A　novel　cross-step　word　segmentation　algorithm　is　proposed　to　process　real-time　traffic　information　represented　in　natural　Chinese　in　this　paper,　to　meet　the　urgent　need　of　real-time　traveling　information　service,　for　dynamic　traffic　information.　Considering　the　record　length　distribution　of　the　word　libraries　depicting　real-time　traffic　information,　this　algorithm　sets　corresponding　steps　of　word　segmentation　for　address,　direction　and　event　libraries,　and　improves　the　one　step　running　of　the　string　pointer　in　classical　Chinese　word　segmentation　to　flexible　multiple　steps　running,　so　as　to　aggregate　possible　Chinese　words　efficiently.　A　case　study　shows　that　the　proposed　algorithm　runs　10　times　faster　than　an　improved　MM　algorithm,　whilst　keeping　similar　accuracy　and　robustness.　The　authors　argued　that　the　presented　algorithm　is　greatly　helpful　to　the　automatic　and　intelligent　processing　of　the　real-time　traffic　information,　and　facilitate　the　development　of　travel　information　services.