or Africa, Middle East, Europe, etc. So a lot of times when the news of that information hits the English news cycle, it usually was released at least a few days prior to that in another language and potentially with even more information. So I remember one of the layers of our tools is the alerts layer where we'll write out a long form analysis on transactions that we deem strategically significant, and there was one, I remember last year it was a Chinese mining company in Latin America that was working on trying to get a mining license and the government approved it, and we were going to write an alert on it, but one of the requirements for an alert is it has to be time sensitive, it has to be done within a certain amount of time, otherwise it's not an alert anymore. But even though it's just in Spanish to English, which a lot of people think are so much closer and more similar to each other than other languages in terms of timing and linguistically speaking, that news article had actually originally come out months prior, but we only caught it in the English news cycle a few months later. So it was no longer time relevant. There's also additional information and the Spanish article that was not included in the English article. So finding property information is definitely the most important thing. A lot of times too, we can sometimes get PDFs, especially if we're looking at company structure like Chinese or Russian company structure on the websites, a lot of times they'll have PDFs on the website of what the structure looks like. You'll have to do keyword searching, you can't just plug it into Google Translate or you can't translate it because it's a document and you can't click on individual words. But keyword searching is also really interesting because it doesn't always work like it works in English. If I found an article talking about, let's say, Chinese Iranian relations and I want to find all mentions of referrals to China. I would have to type in C- H- I- N to find China or Chinese, and I could also type in S- I- N- O for Sino, but if I typed in China, I'm going to miss Chinese. If I type in Chinese, I'll miss China. Same thing is true in other languages. Naturally for Arabic, it's the exact same thing. So alsiyn is the word for China, but if I said foreign language, that's the adjective as such, the translation of Chinese and referring to a feminine noun, but if I type in foreign language, I'm missing foreign language and foreign language, and foreign language is the adjective one, but for the male noun. So you have to type in just the root foreign language, but it gets even more complicated when we get to actually the Chinese language. And the Chinese language is so beautiful, it's so profound and deep. So the way to say China in Chinese is zhongguo. Zhong means middle and guo by itself means country. So together, it literally means middle country and the reason why that is China is because when the language was first developing all these years ago, they saw themselves at the center of the world. So the center country. So that's why it's zhongguo, but if you take zhong by itself, it can be put into a number of other circumstances such as shui is the word for study. So zhongshui middle study means middle school. So you would think that oh, if you just like keyword search zhongguo into an article, then you would find all references to China, but that's not the case. So one example I thought of was if you were looking for let's say Chinese- Iranian relations, So we know that phrase. So we know China is zhongguo, we know Iran is yilang, relations is guanxi. So you would think it would be zhongguo yilang guanxi, China Iranian relations, but it's not. So they take out the second character each country. So it's actually zhongyi guanxi. So if you type in just zhongguo, you're going to miss that, but if you type in just zhong, then you might get a bunch of other words that are using the character zhong, but in referring to the word middle.

Harry Kemsley: I think that is probably one of the most complete answers to a question I've had in a very long time and actually brilliant answer. So what I think I've just heard is that not only is there a timeliness aspect to it, there's also an accuracy and a relevance aspect to the ability for a linguist to identify things that are non- linguist just simply couldn't, you just couldn't possibly detect it.

Speaker 3: Absolutely, and that's why I qualified my second question with a no because Claire will be too modest, but being a linguist is a real skill and takes natural capability. Not everyone can be a linguist, and so if you try and make everyone a linguist, you're going to miss all the nuance that just came out there and you need to keep it going and keep it practicing. So I think every analyst needs to be aware of the power of languages and have access to a linguist. So as you know, I used to govern all the military linguists within the UK, MOD and there were incredibly precious resources that you just wanted to keep there doing the same thing. Now that was career limiting for a lot of them because if we use link, but they were so valuable, so be I used to protect them and call my jewels and the crown. But not everybody can be a linguist just because A, the time, the effort it takes as well as the natural talent. So having an awareness of nuances and also understanding why you can't just send algorithms on it and just translate, and we've all been there, haven't we? So we've just had the PhD level version of what I was going to say is if you go to Google Translate, it will come out with all sorts of spurious things, some of which are quite amusing, but that just at a very base level shows you why it's not as easy as it thinks and why you can't just rely on the algorithms.

