<p>The field of quantum computing has evolved a lot in the past few years with doors opening for novel applications in a variety of fields. Consequently, natural language processing (NLP) has also been affected due to this disruption, leading to various proposals for doing NLP on a quantum computer. One such promising framework is DisCoCat, a linguistics-oriented model designed to capture the meaning of language without compromising its structural information. Recently, DisCoCat models have been run on real quantum hardware. However, a few issues in the current formulation have led to problems like the out-of-vocabulary (OOV) problem. In this work, our aim is to tackle this OOV problem with respect to DisCoCat-based quantum natural language processing. We first design a novel functor mapping from string diagrams, a primary construction in DisCoCat, to quantum circuits. Later, we test the correctness and validity of this new technique on state-of-the-art datasets for tasks such as sentence similarity and paraphrase identification. We present clear empirical evidence that this novel functor mapping not only performs better than the pre-existing technique in most cases, but also it does so with relatively fewer parameters, thus ensuring better scalability with improved performance.</p>