I suspect that you could do so; but not one that escapes (detectable) requirements for structural complexity, energy use, and various other 'life-like' properties. With enough bodging, you probably could get some sort of complex neural network into a plant. I'm not sure how fast you could run it, given the metabolic demands of a high-end nervous system; but even if it were a very slow sentience, it would presumably qualify. Such a plant, though, would raise serious eyebrows on anatomical inspection.
Similarly, mostly-silicon crystals with complex doping patterns and metal layers look almost nothing like silicon crystals with stochastically scattered impurities, except in bulk. Even if they weren't sentient, if we discover that the crystals of Ganymede show complex internal dopant patterns and electrical activity, and have surfaces that promote deposition of materials from the environment to expand the crystal, they'd definitely have a good shot at being 'life' or at least 'not life, because something; but a very, very interesting phenomenon indeed'; but that sort of thing would also be detectable, with the right instrument package.
I should probably have clarified this orginially; but it's the "What if something we've examined up and down for interesting functional structures, with no luck, is capable of those functions anyway, because vitalism!" school that irks me, not the suggestion that quite novel chemical/structural substrates might be more than capable of doing chemistry-akin-to-life. To the best of my knowledge, we've had very limited luck prototyping or theorizing such things in any detail; but it'd hardly be the first time that evolutionary brute force has shown us a solution more elegant than we would ever have expected.
At the risk of rambling, I'd give the example of Mycorrhizal networks(a subject near to my heart, if fungi had those). I suspect that, until fairly recently, people wouldn't have expected 'just plants' of communication; but it is experimentally demonstrable that they do communicate, albeit about a fairly limited range of things(pests, mostly). Once demonstrated, time to go hunting for the method. And, once we started hunting, we found at least a few, the fungal networks being one, and I think some amount of quasi-olfactory sensitivity to chemicals released by a plant under attack by plants nearby.
The point is not to dismiss the capabilities of everything that doesn't 'look' like it should be capable of something; but to keep in mind that 'life', 'communication', 'sentience', etc. can be accomplished by a variety of fascinatingly clever techniques, some almost certainly too subtle or unexpected to be detected by relatively primitive fixed-function chemistry sets on robots; but that all those things do impose certain requirements: can your 'life-but-not-as-we-know-it' shovel entropy into the world and energy out of the world fast enough to keep itself together? Your 'communication' candidate (even if we don't know what it's saying, or how it works yet) needs some sort of signal channels, electrical, chemical, acoustic, optical, magnetic, something, that should show up with suitable sensors. In the absence of channels, or if all we see is silence and thermal noise on all the channels, I doubt communication. 'Sentience' seems overwhelmingly likely to require considerable connectivity across a relatively complex set of structures, the details might be a surprise; but neither communication nor complexity should be able to hide, even if we don't have a damn clue about how 'sentience' works even in our own brain-meat.