Harry Kemsley: Yeah, I absolutely agree with all of that. However, of course what we're going to need to do is find a way of bringing the expertise of the linguist into the conversation. I think what your answer a second ago gave us Claire, is an incredibly clear insight into lack of linguistic skill or understanding of potential nuances in a language means you could be missing a huge amount of the available content. So the benefit of language skills in the discipline of intelligence analysis isn't just the fact that you can read a primary source written in first language, it's that you can actually identify the key elements of it and use that as a platform for further investigation. What other advantages have you seen though Claire, given that you have so many languages, is there a particular advantage that gives you over and above somebody like me who only speaks one other language?

Claire: Yeah, absolutely. When doing OSINT, when conducting OSINT, sometimes while also get videos or interviews and Google Voice can get you so far when you've got Google Translate and you can hit the play button and put it up to your computer and try to get as much of it as possible, but last time I tried it just to see how much it hatched, it only caught 50% of what was being said. So definitely for understanding videos, interviews and audio. I don't think that the apps we have on our phone and our computers really meet the mark yet. However, another really important thing is understanding idioms and over- interpreted expressions, especially religious expressions because a lot of times to means it's the same with English if someone's never studied English before and they hear it's raining cats and dogs, they're going to be like, "What do you mean it's raining?" The first time I said it to my grandparents in Belgium, they're didn't understand it and it's the same in other languages of course. So I have actually a few examples that I want to test you guys with to see, I have a few idioms in Arabic and Chinese and I'll tell you the translation and I want to see if you can get what the meaning is behind them. So are you ready?

Harry Kemsley: Just before we start this, so in almost three years of doing podcasts, the podcast runners, the co- hosts, are now going to get challenged by a series of questions from the guests and just be clear about that.

Claire: Yes.

Harry Kemsley: Perfect, all right, let's do it. I'll go first.

Claire: All right, so the first one is Arabic and it's foreign language, which means the door is big enough for a camel.

Harry Kemsley: Meaning of a door is big enough for a camel.

Claire: foreign language.

Harry Kemsley: Is that a reverse logic that it isn't actually that big to get a camel, like the camel through the eye of a needle, is it related to that?

Claire: It is not, no.

Harry Kemsley: All right, crashed and burned. What does it mean?

Claire: All right, it means if the door can fit a camel, it can fit you too. So please leave.

Harry Kemsley: Oh, okay. It's a bit of an insult.

Claire: Yeah, if someone says it to you, it's either they're saying it sarcastically or being either culturally disrespectful when they want you to leave. So I actually saw this quote when I was sent for a project and it was related to a few people were talking on Twitter and they were saying this in reference to tourists who were apparently not respecting the culture and they said, " foreign language." The door can fit a camel on you as well and asking them to leave the country. So the second one I have for you, Sean. So this one I have not seen in formal articles, but I have seen it in stock managers, social media intelligence. So it's pretty famous praise as well. So inaudible used to actually say this and it's on my power lifting bill as well to give you a little pressure, but the Mandarin is foreign language. It means you cannot catch the cubs without entering the tiger's lair.

Speaker 3: So it's got to be something about you've got to face the big things head on before you can look at the little things, something like that?

Claire: Yes, so like no pain no gain.

Speaker 3: Right, awesome.

Claire: You got it.

Harry Kemsley: One nil to Sean. Okay, well just because I now feel it at a loss, I'm going to give one to you Claire from Spanish. I know you speak Spanish, you ready for this?

Claire: Sure.

