Abstract: 【Purpose/significance】In the age of big data, the information retrieval (IR) models of the information retrieval systems in
different industries are required to have high precision. However, currently the research of the traditional IR models hit a plateau,
which is characterized by that the lately proposed traditional models always only outperform their baseline models by a very small mar? gin. Thus, the traditional models cannot well satisfy the need of the IR system users. Therefore, the IR models with high precision need to be explored urgently. In recent years, a new framework of IR models basing on the theory of digital signal processing (DSP) has been proposed and the models basing on this framework have been testified to have obvious advantage on precision compared with the tradi? tional ones.【Method/process】Accordingly, based on the DSP framework (DSPF), this research introduces the term weighting methods from the classic probabilistic models F2LOG and F2EXP and proposes two new IR models DSPF-F2LOG and DSPF-F2EXP. For testi? fying the precision of our models, this research conducts extensive experiments to compare our models with various classic baseline models based on various categories of standard datasets in terms of different metrics.【Result/conclusion】The experimental results show that our proposed models outperform the classic IR models and are comparable with the lately proposed DSPF-based IR models with the highest precision in the DSP-based IR models at present. The two high-precision IR models proposed by this research can improve the precision of the IR systems in different industries when being used for unstructured data.【Innovation/limitation】This study proposes two high-precision information retrieval models, denoted as DSPF-F2LOG and DSPF-F2EXP, based on the theory of digital signal processing.