This paper presents a novel method to construct symbol letters for Feynman integrals without prior knowledge of the canonical differential equations (CDEs). The method is based on the recursive structure of Baikov representations and the d log-integrand construction. It identifies rational letters using d log construction under maximal cut and selects combinations of Gram determinants in the irrational letters based on certain criteria derived from the recursive structure. The method has been validated by comparing it with existing results for two-loop five-point and three-loop four-point integral families. It has also been applied to several exceptionally intricate integral families that have not been previously documented in the literature. The method successfully bootstrap the CDEs for a two-loop five-point family with two external masses and for a three-loop four-point family with two external masses, which were previously unknown in the literature. The symbol letters obtained through this method can be used to bootstrap the CDEs for the corresponding integral families. This can be effectively accomplished with numeric IBP reduction and d log-integrand construction. The method has been implemented in a Mathematica package, which is currently in a proof-of-concept stage. The method is expected to be applicable to a wide range of similar calculations in the future. The paper also discusses the relationship between the method and the Schubert approach. The cross ratio derived from the Schubert approach is shown to correspond to the symbol letters obtained through the method. The method is particularly effective for planar integral families, where the symbol letters can be determined without the need for explicit construction of d log-integrands under maximal cut. The method is also applicable to non-planar integral families, although identifying the LS in non-planar families requires explicit construction of d log-integrands under maximal cut, which is not easily automated. The paper also discusses the possibility of exploring these singular points in a more systematic manner using multivariate discriminants and resultants. Such investigations could provide insights into an algorithmic implementation of the method for non-planar integral families. The paper also notes that there are situations in which CDEs exist, but the symbol letters cannot be expressed in the d log form. These cases typically involve LS with nested square roots. It remains an open question to determine whether these symbol letters can still be derived from specific Gram determinants. The paper also discusses the relationship between the method and the approach based on Schubert problems. The cross ratio derived from the Schubert approach is shown to correspond to the symbol letters obtained through the method. The method is particularly effective for planar integral families, where the symbol letters can be determined without the need for explicit construction of d log-integrands under maximal cut. The method is also applicable to non-planar integral families, although identifying the LS in non-planar families requires explicit construction of d log-integrands under maximal cut, which is not easily automated. The paper also discusses the possibility of exploring these singular points in a more systematicThis paper presents a novel method to construct symbol letters for Feynman integrals without prior knowledge of the canonical differential equations (CDEs). The method is based on the recursive structure of Baikov representations and the d log-integrand construction. It identifies rational letters using d log construction under maximal cut and selects combinations of Gram determinants in the irrational letters based on certain criteria derived from the recursive structure. The method has been validated by comparing it with existing results for two-loop five-point and three-loop four-point integral families. It has also been applied to several exceptionally intricate integral families that have not been previously documented in the literature. The method successfully bootstrap the CDEs for a two-loop five-point family with two external masses and for a three-loop four-point family with two external masses, which were previously unknown in the literature. The symbol letters obtained through this method can be used to bootstrap the CDEs for the corresponding integral families. This can be effectively accomplished with numeric IBP reduction and d log-integrand construction. The method has been implemented in a Mathematica package, which is currently in a proof-of-concept stage. The method is expected to be applicable to a wide range of similar calculations in the future. The paper also discusses the relationship between the method and the Schubert approach. The cross ratio derived from the Schubert approach is shown to correspond to the symbol letters obtained through the method. The method is particularly effective for planar integral families, where the symbol letters can be determined without the need for explicit construction of d log-integrands under maximal cut. The method is also applicable to non-planar integral families, although identifying the LS in non-planar families requires explicit construction of d log-integrands under maximal cut, which is not easily automated. The paper also discusses the possibility of exploring these singular points in a more systematic manner using multivariate discriminants and resultants. Such investigations could provide insights into an algorithmic implementation of the method for non-planar integral families. The paper also notes that there are situations in which CDEs exist, but the symbol letters cannot be expressed in the d log form. These cases typically involve LS with nested square roots. It remains an open question to determine whether these symbol letters can still be derived from specific Gram determinants. The paper also discusses the relationship between the method and the approach based on Schubert problems. The cross ratio derived from the Schubert approach is shown to correspond to the symbol letters obtained through the method. The method is particularly effective for planar integral families, where the symbol letters can be determined without the need for explicit construction of d log-integrands under maximal cut. The method is also applicable to non-planar integral families, although identifying the LS in non-planar families requires explicit construction of d log-integrands under maximal cut, which is not easily automated. The paper also discusses the possibility of exploring these singular points in a more systematic