Searle says syntax is neither sufficient for nor constitutive of semantics, all a computer gets (eg from sensors) is syntax (tokenised shapes) therefore computers will never understand the world. Searle: "There is no way to get from syntax to semantics" (Minds, Brains and Science, p34); "digital computers insofar as they are computers have, by definition, a syntax alone" (ibid); "semantics is not intrinsic to syntax" (Mystery of Consciousness, p17).
The premiss that semantics is not intrinsic to syntax to me crystallizes the powerful appeal of the Chinese room argument. And all a computer gets according to Searle is syntax: in the Chinese room, this is the shapes of the Chinese ideograms that drop from the slot in the door and the simple manipulations based solely on symbol shape.
Strangely, Searle never discusses relationships between symbols. In his 1990 Scientific American article he calls a basket of symbols a "database". Elsewhere he calls input symbols "bunches". He never talks about relations between symbols. A database is symbols related together. An input stream is a temporal sequence, also symbols related - related by their adjacency in time as they enter the machine.
So sure, shape is syntax, but what about relations between tokenised shapes? Computers get the relations as well as the shapes. Why does Searle never discuss these relations? If they were part of syntax, surely he would have said so, but he hasn't. So they must be separate from syntax. A token has a shape, syntax, but the relations between tokens are a different matter.
Computers get the relations as well as the tokens related. Computers can react purely to temporal contiguity between input symbols and ignore the syntax, eg alternate input symbols can be stored in alternate locations. The shape, the syntax, of the symbol is irrelevant.
So Searle's premiss "digital computers insofar as they are computers have, by definition, a syntax alone" must be false and, hence the Chinese room argument is unsound. Is this a rebuttal of the CRA? Since there are symbols and relations between them inside the machine, we can grant that syntax is neither sufficient for nor constitutive of semantics, but point out that there is more than syntax inside the machine. Syntax could be a component. True, syntax alone can't yield semantics, semantics is not intrinsic to syntax, but semantics could be a compound of syntax plus relations, and digital computers might think.