Providing effective recommendations in discussion groups using a new hybrid recommender system based on implicit ratings and semantic similarity

Highlights

• A new recommender system is represented which has three parts.

• Content-based, collaborative, and hybrid filtering are three sections of the proposed system.

• This work is developed on discussion groups with tagging feature.

• Semantic relevancies of tags are extracted using WordNet database.

• The tags are organized in a hierarchical structure based on their semantic relevance.

Abstract

Discussion groups are one of the most important elements of collaborative learning which utilize recommender systems to improve their performance in several aspects. This type of learning facilitates a comfort communication between users to share their problems and questions and receive the appropriate solutions. Most of recommender systems of discussion groups are based on using collaborative filtering techniques and a few numbers of them use content-based or hybrid filtering. Experimental results of previous works show that using hybrid recommender systems on discussion groups’ databases cause significant improvement in accuracy of recommended posts in comparison with other filtering techniques (Kardan and Ebrahimi, 2013). To improve performance of (Kardan and Ebrahimi, 2013), in this paper, a new recommender system is represented, which includes three parts, namely content-based, collaborative, and hybrid filtering parts. The proposed recommender system uses the tagging features to provide more appropriate recommendations on discussion groups. For this purpose, semantic relevance of tags is extracted using WordNet lexical database and the tags are organized in a hierarchical structure based on their semantic relevance. The hierarchical structure is used for searching relevant posts in content-based filtering part, and the user’s query is extended using related semantic tags. The implicit ratings of the users are calculated in the collaborative filtering part using similarity measures. Finally, the results of these two parts are combined in the hybrid filtering part of the proposed system to recommend the posts of the discussion group which are similar to the query of the active user. Experimental results show higher precision of the proposed system comparing to the former recommender systems.

Introduction

Various communicative environments in different domains have been provided for users via the Internet platform using different web technologies. Asynchronous discussion groups are important examples of these communicative environments which allow users to find proper answers for their raised questions (Kardan and Ebrahimi, 2013). Knowledge extraction from discussion groups and communities is an important research issue that is considered in a variety of recent studies (Kaššák et al., 2016, Christensen and Schiaffino, 2011, Ortega et al., 2016, Wang et al., 2016, Jhamb and Fang, 2017, Lee and Brusilovsky, 2017, Xu, 2018, Li et al., 2018). Unstructured nature of the posts of the members and large volume of information are the main problems of the knowledge extraction process from discussion groups. Recommender systems can be exploited to extract useful knowledge from discussion groups.

This study aims to deliver the appropriate contents that are posted by members of discussion groups to the inquirer users. For this purpose, the related contents are identified at the first step and then, according to the member's interests, the appropriate recommendation will be provided. Most of recommender systems use the similarity of users to make recommendations in discussion groups. Considering similarities of content, or user and content, for this purpose has not received much attention in these systems. The structure of discussion groups is similar to a tree structure and their contents can be as main groups, subgroups, discussion topics, and posts (Fanaeetork and Yazdi, 2013). Each discussion in the subgroups is called a “thread” which has the date and subject by default. This structure allows users to search related topics more quickly. There are some drawbacks in this structure. For example, the users cannot search whole of the group when one topic is included at several threads and while the size of the group is extra-large and thus, they cannot find their related information. Since users cannot usually explain their interests using keyword-based queries. In these situations, common keyword-based searching of discussion groups cannot works properly. Recommender systems work based on users’ activities, behaviors, and preferences.

Sparsity and cold start problems have made collaborative filtering techniques inappropriate for developing recommendations in discussion groups. The semantic of each post should be discovered in order to increase the quality of recommendations in discussion groups. Therefore, in content-based filtering techniques, in order to extraction of related content, the semantic similarity should be applied instead of other methods which use keyword-based measures. Furthermore, the contents of the posts are ignored by the collaborative filtering recommender systems of discussion groups. A proper solution to overcome the mentioned challenges is to use hybrid techniques which consider information of both user and content. The hybrid recommender system is rarely presented in discussion groups’ domain.

In this study, we are going to resolve the mentioned challenges. For this purpose, a new hybrid recommendation technique for discussion groups is suggested. Since the users’ explicit information is usually unavailable, their implicit information should be used. Some ideas of collaborative filtering part of the proposed technique are based on activity analysis and user’s behavior in discussion groups, which retrieve the user's implicit information. Based on this implicit information, some functions are introduced to calculate implicit ratings of posts. Similar users are identified according to the extracted implicit ratings. The semantic similarity of posts is the basis of suggested plan of the content-based filtering part of the proposed method. For this purpose, the present tags in each subgroup should be identified separately and they should be also organized according to their semantic relationships. Then, during searching for the similar posts, these identified semantic relationships are used and related tags are discovered and similar posts are found based on them. The results of two mentioned parts are combined in hybrid filtering part and the final recommendations are made in the last step.

The main contributions of this paper can be summarized as follows:

• A new hybrid method is proposed to provide recommendation in discussion groups. This method includes 3 main parts, namely, collaborative filtering part, content-based filtering part, and hybrid filtering part.

• In collaborative filtering part of the proposed method, the users’ implicit ratings obtained based on presented ideas and users which are similar to the active user are identified.

• In content-based filtering part, similar posts to the user’s query are identified according to the semantic relevance between user’s question and existing posts.

• Finally, in the hybrid filtering part a relation is provided in which obtained posts are recommended to active user based on what extent similar users to the active user have contributed in them.

The remaining of this paper is organized as follows: In Section 2, related works are explained. Section 3 introduces the proposed method. In Section 4, the experimental results and their related evaluation and analytical discussions are described. Finally, the work is concluded in Section 5.