A Method to Success of “Oshigatari” Recommendation Talk by Asking to Create Search Queries While Listening

A Method to Success of “Oshigatari” Recommendation Talk by Asking to Create Search Queries While Listening Meiji University Yukina Funazaki, Satoshi Nakamura IS29

Table of Contents ⚫ Contribution ⚫ What is Oshigatari? ⚫ Background ⚫ Survey on the Oshigatari ⚫ Our Proposed Method ⚫ Experiment ⚫ Result ⚫ Discussion and Future Work ⚫ Summary Yukina Funazaki’s paper kes2022

Contribution Yukina Funazaki’s paper kes2022 ⚫ We conducted a basic survey on Oshigatari recommendation talk and found that over 80% of people have unsuccessful experiences of Oshigatari . ⚫ We proposed a method that increases the success of Oshigatari by asking listeners to think about search queries while listening to recommendation talk.

What is Oshigatari ? Yukina Funazaki’s paper kes2022 Recommending your favorite things (actors, music, movies, artists, hobbies, etc.) to others in conversation.

Background Les Misérables musical Jean Valjean Fantine Marius Pontmercy Inspecteur Javert Éponine Cosette

Background Les Misérables musical I want to recommend what I like! You should see the musical “Les Misérables”, because it's fantastic!

Background Les Misérables musical I want to recommend what I like! You should see the musical “Les Misérables”, because it's fantastic! Not really interested.

Survey on the Oshigatari People We conducted an online survey of 500 men and 500 women to find out how many of them had similar experiences to mine. Examples of "favorite things and people” as answered by survey participants. ➠ Things Activity Actor Comic Cooking Idol TV animation Singing Character TV programming Shogi (Japanese chess) Voice Actor YouTube Traveling YouTuber Movies Make something by hand Pianist Cosmetics Drawing a picture Artist Games Collecting points Baseball Team Music Yoga Fashion Sports Books Visiting cafes Cars Building a computer Railroad Fishing Motorcycle Touring Plastic Model Blogging Musical Instruments Mountain climbing Opera Gardening DIY Investment

Survey on the Oshigatari People We conducted an online survey of Things Activity Actor Comic Cooking Idol TV animation Singing Character TV programming Shogi (Japanese chess) Voice Actor YouTube Traveling YouTuber Movies Make something by hand Pianist Cosmetics Drawing a picture Artist Games Collecting points Baseball Team Music Yoga Fashion Sports Books Visiting cafes Cars Building a computer Railroad Fishing Motorcycle Touring It is clear that they have Plastic Model Blogging Musical Instruments Mountain climbing many different types of "favorite things". Opera Gardening 500 men and 500 women to find out how many of them had similar experiences to mine. Examples of "favorite things and people” as answered by survey participants. ➠ DIY Investment

Survey on the Oshigatari Good at talking Not good at talking 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% People who recommendations have never been unsuccessful. People who recommendations have been unsuccessful. 80% 80% of of the the respondents respondents have have experienced experienced aa failed failed Oshigatari Oshigatari recommendation recommendation talk. talk.

Survey on the Oshigatari Why their recommendation talk was not successful. I could not convey the merits of what I liked The other person got bored with the recommendation talk It is problematic if the user’s way of talking determines whether something is interesting or not, when we originally want the user to know how interesting the recommended thing itself is.

Survey on the Oshigatari Why their recommendation talk was not successful. I could not convey the merits of what I liked The other person got bored with the recommendation talk It is problematic if the user’s way of talking determines whether something is interesting or not, when we originally want the user to know how interesting the recommended thing itself is. ➠ We aim to successfully realize person-to-person recommendation talk ( Oshigatari talk) regardless of the user's speaking style.

