BotDCAT-AP: An Extension of the DCAT Application Profile for Describing Datasets for Chatbot Systems
2 likes515 views
The document introduces 'botdcat-ap', an extension of the dcat application profile aimed at enhancing chatbot systems by facilitating the integration of web information sources. It outlines the problem of existing frameworks lacking suitable solutions for this integration and proposes an architecture and vocabulary to describe chatbot information sources effectively. The proposed framework aims to improve the discoverability and reusability of these sources for chatbot applications.
BotDCAT-AP: An Extension of the DCAT Application Profile for Describing Datasets for Chatbot Systems
- 1. BotDCAT-AP: An Extension of the DCAT Application Profile for Describing Datasets for Chatbot Systems Paolo Cappello, Marco Comerio and Irene Celino Cefriel – Politecnico di Milano, Italy – www.cefriel.com
- 2. Introduction: what’s a bot Chatbots are software programs capable of simulating a conversation between humans and computers The main goal of chatbots is providing suitable answers, possibly carrying out specific actions, based on the context of conversations and users' intentions
- 3. Types of Chatbots The term “chatbot” refers to a variety of solutions that differ w.r.t some essential characteristics: information domain and level of understanding We focus on “Standard Bots”: • Interpret users’ utterances in Natural Language • Cover a specified domain Command-Based Bot AI Machine Standard Bot UNDERSTANDING DOMAIN OPEN COMMAND-BASED NLP-BASED CLOSED
- 4. State of the art – Bot Frameworks PROS: Communication channels and NLU engines are already provided CONS: Integration of (Web) information sources is left in the hands of the developer IBM WATSON CONVERSATION SERVICE MICROSOFT BOT FRAMEWORK
- 5. Problem statement & research questions Existing bot frameworks do not provide suitable solutions to integrate chatbot applications with external Web information sources Literature lacks a standard to optimally describe those sources for chatbot purposes 2 How to reduce the effort required by the integration of Web information sources into chatbot systems? 1 How to promote the reuse and discoverability of chatbot information sources published over the Web? DATA PUBLISHER DEVELOPER
- 6. Proposed solution – Architecture ENHANCED ARCHITECTURE DEVELOPER DATA PUBLISHER
- 7. Proposed solution – Vocabulary DCAT-AP BotDCAT-AP extension bot:http://swa.cefriel.it/ontologies/botdcat-ap
- 8. BotDCAT-AP vocabulary Vocabulary modelling process (similar to GeoDCAT-AP, StatDCAT-AP, …) Reuse of DCAT-AP to describe basic metadata about chatbot information sources Application profile (soft and hard constraints on the vocabulary usage) Extension of DCAT-AP with chatbot-related concepts Intents: goals from the user point of view and method/query from the data publisher point of view “tell me when next train passes on U2 line from Messe-Prater to Rathaus” Entities: named resources to be identified by the NLU engine in users’ requests “U2 is a line of Vienna subway, Messe-Prater and Rathaus are stops of the U2 subway line” Predicates to relate the various classes in the vocabulary
- 9. Evaluation – Talkin’Piazza Bot A chatbot application exploiting BotDCAT-AP descriptions of sources (e.g. events, POIs, and transport) to ease the development and maintenance processes
- 10. Conclusions and Future work BotDCAT-AP is a vocabulary to describe Web information sources so to ease the development of chatbot applications interacting with them Future improvements of BotDCAT-AP aim to increase reusability and interoperability of datasets for chatbots systems; some examples: Intent taxonomy, to describe possible intent categories (e.g. follow-up intents, fallback intents) Intent “normalization”, by creating an ontology that includes concepts related to the most common intents related to a specific domain (e.g. places, help desk, events, transportation, weather forecast, e-commerce, booking, reminders, etc.)
- 11. Thanks for your attention! Any question? bot: http://swa.cefriel.it/ontologies/botdcat-ap Irene Celino Cefriel – Politecnico di Milano irene.celino@cefriel.com