A non-reciprocal fuzzy preference relation (NrFPR) can express partial relations of alternatives, including indifference relations, preference relations and incomparability relations, but cannot depict linguistic preference intensities. A probabilistic linguistic preference relation (PLPR) can represent preference intensities in forms of probabilities and linguistic terms, but the incomparability relations of alternatives were not defined in a PLPR. Given that the NrFPR and PLPR can overcome each other’s drawbacks, this study proposes the non-reciprocal probabilistic linguistic preference relation (NrPLPR). Six conditions are given to express the partial relations of alternatives. Since the $P$-cut of probabilistic linguistic term sets (PLTSs) is effective in the operations of PLTSs without information loss, we construct the $P$-cut matrix of an NrPLPR by an adaptive $P$-determination method. Afterwards, nine rules are provided to define the partially consistent NrPLPR. To repair the inconsistent NrPLPR, a two-stage consistency repairing process, containing the linguistic information and probability repairing stages, is introduced. In addition, a non-reciprocal probabilistic linguistic nondominated selection procedure is proposed to rank alternatives. A case study on selecting social donation channels under the COVID-19 outbreaks is given to demonstrate the applicability of the proposed method. A comparative analysis is done to show the effectiveness of the proposed method.

不可逆的模糊偏好关系（NrFPR）可以表达备选方案的部分关系，包括冷漠关系，偏好关系和不可比关系，但不能描述语言的偏好强度。概率语言偏好关系（PLPR）可以以概率和语言术语的形式表示偏好强度，但是PLPR中未定义备选方案的不可比关系。鉴于NrFPR和PLPR可以克服彼此的弊端，因此本研究提出了非对等概率语言偏好关系（NrPLPR）。给出了六个条件来表示备选方案的部分关系。自从$P$切概率语言术语集（PLTS）在PLTS的操作中是有效的，不会造成信息丢失，我们构造了 $P$自适应的NrPLPR截割矩阵 $P$测定方法。此后，提供了九个规则来定义部分一致的NrPLPR。为了修复不一致的NrPLPR，引入了包含语言信息和概率修复阶段的两阶段一致性修复过程。另外，提出了一种非对等概率语言非支配选择程序来对备选方案进行排名。通过案例研究选择了COVID-19爆发时的社会捐赠渠道，以证明该方法的适用性。进行了比较分析以显示所提出方法的有效性。