Please bare with me a minute. This will all make sense.
I will take you on a bit of a side trip before getting back to your question. I do this as a teaching moment and to make a point. So please stick with me. I will answer your question.
There are a few truths in this world that remain true regardless of any effort mankind may make. One truth is that social engineering never works. Why? Because there are always unintentional consequences. The more mankind engineers his outcome without regard to natural order, the more opportunities there are for failure and failure becomes assured. It is funny how those who aspire to engineer society forget the premise of evolution and yet those who practice social engineering will quickly espouse the theories of Darwin while ignoring them. It is a paradox so obvious if you look.
So why is it that so-called SEO experts are teaching social engineering content without regard to what is natural? Because SEO is premised upon the notion of keywords. This is funny, because this is not how search engines work. Instead, you have been taught to go against the natural order of search with the intent of engineering your own outcome instead of the outcome you should have. You have been taught by those who have ignored the evolution of search and opted for a more primitive and severely limited understanding of search that for the most part was never true and only true for a very short period of time.
Google, for example, was never about keyword matches. And Google was not the first. Nor will it be the last. All search engines that relied upon keyword matches dropped out of sight quickly after the success of Google upon it's launch. Why? Because Page and Brin were absolutely right in 1997 when the wrote in the original Google research paper that keyword matches were severely limited and returned poor results. It was at this time that semantic search began to appear on the web. A technology that has existed since the 70's was to finally find an ontology large enough and worthy of analysis. Semantics was a study waiting for the right opportunity and that opportunity was now found.
So what does social engineering have to do with semantics? The more you try and create your own outcome by manipulating language, the further you drive yourself from the success you should have.
Freud is right.
The term Fruedian should come to mind. If you were to write without regard to outcome other than saying exactly what you have in your mind without filter or manipulation, what you have done is to express much more than a subject. In all that we write, regardless of how careful we are, we leak information that is important to be said and understood. It is the human condition that we express not only through words but through signals what is inside our heads - what we really mean. This is inescapable. This is what makes us human and what separates us from others. It is what is appealing about ourselves and why some writers are more popular than others. It is the reflection of the soul in our communication that makes communication desirable.
Have I gone too far afield?
Not at all. Say what you mean and not what you think others want. This includes search engines. Cripes! You are trying to talk to a machine that understands what you are saying better than you can ever realize. And yet, you want to modify down your speech out of fear? What for?
Your example using semantics.
Fortunately, your example is rather simple. It is also perfect.
How to bake an Irish carrot cake.
I have formatted your example to a fully formed sentence. Punctuation and capitalization is extremely important in semantics. These are clues to meaning. While I am sure you wrote your example with the understanding of a sentence, I wanted to make sure that those who read this understand that the new trend of shortening or simplifying communication should be avoided. Communication is an effort that requires precision. This is an important point. One comma can really make all the difference in meaning if you are not careful.
Breaking down your example using the simplest understanding of semantics yields much. How to Bake contains a semantic subject, predicate, and object. The predicate to ties how and bake. Duh, right? Changing the predicate changes the meaning. How we Bake, for example, yields how we and we bake. We takes it's understanding elsewhere. Being a pronoun, we must be defined prior to use and refers to someone. In just these three words, significant meaning can be derived.
Bake an Irish carrot cake. also contains a semantic subject predicate and object, in this case, bake a cake. Irish and carrot are modifiers of cake. They describe what kind of cake we are talking about.
Semantics is broken down into segments of understanding. How to bake is the first segment containing a full semantic subject, predicate, and object and bake a cake is another. This is important to know. Segments are take by importance. How to bake is more important to understanding for two reasons. One, it appears first. Going back to Freud, what is most important to us we often state first. Second, it contains a broader understanding. When we compare the second segment to the first, we find it is more specific. Bake a cake separated that act of baking a cake from other forms of baking such as baking a cookie or a scone.
People communicate and perceive naturally according to how the mind works. For example, if how to bake is changed to instruction for baking the meaning becomes less clear. While semantics uses ontologies for disambiguation and instruction for baking would somewhat equal how to bake, it is the requirement in semantic analysis that all be understood. Instruction is not understood exactly the same as how simply because they have two similar but different meanings. This is the same for when a person reads a sentence. Why? Because semantics is the study of communication, language, and perception as it exists naturally. This is important to know because what and how you write has consequences.