Approach “Do You Hear The People Sing?” ⚫ Effectively generate interest by inducing active behavior [Setosaki et al. 2011] ⚫ The action of search creates the illusion that the human brain is knowledgeable about the content [Fisher et al. 2015] ➠ It is inferred that search is an effective means of arousing interest in content. AISAS (Consumer behavior model) Users who are interested in a product or service are likely to search for it afterwards ➠ There may be a relationship between interest and search behavior

Approach “Do You Hear The People Sing?” “A story” I enjoyed the movie "A" so much “A cast” that I want you to see it too! “A release date”… This movie is… Uh-huh. Uh-huh. Speaker Listener Proposed Method We want to reduce the listener's effort when listening to the Oshigatari General recommendation talk. ➠ we assumed that the only search process is “the user imagining some search queries,” because we believe that imagining some queries is less burdensome for the listener and increases interest.

Experiment “Do You Hear The People Sing?” “A story” I enjoyed the movie "A" so much “A cast” that I want you to see it too! “A release date”… This movie is… Uh-huh. Uh-huh. Speaker Listener Proposed Method We actually conducted an experiment to see if our suggestion method was effective General in increasing their interest. ➠ We compared how interest changed between listening to a general recommendation talk and listening to a recommendation talk while thinking about search queries.

Experiment “Do You Hear The People Sing?” “A story” I enjoyed the movie "A" so much “A cast” that I want you to see it too! “A release date”… This movie is… Uh-huh. Uh-huh. Speaker Listener Proposed Method We actually conducted an experiment to see if our suggestion method was effective General in increasing their interest. ➠ We compared how interest changed between listening to a general recommendation talk and listening to a recommendation talk while thinking about search queries.

Experiment “Do You Hear The People Sing?” “A story” I enjoyed the movie "A" so much “A cast” that I want you to see it too! “A release date”… This movie is… Uh-huh. Uh-huh. Speaker Listener Proposed Method We actually conducted an experiment to see if our suggestion method was effective General in increasing their interest. ➠ We compared how interest changed between listening to a general recommendation talk and listening to a recommendation talk while thinking about search queries.

Experiment “Do You Hear The People Sing?” “A story” I enjoyed the movie "A" so much “A cast” that I want you to see it too! “A release date”… This movie is… Uh-huh. Uh-huh. Speaker Listener Proposed Method We actually conducted an experiment to see if our suggestion method was effective General in increasing their interest. ➠ We compared how interest changed between listening to a general recommendation talk and listening to a recommendation talk while thinking about search queries.

Experiment Popular Japanese Manga to Read in English Participants in the experiment watched two recommendation videos ⚫ Comic “ The Lady Becomes an Idol ” (Content A) ⚫ Comic “ Oten no Mon ” (Content B) ➠ We wanted to confirm the difference between the proposed and conventional methods, so we decided to watch the video so that the results would not change under the other conditions.

Experiment Popular Japanese Manga to Read in English Participants in the experiment watched two recommendation videos ⚫ Comic “ The Lady Becomes an Idol ” (Condition A) ⚫ Comic “ Oten no Mon ” (Condition B) ➠ We wanted to confirm the difference between the proposed and conventional methods, so we decided to watch the video so that the results would not change under the other conditions. To prevent their interest from changing based on their impressions, we created two recommendation videos using avatars.

Experiment Without Query Thinking Group Popular Japanese Manga to Read in English With Query Thinking Group (Proposed Method) “Les Mis e ra ble s Japan Theater” “Popular Japanese Manga English” ⚫ Participants in the experiment were randomly divided into two groups. ⚫ Participants in the group with query thinking were instructed to listen to the recommendation talk while thinking of search terms.

