I've extracted all the dialogue captions from the game's *.dlg files and posted them in a github repository here: github.com/CahootsMalone/starship-titanic-stuff

Some of the dialogue appears to be unused. For example, I was unable to persuade the BellBot to utter his remarks regarding various contemporary politicians.

The repo also contains a short Python 3 script I wrote (github.com/CahootsMalone/starship-titanic-stuff/blob/master/scripts/dlg-parser.py). It performed the caption extraction and can also extract the wave files if the "save_waves" variable near the start is set to True.

Amazing!

Is there any way to get the STVOCAB.TXT list?

I'm having a hell of a time trying to figure out what the text parser actually wants. I want to try to record a complete guide for how best to interact with the text parser.

I don't actually see a file named STVOCAB.TXT in my installation of Starship Titanic; however, this may be because I downloaded it from GOG in September 2021 (GOG has apparently modified it at least once since then, e.g., see this thread) and/or because I used innoextract to extract files from the installer (to run the game with ScummVM as described here: https://github.com/CahootsMalone/starship-titanic-stuff/blob/master/README.md) rather than installing the game the normal way. Speaking of ScummVM, it looks like STVOCAB.TXT is actually created by the ScummVM Starship Titanic Engine (see https://github.com/scummvm/scummvm/blob/master/devtools/create_titanic/create_titanic_dat.cpp#L1560), which makes it somewhat odd that I don't have it. Where is it located in your installation?

Coming back to your main question, I looked into this topic a while ago (like you, I'm also curious about what keywords the text parser looks for) and if I remember correctly the data related to text parser keywords is stored in the game's executable.

Taking a quick look at ST.exe in HxD (a hex editor), it looks like some keyword-related ASCII strings are present from decimal offset 3,565,604 to offset 3,643,937. For example, the words "baggage", "luggage", and "suitcase" (as well as misspelled versions, which I assume were intended to catch spelling errors players might make) appear near offset 3,570,126.

These strings might be worth taking a look at, although I doubt every word the parser responds to is here (for example, I recall that entering "zaphod" when talking to the BellBot would elicit a response regarding the player having "the wrong universe", but the word "zaphod" doesn't appear in the executable, at least not in uncompressed form).

It may also be worth taking a look at the code for the ScummVM Starship Titanic Engine (https://github.com/scummvm/scummvm/tree/master/engines/titanic) mentioned above as its creators have presumably figured out how the text parser works!