Search engines use semantics to both understand the search query and content. While there are quite a few semantic analysis methods used, two primary methods serve you best. One is semantic topical analysis. Without getting too deep into topical analysis, assume that each word you use will be tied to a topic using an ontology. A word can have more than topic, however, with further analysis, disambiguation makes topical understanding clear. Each content segment, a sentence, paragraph, header, will be analyzed into a table of metrics. Why? Because topical analysis of the search query and content and matching based upon topical scores yields a result set that is rather clean in understanding intent though it may not match specifically. For this, PoS (parts of speech) come into play. Starting to sound familiar?
Using your example, How to bake is rather broad in meaning. Bake a cake is more specific. The modifiers Irish and carrot even more specific. This is important to know. The order of PoS analysis and matching follows. You were concerned with competing with how to bake. This is inescapable. However, bake an Irish carrot cake differentiates all of the how to bake matches into a smaller result set. Over simplifying it a bit, if I asked you to get me a red ball, you might go to the store where red balls can be found, go to the toy section, go to where the balls are kept, and extract the a red ball from all of the other colors. In the same way, semantics separates Irish carrot cake from all other forms of baking.
This is what signalling is about.
Careful use of semantic meaning in the title tag, header tags, and content signal what the content is about. Where you were taught to signal through the use of single terms (keywords), the reality is that signals are made through semantic meaning. Keywords alone have no meaning. For example, doing single word search yields poor results. However, the more conversational (sentence like) a query is will result in more specific matches. The more specific in semantic meaning a query is, the better the results become. This gives rise to the notion of semantic meaning as being more important than keywords. For example, What is mouse the click? will result in What is a mouse click? Why? Because search is not about keywords. If it were, the results would have been significantly different. Search is about meaning and intent.
A word on long-tail searches.
SEO experts will define long-tail searches as matching terms less used. In a manner of speaking this is correct, however, the meaning is more Freudian. Am I coming full circle here?
Remember that people will search and write differently. The more semantic meaning that can be drawn from content and search queries the more this becomes evident. Search is about intent drawn from semantic meaning. Long-tail searches are a result of intent where intent can be more specifically understood. The longer the search query, the more specific the search results. This is where long-tail lives. Long-tail almost always lives within content primarily because title tags and header tags are short in understanding.
I have always espoused writing naturally. Of course your language should be precise and should use terms related to the subject, however, with consideration to natural writing, you attract a more organic audience. Worrying less about how search engines may perceive what you have written means that you attract the right audience. Hence, the advice to write content for users and not machines is perfectly valid and powerful.
Competition in Search
Competition in search first comes with satisfying intent. Keyword analysis is really intent research in disguise. Keep this in mind. Do not search for keywords, but rather how people search for topics related to your content and use language to match intent. Here, you can use semantic meaning for stronger results. I did an experiment a long time ago. I wrote a paragraph plainly as I would naturally. From there, I added greater semantic meaning using my understanding of linguistics to make the paragraph more semantically meaningful without changing it's actual meaning. This included paying attention to semantic clusters and making each one as specific as it should be, ordering semantic clusters appropriately, using proper punctuation, using proper capitalization, etc. Guess which paragraph resulted in more search results? Did I just make your point? Not really. I made my point. While the meaning of the paragraph did not change, how the paragraph could be understood was. In otherwords, I used precise language to communicate meaning carefully.
Your example is actually a good one. However, there are some things you can do. For example, How to bake an Irish carrot cake at home. and, Recipe for baking an Irish carrot cake. How many how to bake queries are actually looking for recipes? Search intent is about matching content language to search query language. It is my belief that how to is over used when writing content. Partly this is because search ontologies will match how to to many other semantic segments. For example, using your example in search will result in recipes, however, you may find them lower in the SERPs than how to. If recipes is searched for more than how to for your subject, then recipes should be used over how to. Intent research becomes more important. And who is to say you cannot have both? Title tags are more important than header tags thus the title could contain recipes while the h1 tag contains how to.
Are you picking up what I am putting down?