This preliminary study aimed to screen non-coding RNAs (ncRNAs) from plasma exosomes as a new method for cervical cancer diagnosis. Differentially expressed RNAs were initially selected from among a group of 12 healthy individuals (normal group) and a pretreatment group of 30 patients with cervical cancer (cancer group). Then, we analyzed the association between an ncRNA-mRNA network and cancer using ingenuity pathway analysis after secondary selection according to the number and correlation of mRNAs (or ncRNAs) relative to changes in the expression of primarily selected ncRNAs (or mRNAs) before and after chemoradiotherapy. The number of RNAs selected from the initial RNAs was one from 13 miRNAs, four from 42 piRNAs, four from 28 lncRNAs, nine from 18 snoRNAs, 10 from 76 snRNAs, nine from 474 tRNAs, nine from 64 yRNAs, and five from 67 mRNAs. The combination of miRNA (miR-142-3p), mRNAs (CXCL5, KIF2A, RGS18, APL6IP5, and DAPP1), and snoRNAs (SNORD17, SCARNA12, SNORA6, SNORA12, SCRNA1, SNORD97, SNORD62, and SNORD38A) clearly distinguished the normal samples from the cancer group samples. We present a method for efficiently screening eight classes of RNAs isolated from exosomes for cervical cancer diagnosis using mRNAs (or ncRNAs) altered by chemoradiotherapy.