Embedding the quotient of a vector space Q: Given a vector space $V$ over an arbitrary field and a linear subspace $W$, can we embed $V/W$ within $V$? 
An embedding would be an injective morphism $i: V/W  \to V $ which is an isomorphism from the domain to its image?
Also are there similar results on groups and embeddings of the quotients by their normal subgroups?

In the case of finite-dimensional real or complex vector spaces, we can use orthogonality, but I am unsure how this would work otherwise.
Related question: Is quotient space a subspace of the underlying vector space
 A: SO, let's back up. For groups or rings, there is no general guarantee that we can imbed a quotient into the group/ring itself. This is true exactly when the kernel has a compliment in the group/ring itself. 
Vector spaces are special though. They can always be written as a direct sum of their one dimensional subspaces (with respect to a given ordered basis). So, in this case where we have (I will only demonstrate the finite dim. case but it is similar for infinite dim. at least with countable bases)
$$V=V_1 \oplus V_2 \oplus \dots \oplus V_n$$
So, if we quotient out by a particular subpsace then we are just quotienting at by some ``subsum'' of the above essentially. That is, the kernel of that map (the thing you're quotienting out by) you have always has a compliment. That compliment is how we imbed $V/K$ into $V$. It takes very little work to show the isomorphism $V \cong K \oplus V/K$
A: Yes. Take a basis $\{e_i\}$ of $V/W$, and choose $f_i$ in $V$ such that $f_i$ maps to $e_i$ under the quotient map. The map from $V/W$ to $V$ linearly extending the above gives an embedding. If $\sum a_i e_i = z \in V/W$ (a finite sum) then the image in $V$ would be $\sum a_i f_i$. etc
A: I completely agree with the previous answers. However, please be aware that any such an embedding is actually a choice! So there is no canonical way on doing it (it is only canonical if you quotient out a trivial space, i.e. the whole space or 0). Furthermore, if you are interested, this property is usually referred to as semi-simplicity of a category (something quite powerful).
and to get back to Rhythmink's answer: an example for rings or groups were this is not possible is the map $$\mathbb{Z} \to  \mathbb{Z}/(p).$$
I leave it to you to show that this admits no leftinverse ,i.e. an embedding 
$$\mathbb{Z}/(p) \to \mathbb{Z}.$$
