In real-world-job speak, Oxbridge interviews are almost 100% technical (as opposed to behavioural, like the ones you may get if you apply to American universities). The goal of the interview is to evaluate your academic potential in the subject and your fitness to the tutorial system. This advice is based on my experience of maths interviews at Oxford, but it should be generalisable to other STEM subjects as well.
Anyway, best of luck! Getting an interview by itself is an incredible achievement that you should be proud of. No matter the outcome, you are going to do fine.
In a rough order of increasing difficulty:
Vlad Neacșu wrote Linguistics Olympiad: Training guide which is probably the first English book on LO. You should read it in the same way you would read a maths textbook—with a pen and paper.
Most people in the LO sphere classify LO problems into 5 types, ordered roughly by the frequency with which they appear in the IOL:
My advice for beginners is to master morphosyntax first. At least in Thailand, you could make the IOL team just by solving morphosyntax problems. On average, morphosyntax questions compose about 40-55% of the IOL, and they are even more prominent in lower level olympiads. Morphosyntax problems also have a more predictable marking scheme. If you get past a certain level, you are almost guaranteed to get at least 8-12 points in any morphosyntax problem, whereas most numbers/semantics problems are 0-or-20, all-or-nothing.
I draw the following vague equivalences between linguistics and maths olympiad: semantics = combinatorics, numbers = number theory, morphosyntax = algebra, and phonology = geometry. Obviously I prefer semantics.
First I will screen through the entire problem, paying particular attention to the descriptions (introduction, figures, and diacritics). Then I will look at the translations to understand what features are being tested, or more accurately, what are not being tested. For example, if all the translations are in past tense, then tense is not tested. After you isolate those features, I find that the best way to see the pattern is to abstract everything.
What do I mean by abstraction? At the simplest level, it is to write the translation using another notation so that you do not have to read the entire translation again. For example, one can write personal pronouns as pqr, where p indicates the grammatical person, q the number, and r the gender. Another, more sophisticated abstraction is drawing graphs in semantics problem and find an isomorphism between the graphs for the unseen language and the translations.
Once I am satisfied with my model, it is time to write a solution. LO solutions come in 2 parts: (i) answers and (ii) explanations. Many beginners make mistakes with part (ii), by either explaining too little or too much, but mostly too much. The best solutions in the IOL rarely exceed 2 pages. You should aim to explain just enough for a linguist to be able to construct your answers from your explanation. It is a plus if you can do this by using diagrams or abstraction (which if not common you should explain how it corresponds to the translations) because it increases clarity. Do not write how you came up with the model. You will never get any mark from it, and some markers will be really mad.
I like to write part (i) before part (ii), because it allows me to update my model if I see something fishy in my answers. Some people do the opposite, they will try to have a concrete model first before they answer the tasks.
Read this advice from UKLO: 20 tips by Ellie Warner.
I cannot myself remember the IPA table. At the IOL level, it is good to know the most frequently used symbols, especially the voiceless-voiced pairs of consonants and the vowel chart (for vowel harmony). At lower levels, you do not need to know that the IPA table exists. Of course knowing more is good, especially if you want to study linguistics, but there is a tradeoff between the time you spend learning the IPA table and the time you spend solving actual problems.
No. Although you need to know at least one language, and know it well, in the sense that you should at least understand the parts of speech in that language and how the particles of that language interacts. Otherwise there is no way you can solve an LO problem.
An exposure to other olympiads is advantageous. The most important translational skills are knowing how to get unstuck and how to do hard exams under tight time conditions. However, there is no singular pathway to LO. People at the IOL are among the most diverse group of people I have ever met in terms of educational background.
No. Throughout my LO journey I have always tried to think of a way to “teach” LO to a general audience, but I failed. I believe that the most time-effective way to improve your LO skills is to solve the problems.
I myself cannot, but some former Thai IOL contestants who are not me made the Facebook page ภาษาศาสตร์โอลิมปิก LOs Today, and they might be willing to help.
You may also want to consider joining an LO Discord server, the link of which I believe can be found on the mysterious place that we call the Internet.