Harry Kemsley: It's a phrase that I used to hear all the time at home when I would tell my parents, which kind of gives you the sense of it that I was going to do something and I was going to get something done and my mother would say, " foreign language." Which translates, when the frogs grow hair. What do you think that-

Claire: Like it's never going to happen, right?

Speaker 3: Yeah, never going to happen.

Harry Kemsley: Exactly, never going to happen. My mother knew me well enough to know that I'd say, " I'm going to do this." You're like, "It's never going to happen." Now, the point of these translations and these idioms I think is the culture, the first language first culture, understanding of what it is that you're engaging with that is otherwise lost. I think Sean, you and I can both to experiences we've had where a lack of cultural understanding has led to some pretty horrific outcomes in terms of actions arising from what we've seen. I remember and I think it's well documented and therefore not difficult to talk about in terms of releasing it. A scene where lots of machine guns were being pointed in the sky and fired at a gathering of people, which in that particular part of the world in that kind of gathering was a celebration to a wedding. It wasn't an attack on allied forces as it was regarded by the allied forces. That's a cultural understanding, and Sean, you and I both remember the days when we had legal advisors stood next to us and then later we had cultural advisors and those cultural advisors were there to help us understand the kind of things we were seeing, hearing, and reading in the environment. It sounds to me Claire as though I want to give Sean a moment to think of an example where that cultural advice was critical. I know there'll be some, Sean, I'll give you a moment to think of one. Where do you think cultural and language really pays the highest dividend, Claire? Is it in the initial collection of content? Is it in the analysis of the content or is it both? Where does that cultural and first language understanding really pay the highest dividend?

Claire: I think it's actually both because if you see an idiom religious expression that has a meaning outside of the words, then it might be referring to something that if you don't understand what it actually means, you're going to miss the reference. So one actually example I saw was going through Twitter the other day is the Arabic word for Nasabi or Nawasib, it's a word used by Shias derogatorily towards Sunnis and it refers to someone who hates the family of the Prophet Muhammad. So when you see that, if you're looking for a negative sentiment of maybe of Shias towards Sunnis, whether it's in Iraq, Lebanon or Iran, and you see that term Nasabi. If you don't know what Nasabi or the plural is Nawasib, if you don't know what that means, then you're going to completely miss that whole phrase. Another one, this is more in sacrament, but Halsh is like a derogatory term referring to Hezbollah. So if you're looking for negative sentiments from Sunni and Christian Lebanese towards Hezbollah and you see a bunch of tweets with Halsh and you don't know what that means, you're going to completely skip it. So I think when collecting, absolutely it's important because it helps you with that insight, but also of course for analyzing. If you're looking at a phrase and a series of phrases and you don't know what the deeper meaning behind it is, then you're going to miss it, right? And then there are some phrases, even like my friends and I used to joke about this one too, with an Arabic, it's inshallah, it means God willing, but there's two meanings to it. So there's the real meeting of like, " Yes, God willing, I want to see you tomorrow." Let's say I made plans with a friend and then I say, " See you tomorrow, inshallah." That means by God's will, I will see you tomorrow, but there's also the meeting that can be more sarcastic, like let's say you ask your mom if you can go spend the night at your friend's house. If she says inshallah, that means no. So she could be saying, God wills it, sure, but for me it's a no.

Harry Kemsley: It's a solid no, right.

Claire: So there's some terms like that too where they have double meanings and that takes a lot of cultural nuance. So I would say it's important in both collecting the intelligence but also in analyzing it.

Harry Kemsley: Yeah, I agree. Sean, give us an example if you can from your fairly extensive experience.

