Korean is a language that heavily uses topicalization, which is a process of letting your listener/reader know what the sentence is about, so that you don't have to have a subject later on. You could, for example, say, As for me, at 7:00 going to a party. We used to get on our Korean students for using In my case at the beginning of every sentence they made, but they did it because that's how you would set up a sentence in Korean. The basic method of setting up a sentence is different. It seemed to us English speakers that it was much more efficient (and natural) to simply say I am going to a party at 7:00. But we are using an entirely different system.
So I was helping a Korean student the other day, and he had produced several sentences that said, in effect, According to this issue, the problem is x and y...I could see that this was in effect the same problem. The topic was the issue at hand, and he'd identified the topic. According to was the wrong expression; With respect to might have worked better; but the real problem was simply that it's very rare to even need to identify the topic in English; we have myriad other ways to put the reader's attention on a different topic. We tend not to use that way of establishing the topic.
The reason I mention this is that computers can't pick up the cultural preference of ways of alerting a reader to a new topic. The computer could make the According to expression grammatical; it could ensure that there was a noun after it, and that it wasn't the main structure of the sentence. But it couldn't tell him what it meant. And it couldn't tell him that in English one tends to establish the topic differently.
Machine translator grammarians are proud that they have been able to work in the grammar of concordances, such that their software will catch the fact that, for example, we generally say interested in rather than interested by or interested with. If the computer can catch the expression, or the opportunity, to put you on the same path as 90% of English speakers, then it has made your output make more sense to English readers and has made you a better writer. But if it's unable to get you to construct sentences differently, or it can't alert you to a subtle difference in what we take things to mean, then its job isn't done.
Our job as teachers is to do the jobs that computers can't do. We might take on the whole language and decide that it's best to teach students without ever using computers, and in this case we might be better off, because we're talking about the mindset of the person who sits down to write, and the tendency to start off in a certain way, or set up a sentence in a certain way. But our problem is that the student has had the computer with him/her for every moment of his/her learning, so the English he/she has been producing is already without the benefit of this direction of teaching, and this student is quite used to putting these topicalizing expressions in front of everything he/she creates, without the benefit of auto-correction essentially reminding him/her that it's inappropriate. It's similar to any habit-forming experience that basically had to be undone in the past.
We once had a group of Malaysian speakers who basically reported to us that 1) they spoke quite a bit of pidgin English in their lives before they had arrived in the US; and 2) in this form of English it was quite common to omit the -s ending in the process of matching subjects and verbs. It was difficult to convey to them how important this was in the construction of our language, when in fact, even if they were able to understand its importance, they were unable to undo years of bad-habit overlooking of the requirement. These days the computer automatically fixes the failure to match; it simply puts the -s on the verb and forces you to remove it deliberately. But what we are seeing is a higher level of bad habits, because students find it easier to create the basic sentences, allowing the computer to match subject and verb, and allowing the electronic dictionary to provide the necessary words; therefore they can make longer and more complex sentences that all have easily obtained, but not always appropriate, words. We might even mistake what they are doing for a very high level of sentence construction. It would be better understood as a high level of misguided sentence construction habits. And the computer doesn't recognize what it sees well enough to point that out.