This thesis fits into the first operating years of the Large Hadron Collider. This monumental machine was built to explore the infinitesimal structure of matter at the multi-TeV scale. The LHC aimed primarily at searching for the Higgs boson, the discovery of which would confirm the electroweak symmetry breaking model. This mechanism, which provides W and Z bosons with a mass, describes the transition from a unified electroweak interaction to a weak interaction (short range) and an electromagnetic interaction (infinite range). The LHC's proton collisions, operated at a 50 ns period, are analysed by 4 large detectors, including the Compact Muon Solenoid (CMS). This small period allows to observe very rare phenomena, such as the Higgs boson production and decay, but it requires a fast online selection of the interesting collisions: the trigger system. The computing resources available for the data's storage and analysis set a limit to the trigger rate. Therefore the bandwidth, which is split into several physics signals, must be optimised. Firstly, I studied the electron trigger: electrons are a clear signature in the intense hadronic environment within the LHC and allow a high measurement accuracy, as well as a search for rare signals. Besides, they are part of the final states investigated by a large number of analyses (Higgs, electroweak, etc). From the first collisions in 2010, anomalous signals in the CMS electromagnetic calorimeter (ECAL) were a source of uncontrolled trigger rate increase. Indeed, their production rate increased along with the collisions' energy and intensity: they were likely to saturate the bandwidth as early as 2011, crippling drastically the CMS physics performances. I optimised the anomalous signal rejection algorithm, while conserving an excellent electron triggering efficiency, as regards the data collected in 2011. Moreover, the increasing intensity of the LHC collisions causes a loss of transparency in the ECAL crystals. The setting-up of weekly corrections to the ECAL trigger calibration helped make up for the inefficiency caused by this loss of transparency. Secondly, I contributed to the search for the Higgs boson decaying to 2 tau leptons. So far, this analysis proved to be the only possible method to check the coupling of the Higgs boson to leptons. The tau lepton decays either into lighter leptons (electron or muon), or into hadrons: hence the study of six final states. I focused on the semileptonic final states, in which the expected signal is the most statistically significant. The trigger algorithms dedicated to this analysis select a lepton and a hadronic tau, with high transverse momenta. However, this selection removes half of the signal, which motivated the elaboration of new algorithms selecting low momenta leptons, including a cut on the missing transverse energy. This cut helps controlling the trigger rate and selects events containing neutrinos, which are a distinguishing feature of the tau lepton decay. The invariant mass distributions for all background and signal processes allow to quantify the compatibility between the acquired data and the presence of a signal. The combination of all final states leads to the observation of an excess of events over a large mass range. Its statistical significance is 3,2 standard deviations at 125 GeV ; the boson mass measured in this channel is 122 ± 7 GeV. This measurement is the first evidence for a coupling between the Higgs boson and the tau lepton.
