The interference pattern on the screen is altered due to our observation, but how could this happen, as we have been watching it since we started the experiment? Is there any other meaning of 'observation' in this experiment?
It depends on what, exactly, is observed. In the standard version of the experiment, only the position of the electrons as they hit the screen is detected. In this case there are interference patterns. It's only when an additional observation is made, the passage of the electrons through one or the other slit, that the interference patterns disappear.
Let's compare two experiments. Experiment A is the classical double slit case. The wave function consists of two parts, $\psi_L$ and $\psi_R$, which are coherent and overlap on the detection screen. Thus an interference pattern results from the term $2\psi_L \psi_R$. Experiment B is altered with a detector that can unambiguously determine through which slit the particle went. The detector has two, orthogonal, states, $d_L$ and $d_R$. The total wave function now is $\psi_L d_L + \psi_R d_R$. The two parts are orthogonal, $2 \psi_L d_L \psi_R d_R=0$, due to the entanglement of the particles with the detector, so no interference occurs, even if the experimenter forgets to watch the detector. Interference can only occur if it cannot be determined through which slit the particle passes.