=Paper=
{{Paper
|id=Vol-1866/paper_197
|storemode=property
|title=An Automatic Procedure for Generating Datasets for Conversational Recommender Systems
|pdfUrl=https://ceur-ws.org/Vol-1866/paper_197.pdf
|volume=Vol-1866
|authors=Alessandro Suglia,Claudio Greco,Pierpaolo Basile,Giovanni Semeraro,Annalina Caputo
|dblpUrl=https://dblp.org/rec/conf/clef/Suglia0BSC17
}}
==An Automatic Procedure for Generating Datasets for Conversational Recommender Systems==
https://ceur-ws.org/Vol-1866/paper_197.pdf
An Automatic Procedure for Generating
Datasets for Conversational Recommender
Systems
Alessandro Suglia1 , Claudio Greco1 , Pierpaolo Basile1 , Giovanni Semeraro1 ,
and Annalina Caputo2
1
Department of Computer Science, University of Bari Aldo Moro
2
ADAPT Centre, Trinity College Dublin, Dublin, Ireland
alessandro.suglia@gmail.com, claudiogaetanogreco@gmail.com,
{firstname.lastname}@uniba.it,annalina.caputo@adaptcentre.ie
Abstract. Conversational Recommender Systems assist online users in
their information-seeking and decision making tasks by supporting an
interactive process with the aim of finding the most appealing items ac-
cording to the user preferences. Unfortunately, collecting dialogues data
to train these systems can be labour-intensive, especially for data-hungry
Deep Learning models. Therefore, we propose an automatic procedure
able to generate plausible dialogues from recommender systems datasets.
People have information needs of varying complexity, which can be solved by
an intelligent agent able to answer questions formulated in a proper way, eventu-
ally considering user context and preferences. Conversational Recommender Sys-
tems (CRS) assist online users in their information-seeking and decision making
tasks by supporting an interactive process [1] with the aim of finding the most
appealing items according to the user preferences.
Unfortunately, collecting dialogues data required for the training phase of
these systems can be really labour-intensive, especially for the latest data-hungry
Deep Learning models. For this reason, synthetic dialogue datasets can be ex-
tremely useful in order to bootstrap effective dialogue systems able to support
a goal-oriented conversation with the user. Therefore, we propose an automatic
procedure able to generate plausible dialogues directly from well-known recom-
mender systems datasets exploiting data coming from the Linked Open Data
Cloud and contextual information related to the user.
Given a user u and his/her set of binary preferences, we trained a decision tree
from the user u preferences expressed towards items represented using Linked
Open Data binary features extracted from the Wikidata 3 knowledge base. In par-
ticular, each pair predicate-object is represented as a binary feature which is 1 if
The dialogue datasets generated from MovieLens 1M and MovieTweetings datasets
can be found at: http://github.com/swapUniba/ConvRecSysDataset.
The source code of the automatic procedure for generating conversational recom-
mender systems datasets will be released when the paper will be accepted.
3
http://www.wikidata.org
the knowledge base contains the triple (item, predicate, object), 0 otherwise. The
considered predicates are wdt:P57 (director), wdt:P161 (cast member), wdt:P136
(genre) 4 . The dialogue generation procedure is an iterative algorithm which is
executed until all user preferences have been used. At each step of the dialog
generation procedure, a top-n list of items composed by positive and negative
items is generated by randomly choosing from positive and negative preferences
of the given user u. Then, paths from the root of the decision tree to the con-
sistently classified examples are exploited to generate a sequence of questions,
randomly chosen according to a binomial distribution over the item features,
to elicit user preferences. Depending from the percentage of positive items in
the top-n, a “refine” step is triggered which extends the dialog with additional
questions that lead to a list of suggestions which contains only positive items.
Table 1 shows a conversation generated by applying the designed procedure
to the well-known MovieLens 1M recommender systems dataset. In the first
part of the conversation, utterances with the aim of introducing the user are
generated by exploiting the contextual information included in the dataset.
Agent Utterance
user Hey
bot What is your name?
user I am Sofia
bot How old are you?
user I am 25
bot What is your occupation?
user sales/marketing
bot Which are your favourite movies?
user I love Speed
bot I suggest you Erin Brockovich Witness Pocahontas Four Weddings and a Funeral Anna and the King. Do you like them?
user I hate them
bot What are your favourite directors?
user I like John Waters
bot What are your favourite actors?
user I like Josef Sommer Wade Williams Marg Helgenberger Jeroen Krabb
bot What genres do you like?
user My favourite genres are teen film romance film biographical film
bot I suggest you Erin Brockovich Witness Ever After Simply Irresistible Hairspray. Do you like them?
user I like them
bot I am glad that you like them
Table 1. Conversation generated from the MovieLens 1M dataset. To help reading,
Wikidata ids have been replaced with the corrisponding entities.
In this work we have proposed an automatic procedure able to generate syn-
thetic dialogue datasets starting from well-known datasets in the recommender
system field. The presented procedure is completely generic and can be applied
on any dataset containing binary user preferences and whose items have a cor-
responding identifier in the Linked Open Data Cloud.
References
1. Mahmood, T., Ricci, F.: Improving recommender systems with adaptive conversa-
tional strategies. In: Proceedings of the 20th ACM conference on Hypertext and
hypermedia. pp. 73–82. ACM (2009)
4
The prefix wdt stands for http://www.wikidata.org/prop/direct/