Experiment Popular Japanese Manga to Read in English ⚫ Knowledge of recommended comic before watching the video Before viewing the recommendation video ⇩ Watch the recommendation video ⇩ After viewing the recommendation video (How much do you know about it?) ⚫ Interest in the recommended comic before watching the video (How interested are you in the content?) ➥ Responses on a 6-point Likert scale (1Not interested at all to 6Very interested)

Experiment Before viewing the recommendation video ⇩ Watch the recommendation video ⇩ After viewing the recommendation video Popular Japanese Manga to Read in English

Experiment Before viewing the recommendation video ⇩ Watch the recommendation video ⇩ After viewing the recommendation video Popular Japanese Manga to Read in English ⚫ Were you interested in what was recommended? ⚫ Did you want to investigate further on your own? ➥ Responses on a 6-point Likert scale (1Not interested at all to 6Very interested)

Result After 5.5 6 5.5 Interest rating value for recommended content 6 Interest rating value for recommended content Popular Japanese Manga to Read in English 5 Before 4.5 4 3.5 3 2.5 2 5 4.5 4 3.5 3 2.5 2 1.5 1.5 1 1 Without query thinking Before viewing the recommendation video With query thinking After viewing the recommendation video Co mic “ The Lady Becomes an Idol ” ( Condition A ) Without query thinking With query thinking Before viewing the recommendation video After viewing the recommendation video Co mic “ Oten no Mon ” ( Condition B) Responses o n a 6 -point Likert scale(1: Not interested at all to 6: Very interested)

Result Popular Japanese Manga to Read in English Interest rating value for recommended content 6 5.5 5 4.5 4 3.5 3 2.5 2 1.5 1 Without query thinking Before viewing the recommendation video With query thinking After viewing the recommendation video Co mic “ The Lady Becomes an Idol ” ( Condition A ) Co mic “ Oten no Mon ” ( Condition B) Responses o n a 6 -point Likert scale(1: Not interested at all to 6: Very interested)

Result Popular Japanese Manga to Read in English 6 Interest rating value for recommended content 5.5 5 4.5 4 3.5 3 2.5 2 1.5 1 Co mic “ The Lady Becomes an Idol ” ( Condition A ) Without query thinking With query thinking Before viewing the recommendation video After viewing the recommendation video Co mic “ Oten no Mon ” ( Condition B) Responses o n a 6 -point Likert scale(1: Not interested at all to 6: Very interested)

Result Popular Japanese Manga to Read in English Examples of queries considered by participants ⚫ “<comic title> and <character name>” ⚫ “<comic title> and <scene type>” ⚫ “<comic title> and <historical fact>” Example ⚫ “Oten no Mon Sugawarano Michizane” ⚫ “Oten no Mon Historical fact ⚫ “Oten no Mon Volume1 Cover Painting” Many respondents looked up the comic’s title and the names of the characters, but other queries varied from person to person.

Result Popular Japanese Manga to Read in English 6 5.5 Rating value of active motivation 5 4.5 4 3.5 3 2.5 2 1.5 1 Without query thinking With query thinking Responses o n a 6 -point Likert scale(1: Not interested at all to 6: Very interested)

Discussion and Future Work Yukina Funazaki kes2022 paper ⚫ These results indicate that listening to the recommendation talk while thinking about the search query of the suggestion method may be effective in generating interest in what is being recommended. ⚫ The query-thinking group showed higher interest ratings and active motivation after viewing the video than the Without-query-thinking group. ➠ It was suggested that having the participants think about the query while listening to the recommendation may increase their interest after the recommendation.

Discussion and Future Work Yukina Funazaki kes2022 paper ⚫ We would like to further increase the number of participants in the experiment to verify whether the suggested method is effective. ⚫ In the experiment, we told the participants to listen to the recommendation talk while thinking about the query. ⚫ It is not common to ask listeners to listen while thinking of a query during a normal recommendation talk. ➠ In the future, we would like to consider ways to induce listeners to spontaneously think about search queries.

Discussion and Future Work Yukina Funazaki kes2022 paper We are trying to build a system to be used during the Oshigatari . ⚫ We believe that showing a third party, the system, entering a query and searching for the query will make it easier for the listener receiving the recommendation to recall the search query. Example of use with zoom

Summary Background I want to recommend something I like. Issue Recommendation not successful. Proposed Method Increase interest by having the listener listen to the recommendation talk while thinking about the search query. Experiment and Results Suggested that the proposed method may be effective in increasing interest in what was recommended. Future work Further verification will be conducted by increasing the number of people in the experiment.