Ontology-driven dialogue management in the spoken dialogue system A SOLUTION FOR AUTOMATION OF CALL-CENTERS
2 Problem statement Spoken dialogue systems still have some drawbacks Not flexible enough for end-user: catching only key words, not catching the real meaning Hard to support – the data is in databases, which makes it available only for programmers Only prototypes exists, still no good examples in use
3 Human labor is expensive Call center’s expenses are majorly in labor Competition is tough CUT THE COSTS!!! Industry cries out
4 Our response An efficient b2b solution for call-centers that allows to minimize human involvement in the process of technical problem diagnosis and problem solving by complete dialogue automation
5 What ontology holds Domain knowledge Linguistics Problem classification rules Dialogue management scenarios
6 Why bother with ontologies? Easy to integrate external data sources Transparency of data Simplifies dynamic classification Openness of data
7 Benefits: each gets his own share flexible and intellectual service Happy customers Transparent and simple to expand data model – no programming needed Happy customers (call centers) Employ non-coding linguists in development
8 How will it work? Call-center customer says: “I need a good Internet tariff in roaming”
9 Down the rabbit hole: classification funnel Call-center customer says: “How can I connect to the Internet without a sign-up fee for the service”
10 We need help! What libraries should we better use? What linguistic approaches and libraries to integrate? What could be tricky in using ontologies
11 Thank you!