Speaker 3: Yeah, I mean there are quite a lot of examples and yeah, sorry and the Kulaks were really important, but putting them next to the linguists as well, the translators was so important. So for several reasons, firstly, so you didn't inadvertently offend people by the wrong customer, the wrong habit. I mean even in terms of using your right hand for certain stuff and all the rest of it, but there was also and this does lead into languages as well, is that there was a thing we used to call$ 5 Taliban and they were distinguished between basically in national sport in Afghanistan was to shoot at the infidel, and so the real Taliban would pay young men in their teens to take pot shots just to keep the security forces on their toes, literally pay them probably wasn't even$ 5, but for them it was like playing football. And then there was very much a distinguished between them and the real Taliban if you want who were the dedicated fighters and all the rest of it, and the Kulak along with the translators could tell quite quickly which ones they were because the$ 5 ones were locals and the other ones were coming across from the Pakistan border, and that was really important in terms of how you act against it and what sort of tools you used to do so.

Harry Kemsley: All right, so thank you. I particularly like the examples we've toyed with here in terms of how this cultural understanding, this first language, first cultural understanding is so important to the collection and the analytical stages as I suspected it would be and we've also talked earlier about the fact that linguists are rare commodities, the jewels in your crown I think you called them Sean, and that we're not going to get all analysts to a place where they can be strong linguists in the kind of languages and the spread of language we might need them to be. Consequently, in recent times, as one of my quickfire questions indicated earlier, we've moved, have we not towards machine driven translations. Now I think it's fair to say, Sean, that one of the reasons we've done that is because certainly in the English- speaking world or the French- speaking world, we seek to bring the intelligence sources to first our first language in order that we think we can work with it better in our first language, right? I don't believe my Spanish is strong enough for me to work in Spanish in first language for intelligence purposes, but I do believe my skill in English is good enough for that. Therefore, I'm asking machines to bring the translation for me so that I can engage with it. It's also addressing this lack of resources. But this now starts to ask big questions, does it not about the machine's ability to find those nuances, those keywords, those key idioms and their real meaning. So back to my earlier question, Claire, to revisit it. Do we think machines are able to do what they need to help the non- linguist identify what you've been able to do as a linguist? And if not, what can we do to mitigate it? What can we do to help that non- linguist work around these fallibilities of the machine translations?

Claire: Yeah, so I think the machine translation is definitely an opportunity because it also allows people to conduct research in languages that they don't speak or languages that they can read but don't speak, right? So all English speakers can technically read or put together all the Latin languages, but it doesn't mean that they know what they mean, and same with Urdu for me and Pashto and all these languages that use the Arabic alphabet, right? So I mean when we use machine translation, it definitely helps with the collection. And of course if you're looking for very specific data pieces or data points, then you might not need to know these idioms and stuff. If I'm just looking for a timeline of when a Chinese company was operating, let's say in Libya and I want to know what they were working on and for how long they were there and what the investment amount was, then knowing these idioms is probably not going to be that beneficial because it's not going to change the output. But if you're working in more sentiment pieces or trying to understand the sentiment of a population towards a particular group or towards the government, then of course you can miss lots of pieces. So I think how to mitigate that, I think creating quick guides, some languages, if it's Arabic, if we're talking about tension between Sunnis and Shias, knowing words on terms like Nasabi and telling the analysts to look out for these terms when they're conducting research. I've created a few quick guides for Chinese military lingo that we used when we were doing some work on Chinese PSMCs, and all of us have this guide of 10 to 15 Chinese words that are specific to the military vernacular.

Harry Kemsley: Sean?

Speaker 3: Yeah, I think the other advantage of there's inevitability, we have to use the translation just because of the scale and the pace that we now need to work up. There was so much data out there as we've talked many times before, you just can't literally manage all that data, doing it manually by translating and all the rest of it and again, it's a hugely time- consuming thing just to listen to transcripts, but so the back to the human machine tuning thing, but I think where you need to really get into nuance, and without going into detail, I had linguists who could actually distinguish a specific voice, let alone just generally. So this is that person and when you get to know that sort of person as well, you know whether they're emotional, you know whether they are actually lying to you, and it's the sentiment analysis if you'd like of understanding somebody might be legitimate or seem to be legitimate when they're just lying through their teeth. And you can in some cases just by their nuance, by their tone, by what they're saying, sort of understand that they are lying. So some of the linguists in some of the facilities we used to have in Afghanistan were very good at that indeed, and you simply can't do that with just machines.

