We present the initial design and development of KidsQR, a query recommendation system tailored exclusively for children. KidsQR aids children in their quest for online information by considering children vocabulary, child-friendly phrases, and entities children are familiar with. Initial experiments conducted based on the assessment of parents and elementary school teacher appraisers verify the promising performance of KidsQR.
Despite the large number of studies conducted in the field of
query recommendation, relatively few focus explicitly on the
young group of Internet users and their difficulties,.
Consequently, literature pertaining query recommendation for
children is very limited. In fact, most of the existing query
recommendation systems are designed based on the information
needs of adults [
In this section we present a brief overview of KidsQR. Generating Candidates. To identify possible queries to be recommended, i.e., candidate queries, in response to a given user’s initial query, KidsQR employs Ubersuggest.org1. Ubersuggest is a query generation tool that provides hundreds of possible suggestions given an initial user query and offers topical diversity among the suggestions. We have verified that phrases provided by Ubersuggest include terms related to children pop culture, such as the name of cartoon characters.
Analyzing Candidates. To determine the adequacy of the candidates being recommended to the user, i.e. distinguishing child-friendly candidates from the non-child-friendly ones, KidsQR evaluates each candidate query based on a number of child-related characteristics that are applicable to them. In other words, KidsQR considers child-related properties to determine how closely a candidate phrase is related to children’s interests, or if the candidate relates to child-friendly content. The properties/characteristics to be observed in quantifying the degree to which a candidate query is likely reflecting children’s search intent are described as follows:
Vocabulary. A fundamental step to differentiate childfriendly queries among the candidate ones, is to examine the existence of children’s vocabulary terms in each query. We consider children vocabulary lists extracted from children dictionaries and schools’ academic vocabulary (such as www.opsu.edu/www/education/BuildAcademicVoc.pdf and 1 While we used Ubersuggest for development purposes, other tools, such as keywordtool.io, can be considered as well. kids.wordsmyth.net/we/) and prioritize candidates that include keywords frequently occurring in children pre-defined vocabularies. We do so, since it is anticipated that children will favor queries including keywords they are familiar with. For example, for the queries “color” and “city,” the candidates “coloring pages” and “pig in a city” are preferred over “color spectrum” and “city infrastructure” since “spectrum” and “infrastructure” are not common words among children. x Popularity. The popularity of terms among children is considered by analyzing term frequency distributions2 on children stories, poems, and blog posts. Candidate queries including popular children terms are also given precedence. x Phrase-Formulating. Examining the child-friendliness of individual terms in candidate queries is crucial, but not sufficient in confirming the appropriateness of a candidate query since it does not consider the query phrase as a whole. For example, having the words “bar” —as in “chocolate bar”—and “open” in children vocabulary does not imply that “open bar” is a child-related phrase. We consider stories and poems written for children, as well as texts, blog posts, and online reviews written by children, to determine the appropriateness of the combination of the words, and capture children’s informal phrasing patterns. Candidate queries that have similar patterns to children’s informal phrasing behavior, or are child-appropriate as a phrase, most likely address a child’s search intention, hence, are prioritized. x Pop-Culture. We observed that candidate queries that do not include children vocabulary, or do not literally make sense as a phrase, can still be related to children’s popular culture. KisdQR examines candidate queries in the context of children pop-culture and prioritizes queries including terms related to children’s movies, songs, and toys (extracted from Pixar.com and Allmovie.com, to name a few). For example, “Mary Poppins” and “Mr. Potato Head” are valid candidates since they refer to a movie character and a toy, respectively, even though the former contains “Poppins”, a word not included in children’s vocabulary, and the latter consists of child-related words but does not have a literal meaning as a phrase. Ranking. KidsQR analyzes each of the candidate queries based on the characteristics mentioned above, and prioritizes candidates that (i) are simple, (ii) refer to children’s topics of interests, (iii) include terms children are familiar with, and (iv) resemble children’s informal phrasing behavior. KidsQR relies on a multiple regression analysis model that simultaneously considers the different contributing factors in determining whether a candidate query is, in fact, child-friendly and generates a single ranking score for each candidate query recommendation. The topN candidates are presented to the user as the corresponding query recommendations that can help capture his search intent and guide the online search process.
As far as we know, a benchmark dataset that specifically
addresses queries conducted by children has yet to be developed.
Thus, we created our own dataset by conducting a user study and
collecting data from 10 appraisers who were either parents of
children between the ages of 3 and 12, or elementary school
teachers. We presented each appraiser with 8 queries and the
corresponding set of query recommendations, comprised of
2 Sample sources considered for determining term popularity and
phrase suitability include kidsblogclub.com and storybud.org.
randomly-positioned recommendations generated by Google,
Bing, and KidsQR. Appraisers were then asked to select the two
recommendations that they found most child-friendly for each
query and their selections were treated as the gold standard.
Using the created dataset we evaluated KidsQR based on Mean
Reciprocal Rank (MRR) and Normalized Discounted Cumulative
Gain (NDCG). We also compared the performance of KidsQR
with that of Google and Bing, two well-known search engines that
offer query recommendations and that are frequently used by
children [
We have developed a query recommendation system, KidsQR, designed specifically to address the challenges of children in query formulation. KidsQR distinguishes the child-friendly query candidates from the non-child-friendly ones by simultaneously considering multiple desired properties on children queries. We aim to further enhance the initial development of KidsQR so that it can adequately handle informal as well as natural language phrasing which are very common among children. We also intent to further enhance the performance of KidsQR by addressing children pop-culture more comprehensively. We believe the more aspects of children’s pop-culture that we consider, the more closely we can predict a child user’s search intention, i.e. recommend queries that are anticipated to be appealing from a child’s perspective can be generated. Moreover, we will examine children vocabulary and words provided by school vocabulary lists more accurately and consider the age gap among young children, i.e., we will group children by age groups and explicitly consider their reading ability in making query recommendations for children in the respective groups.