Chat-based recommender systems are getting more and more attention in recent time given their natural interaction with the user. Indeed, chat-based recommender systems implement a paradigm where users define their preferences and discover items that best fit their needs through a dialog. A chat-based recommender system can be easily integrated in platforms such as social networks, e-commerce websites, bank websites. Therefore, the preferences can be directly provided by the users during the dialog or can be automatically extracted from their activities on the same platform that hosts the chatbot [3]. In this demo, we present a framework for building chatbased recommender systems. The framework, based on a content-based recommendation algorithm, is independent from the domain.
Chat-based recommender systems have the capability of
interacting with the user during the recommendation process
[
Users can provide functional requirements or technical
constraints used by the recommender for finding the items that
best fit their needs. Accordingly, the acquisition of
preferences is an incremental process that might not be necessarily
ifnalized in a single step. A chat-based recommender system
can provide several interaction modes and can ofer
explanation mechanisms [
In this demo we will show a framework, not dependent on the domain, for generating conversational recommender systems. Our framework implements most of the capabilities that Copyright © CIKM 2018 for the individual papers by the papers' authors. Copyright © CIKM 2018 for the volume as a collection by its editors. This volume and its papers are published under 2
The main goal of this framework is to make easy the building of a new chat-based recommender system. Therefore, the components have been generalized making them independent from a specific domain. When the user desire to build a new chat-based recommender system for a new domain she should update the configuration file, and provide the list of entities and properties in the Wikidata2 format. This last requirement depends on the implementation of the Entity Recognizer. In the following we provide a brief description of each component.
Dialog Manager. This is the core component of the framework whose responsibility is to supervise the whole recommendation process. The Dialog Manager (DM) is the component that keeps track of the dialog state. DM receives the user message, invokes the components needed for answering to the user request, and returns the message to be shown to the user.
Intent Recognizer. This component has the goal of defining the intent of the user formulated by natural language. The Intent Recognizer (IR) is based on DialogFlow3 developed by Google. Our framework uses the DialogFlow APIs for sending the user message and to receive the intent recognized. DialogFlow requires a set of example sentences for each intent.
Sentiment Analyzer. The Sentiment Analyzer (SA) is based on the Sentiment Tagger of Stanford CoreNLP4. The Sentiment Tagger takes as input the user sentence and returns the sentiment tags identified. Afterwards, SA assigns 1On Telegram, search for: @MovieRecSysBot, @MusicRecSys, @BookRecSys. 2https://www.wikidata.org/ 3https://dialogflow.com/ 4https://stanfordnlp.github.io/CoreNLP/ the sentiment tags to the right entity identified into the sentence. For example, given the sentence I like The Matrix, but I hate Keanu Reeves, the Sentiment Tagger identifies a positive sentiment (i.e. like) and a negative one (i.e. hate). SA associates the positive sentiment to the entity The Matrix and the negative sentiment to the entity Keanu Reeves.
Entity Recognizer. The aim of the Entity Recognizer (ER) module is to find relevant entities mentioned in the user sentence and then to link them to the correct concept in the Knowledge Base (KB). The KB chosen for building our framework is Wikidata since it is a free and open-knowledge base and acts as a hub of several structured data coming from Wikimedia sister projects5. Moreover, Wikidata covers several domains and this is a key feature for developing a domainindependent framework. The ER module can be adapted to exploit a custom KB for particular domains not covered by Wikidata. The only requirements are: 1) the knowledge must be modeled through triples; 2) each concept must have one or more alias.
Recommendation Services This component collects
the services strictly related to the recommendation process.
The recommendation algorithm implemented is the
PageRank with Priors [
During the demo we will show three diferent implementations of our chat-based recommender system in three diferent domains. More specifically, we will show a music recommender system, a movie recommender system, and a book recommender system. The peculiarity of these systems is that they have been implemented through the same framework with minimal code variations. Furthermore, we will show three diferent interaction modes: a button-based interaction, a natural-language based interaction, and a hybrid interaction (natural language and buttons combined). In Figure 1 two screenshots of the movie recommender system with the hybrid interaction, and the music recommender system with the NLP interaction are reported. The goal of the demonstration will be to show how to build a new chat-based recommender through our framework is an easy task.
This work has been funded by the projects UNIFIED WEALTH MANAGEMENT PLATFORM - OBJECTWAY SpA - Via Giovanni Da Procida nr. 24, 20149 MILANO - c.f., P. IVA 07114250967, and PON01 00850 ASK-Health (Advanced system for the interpretations and sharing of knowledge in health care).