Harry Kemsley: Sure. Actually, that brings to mind for me, Clare, the previous conversations we've had on this podcast about miss and disinformation. Misleading people by mistake, misleading people by deliberate act and what you just said, Sean connects with that for me in that if I know what reads well, what sounds right, what looks right in terms of a language, idiom use, nuance culture and so on, I can probably and this is a question, you could presumably spot where something has been created perhaps by a non- first language trying to convey a message in another language for some purpose, particularly disinformation. So presumably that ability in linguistic terms would help you identify missing and disinformation as well.

Claire: Absolutely because I think when we're talking about using machines for translating, especially when it comes to document sentimental analysis is there's also so many words that are used in social media and slang that are not used in the formal lexicon that these machines are trained in. There's a few words that they start using in Arabic that are copies of English, like cute and foreign language for best or foreign language for share that if you try to translate it directly, it's going to come across as not as correct. So absolutely, yeah.

Harry Kemsley: Yeah, as I recall, sentiment analysis is based on a series of words being ranked in if you like, happiness and sadness and that they were all ranked by a number of people and I don't know if that ranking is done in other languages, but presumably if it's not, you can't get sentiment analysis in first language, you have to translate into English first, and anybody that's done translation from one language to another with the machine as we've been about will know that a lot of things will be lost in that translation. So even the sentiment analysis might not be that accurate. All right, so start bringing this not to a close just yet, but certainly start bringing this to some sort of conclusions. I think what we've heard is that there is immense value in having linguistic skill. We've discussed that in terms of timeliness, accuracy of collect understanding of what we've collected for the analysis process among other things. I think we've accepted that machine translation is an inevitable and a necessity in order for us to do the work we need to do in this vast amount of information that's available to us, but is there a place now for us to start understanding where machines are very helpful and useful, vice where they're less? Is there a way we start to divide this human machine teaming? And also Sean, I'll come to you with this second question in a moment. Is there a place for the linguist, the machine and the standard non- linguist analysts together? They're working in a sort of triangle of trade craft. So first part of that question to you, Claire, is there a place there for that machine and linguist working together to really get the maximum amount of what's available in the information world?

Claire: Absolutely. I think the hardest part of what we do now is this scraping just because of the vast amount of data that there is out on the internet. So I think the first thing that we want to do is using the target language, whether it's Arabic, version of Chinese or Russian, etc., to conduct the analysis to find the pieces of information that we're looking for and then once we go from there, if we have other analysts who do the analysis, that's when we can work with the machine translation to get those to extract the pieces of information that we need. However, of course if it's sentimental analysis, that's when the linguist can step in and say, " Okay, let me do these parts." And then these other less soft pieces of data, hard pieces of data than the other analysts can do. So I think absolutely, so when it comes to scraping, I would use the target language, but then when we're looking for the pieces of analysis and if especially there are other analysts working on the project, that's when we can use the machine translation.

Harry Kemsley: Perfect, and Sean, in terms of the trade craft and that triangle of machine linguist and analyst?

Speaker 3: In an ideal world, of course the answers are firm yes, of course it is. The challenge becomes with as I say, the scarce resources, bringing them together as well. I keep saying this because I am one, analysts tend to be quite insular, quite introverted, getting people to work together genuinely like that all the time and actually there's a trick, it's actually quite complex. So you get your machine translated piece, then how do you integrate that with the physical personal translation and then get that nuance to the analyst. So that needs quite a lot of time and effort and resource intensive, but yeah, absolutely, if we can.

