A binary lifter translates binary code into an intermediate representation (IR), such as LLVM IR. With the growing popularity of binary lifters and their ability to produce semantically accurate IR, even more complex binaries can now be translated successfully. In this study, we focus on two types of LLVM IR generated by binary lifters: high-level style IR (HIR) and emulation style IR (EIR). Although HIR offers advantages for analysis, EIR achieves a higher functional accuracy but is less suitable for analysis. To address this trade-off, we propose two methods for transforming EIR into HIR, aiming to combine the functional accuracy of EIR with the analysis suitability of HIR.