Downloaded 274 times
Uploaded byOracle Developers
Implementing Your Own Chatbot Platform!
Messaging apps have become more popular ways for people to communicate than social networks or phone calls. As a result, chatbots are growing in usefulness, especially for customer service tasks. There are different types of chatbots, from task-oriented bots to predictive, data-driven bots. Current chatbots have capabilities for intent recognition, entity recognition, and sentiment analysis using machine learning, but still face challenges with ambiguity. Future chatbots may be able to pass the Turing Test by more closely resembling human conversations. Oracle's Intelligent Bot platform includes components for natural language processing, customization of bot flows, and integration with backend systems.
More Related Content
Similar to Implementing Your Own Chatbot Platform!
![UiPath Automation Suite Installation (Hands-On) [2/3]](https://cdn.slidesharecdn.com/ss_thumbnails/automationsuitecommunitysession2-251015095633-a6d862f1-thumbnail.jpg?width=600ounds&width=560&fit=bounds)
Implementing Your Own Chatbot Platform!
- 1. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. Lyudmil Pelov @lpelov Oracle A-Team March, 2017 IMPLEMENTING CHATBOT PLATFORMS
- 2. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. | • We Message! – there are more people on messaging apps than social networks – millennials are texting more than calling – teens would rather message than have a phone conversation – 65% of consumers want to message a business to* – > 50% of all consumers would make a purchase via a message – messaging is just easier, faster, and more productive for targeted tasks than mobile or web *Based on survey conducted by Smooch.ai
- 3. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. 3 • most common types of bots • highly specialized and structured interactions • typically user initiated • integrated with enterprise systems • useful in the support or service category Task Oriented Data Driven & Predictive • personalized based on profile, past behavior and contextual information • monitors data and initiates conversation when data crosses threshold • predictive Intelligence based on data - key for bots to be a useful in the long run across use-case categories First Generation Chatbots
- 4. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. Next Generation Chatbots “non supervised general availability bots” “sophisticated prediction models requiring reduced supervising” “auto labeling” “anomaly detection” “concept extraction” “augmentation argument (capable to argument based on facts)” “semi-supervised named entity recognition”
- 5. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. Future Generation Chatbots “be able to take successful the Turing test?!” “Turing Test: If the evaluator cannot reliably tell the machine from the human (Turing originally suggested that the machine would convince a human 30% of the time after five minutes of conversation) , the machine is said to have passed the test.” “The test does not check the ability to give correct answers to questions, only how closely answers resemble those a human would give.” Probably not likely to happen soon but not impossible as the ML models get more advance! https://en.wikipedia.org/wiki/Turing_test
- 6. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. FIRST GENERATION CHATBOT CAPABILITIES AND CHALLENGES
- 7. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. Intent Recognition (classification) “the ability to understand the intent from human ambiguous language” #wordvector, #deeplearning4j, #opennlp, #stanfordnlp
- 8. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. Advance Intent Recognition Challenge “The city councilmen refused the demonstrators a permit because they feared violence” “What does the word “they” refer to here — the councilmen or the demonstrators? What if instead of “feared” we wrote “advocated”?”(1) “Oren Etzioni: When AI can’t determine what “it” refers to in a sentence, it’s hard to believe that it will take over the world”(1) https://techcrunch.com/2017/02/25/conversational-ai-and-the-road-ahead/ (1)
- 9. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. Intent Recognition Training Model Guideline “intents should be different enough to help the ML disambiguate better” “Good Pizza vs Bad Pizza are not two different intents, the intent here would be Pizza Quality, Good and Bad would be entity Rating” “single keywords are not useful” “unique sentences are super helpful, semantically and synonym rich” “do {word} sentence construction in different intent models will falsely give high prediction score for all intents”
- 10. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. Entity Recognition (NER) “the ability to automatically recognize entity meanings from ambiguous language” “despite all improvements in ML, the problem of accurate entity recognition is far from been solved” “main effort currently is directed to find approaches to reduce the annotation labor – semi-supervised NER” #opennlp, #stanfordnlp
- 11. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. Entity Recognition (NER) – example with OpenNLP Model “The other day at work I saw <START:person> Larry Elison <END> at <START:building> HQ600 <END>” Test “The other day at work I saw Lyudmil Pelov at HQ300” “The other day at work I saw lyudmil Pelov at 300”
- 12. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. Sentiment Analysis (emotional intelligence) “evaluate subjective patterns and expressions of sentiment within textual content” “usually identifies if expression was positive, negative, mixed, or neutral” “requires a high degree of supervised ML on big data corpus, usually Twitter” #deeplearning4j, #opennlp, #stanfordnlp
- 13. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. Probabilistic Parser “probabilistic parsing is using dynamic programming algorithms to compute the most likely parse(s) of a given sentence, given a statistical model of the syntactic structure of a language” “parse any address, name, number, date or any unstructured text in useful components” #stanford, #duckling, #parserator http://nlp.stanford.edu/projects/stat-parsing.shtml
- 14. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. Probabilistic Parser – Duckling Example #stanford, #duckling, #parserator https://duckling.wit.ai/
- 15. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. CHATBOT CONCEPTS
- 16. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. https://www.behance.net/gallery/47419375/Chatbot-Powerful-UX-Concepts
- 17. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. ORACLE INTELLIGENT BOT - KEY COMPONENTS
- 18. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. | • active machine learning • NER • sentiment recognition • SDK to consume back-end API’s • declarative bot flow • state machine context driven workflows • cross channel chat interface Channel Configurator Dialog Flow Execution NLP Engine Custom Components in Dialog Flow
- 19. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. ORACLE INTELLIGENT BOT – HOW IT WORKS
- 20. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. Design-time “specify intents and entities” “test intents and flows” “create a flow (depicts bot execution logic)” “wire up the flow to clients and custom components (channels & back-end API integration)”
- 21. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. Run-time “incoming message from channel activates bot flow” “intent classification: determines what the user request is most likely trying to accomplish based on machine learning” “entity resolution: the entity recognizer identifies the important words and phrases contained in each user request” “dialog engine execution: once the entity’s are resolved, the dialog engine or manager formulates the response to return to the user based on a state machine”
- 22. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. ORACLE IBCS: PaaS Multi-Tenant Architecture
- 23. Copyright © 2016,Oracle and/or its affiliates. All rights reserved. | 23 Custom Components Customer MCS/ Customer ACCS/ Customer JCS LBaaS/ Cloudgate ACCS Container C Dialog engine OEHPCS Service Intent Server ACCS Container D ACCS Container A Channel Drivers/Auth UI Endpoints API ACCS Container B DBaaS (1 per tenant)
Editor's Notes
- #3 Bots are transforming the way mobile users interact with businesses and LoB Systems by putting app features in conversational UI’s and eliminating ‘App fatigue’
- #9 What does the word “they” refer to here — the councilmen or the demonstrators? What if instead of “feared” we wrote “advocated?” This changes what we understand by the word “they.” Why? It is clear to us that councilmen are more likely to fear violence, whereas demonstrators are more likely to advocate it. This information, which is vital for disambiguating the pronoun “they,” is not in the text itself, which makes these problems extremely difficult for AI programs. The first ever Winograd Schema Challenge was held last July, and the winning algorithm achieved a score on the challenge that was “a bit better than random.”