Harry Kemsley: So I'm going to ask each of you to give the audience as usual with this podcast one thing that you really want them to take away, and I think it's my turn to go first, Sean, I've had a couple of second or third places on this, so I'll give that one first and I think this for me starts to sum up this frankly really fascinating and useful podcast. It's that even if you don't have linguistic skills, even if you only have machines, you need to be aware that there are almost certainly, if not absolutely certainly, nuances that you are missing that you may need to be looking for in other ways, and I'm thinking Sean here in terms of the triangulation we should be doing, never using a single source, multiple sources, trying to get an understanding of something may be a mitigation to the lack of linguistic skill that some analysts will feel they have to really understand what's actually happening, what the real sentiment is. So for me, awareness of the nuance that's available to a linguist that wouldn't be available to you as a non- linguist and the continuing necessity for great trade craft to triangulate and find the answer. That's my takeaway for the audience. Claire.

Claire: I would say the value of languages is not going away anytime soon. I think a lot of people are saying that, oh, with AI and all these different translation capabilities that we have that linguists and languages that their value is going down, but as we've discussed, especially with Sackman and OSINT, the value of languages and culture is not going away anytime soon, but also with humans and human- to- human connections, that's a little bit out of what we were discussing, but when you can speak to someone in their native language, they go as Nelson Mandela that said, " When you speak a language someone knows, you speak to their head, but when you speak their mother tongue, you speak to their heart." And that is so true when it comes to, I mean of course business connections if you want to put that taste on it, but also when I was working with refugees, it was a night and day difference when I could speak with them in their mother tongue versus through a translator interpreter and I think that that value is not going to go anywhere anytime soon.

Harry Kemsley: Yeah, I completely agree and I think your challenge to us with those idioms in foreign languages proves that the languages evolve actually extremely quickly. That barely a week goes by when my own children don't come back in the house and say something to me that I have literally no idea what they're talking about and they are talking in English apparently, Sean, your takeaway.

Speaker 3: So that's ignore my sandwiches. Now, how does that translate? Anyway, but I would say actually it is slightly cheating because a bit nuance than you said, but I think this particular discussion by the way, which has been fascinating, just shows how complex true open- source intelligence where you are trying to use multiple sources or available sources is far more complex than people think, and there's a lot of, as we know, we'll just politely call them talented amateurs out there that think it's just all about reading documents then putting together an assessment just off stuff that they've read on the internet. It's so much more of that, and if you're going to do the proper trade craft, it's multi- source, so there's a reason why there are different agencies looking at different things in different ways, but it's bringing it all together and this is a classic case of having to do that.

Harry Kemsley: Yeah, I couldn't agree more with that last comment. Claire, what can I say? It has been a genuinely fascinating conversation and one that we are going to pull you back in to have another conversation about. I think what we'll do next time perhaps is actually go and get a case study where we look very specifically at where the linguistic skill that you bring gives a night and day interpretation of the same event and maybe we'll look at Shaun having two competing views of the same content, one from a linguist perspective and one without. I suspect all it'll do is reinforce the lesson from this podcast, which is in the open- source environment to your point, Sean, the variety not just of media types and sources, but the variety of language and then within languages, the variety of dialects, etc., means that for the open- source analyst without linguistic skills somewhere in touch, you've got a real challenge to really understand what's going on. Claire, thank you so much. A real pleasure talking to you. We will let you get back to work now. James needs you to earn your dollar, but thank you so much for your time today. It's been a really fascinating conversation.

Claire: Absolutely, thank you for having me.

Harry Kemsley: All right, and for the listeners, if anybody has any questions about anything we've said today or indeed any topics you'd like us to cover in the future, number of you have given us some ideas recently, which we'll follow up on soon. Thank you for doing so. Thank you for listening and we'll see you again soon, goodbye.

Speaker 1: Thanks for joining us this week on the world of intelligence. Make sure to visit our website, janes. com/ podcast where you can subscribe to the show on Apple Podcasts, Spotify, or Google Podcasts so you'll never miss an episode.