TWIN - Research Project on Recommender Systems with Personality-based Profiles.
  • An important part of machine learning and data mining tasks is the formation of an appropriate dataset;
  • The Internet provides sufficient amount of publicly available data and crawlers can be used to collect and organize it.
datasets and crawlers
  • Datasets allow training algorithms to model interactions between users who purchase goods and services online;
  • People need good advice on what to purchase when the amount of alternatives is extremely high;
  • Single user profile can include information about preferences and personality;
  • Services such as Recommender Systems can utilize such profiles to suggest the most suitable items.
RecSys and personality research
  • We study the quality and trustworthiness of recommendations that people can get from a circle of like-minded individuals;
  • We are interested in the following domains: e-tourism, linguistics and e-commerce.
TripAdvisor datasets

Tripadvisor dataset

(2015)
Includes personality scores
(calculated using Fabio Celli's component).
DOI:­ 10.13140/RG.2.1.5104.8081

- detailed description of users' profiles
- personality scores per each user profile
- samples of 5 or more text reviews (for each user)
- textual content of 1 article (available only for some users)

Dataset description
Download

Tripadvisor dataset

(2015)
Updated version of TripAdvisor Datasets released in 2012-2013.
DOI:­ 10.13140/RG.2.1.3308.0409

- detailed description of users' profiles
- samples of 5 or more text reviews (for each user)
- textual content of 1 article (available only for some users)

Dataset description
Download

TripAdvisor dataset
date of data collection: 23.05.2013

FIELDS:
  1. title of the review
  2. the overall hotel rating
  3. textual content of the review
  4. hotel value rating
  5. hotel rooms rating
  6. hotel location rating
  7. hotel cleanliness rating
  8. hotel service rating
  9. hotel sleep rating

Download


Translations of papers on Personality Recognition (into Russian)


Translations of interviews (into Russian)

A. Roshchina, J. Cardiff and P. Rosso. (2015). TWIN: Personality-based Intelligent Recommender System, Journal of Intelligent & Fuzzy Systems, IOS Press, vol. 28, no. 5, pp. 2059–2071, DOI: 10.3233/IFS-141484.

Рощина А.С. (2017). Песни в одиночку: cборник рассказов / худож. Е. Канаева. — Санкт-Петербург: Галарт+, 2017. — 104 c. ISBN 978-5-90-609010-2

Эти рассказы, в чем-то автобиографичные, в чем-то продиктованные фантазией, стали побочным продуктом во время работы над моим основным проектом. Слова в них — просто указатели на те эмоциональные переживания и образы, которые сопровождали напряженную работу ума на протяжении нескольких лет.
Я хочу пожелать вам захватывающего путешествия по волнам эмоций. Не ищите здесь никакого скрытого смысла и не выстраивайте логических цепочек. Смысл придет сам. Дайте ему время.

Download

  • Группа в VK
  • Группа в FB
  • Группа в Instagram
Abstract: This paper presents the “Tell me What I Need” (TWIN) Personality-based Intelligent Recommender System, the goal of which is to recommend items chosen by like-minded (or “twin”) people with similar personality types which we estimate from their writings. In order to produce recommendations it applies the results achieved in the personality from the text recognition research field to Personality-based Recommender System user profile modelling. In this way it creates a bridge between the efforts of automatic personality score estimation from plain text and the field of Intelligent Recommender Systems. The paper describes the TWIN system architecture, and results of the experimentation with the system in the online travelling domain in order to investigate the possibility of providing valuable recommendations of hotels of the TripAdvisor website for “like-minded people”. The results compare favourably with related experiments, although they demonstrate the complexity of this challenging task.
Abstract: one of the important issues arising in the modern world is the information overload problem. In order to help the person navigate through the sea of all possible choices available online, Recommender Systems have started to appear. They collect preferences of people based on explicit ratings of various products or on the analysis of behaviours of the users working within the system. Therefore Recommender Systems become able to suggest new items to their users taking into account things liked by people with similar tastes.
The process of preferences retrieval and the choice of the recommendation algorithm are key parts of the Recommender System construction. There are a number of classical approaches available: content-based, collaborative filtering, etc. with the tendency to choose the combination of them to create a hybrid system. But recently a new type of Recommender System has appeared that utilises personality information about the users. It provides a more personalised approach to user representation aimed at improving the quality of the recommendations.
In this thesis we propose the TWIN Personality-based Recommender System. In order to produce recommendations it applies the results achieved in the personality from the text recognition research field to Personality-based Recommender System user profile modelling. In this way it creates a bridge between the efforts of automatic personality score estimation from plain text and the field of Recommender Systems. TWIN also serves as a tool for visualizing the resulting scores to perform personality analysis. We show that the application of the TWIN in online tourism domain produces valuable results in recommending tourist facilities to "like-minded" people. We describe the components of the TWIN system, the experiments conducted on the system, and we present an analysis of the very promising results obtained.

Download
Abstract: with the constant increase in the amount of information available in online communities, the task of building an appropriate Recommender System to support the user in her decision making process is becoming more and more challenging. In addition to the classical collaborative filtering and content based approaches, taking into account ratings, preferences and demographic characteristics of the users, a new type of Recommender System, based on personality parameters, has been emerging recently. In this paper we describe the TWIN (Tell Me What I Need) Personality Based Recommender System, and report on our experiments and experiences of utilizing techniques which allow the extraction of the personality type from text (following the Big Five model popular in the psychological research). We estimate the possibility of constructing the personality-based Recommender System that does not require users to fill in personality questionnaires. We are applying the proposed system in the online travelling domain to perform TripAdvisor hotels recommendation by analysing the text of user generated reviews, which are freely accessible from the community website.

Download
Abstract: the information overload experienced by people who use online services and read user-generated content (e.g. product reviews and ratings) to make their decisions has led to the development of the so-called recommender systems. We address the problem of the large increase in the user-generated reviews, which are added to each day and consequently make it difficult for the user to obtain a clear picture of the quality of the facility in which they are interested. In this paper, we describe the TWIN (“Tell me What I Need”) personality-based recommender system, the aim of which is to select for the user reviews which have been written by like-minded individuals. We focus in particular on the task of User Profile construction. We apply the system in the travelling domain, to suggest hotels from the TripAdvisor site by filtering out reviews produced by people with similar, or like-minded views, to those of the user. In order to establish the similarity between people we construct a user profile by modelling the user’s personality (according to the Big Five model) based on linguistic cues collected from the user-generated text.

Download
Abstract: the appearance of the so-called recommender systems has led to the possibility of reducing the information overload experienced by individuals searching among online resources. One of the areas of application of recommender systems is the online tourism domain where sites like TripAdvisor allow people to post reviews of various hotels to help others make a good choice when planning their trip. As the number of such reviews grows in size every day, clearly it is impractical for the individual to go through all of them. We propose the TWIN (“Tell me What I Need”) Personality-based Recommender System that analyzes the textual content of the reviews and estimates the personality of the user according to the Big Five model to suggest the reviews written by “twin-minded” people. In this paper we compare a number of algorithms to select the better option for personality estimation in the task of user profile construction.

Download
Abstract: the construction process of the traditional heavyweight ontology precisely describing a specific area is a time consuming task. But in the field of constantly changing dynamic areas like the Web it is impossible to produce a complete ontology to accurately reflect any particular domain of interest. If information in the domain is changing rapidly the corresponding ontology should be constantly enriched with newly emerged concepts and relations. The problem of ontology enrichment becomes even more crucial now with the emergence of social networking services and e-learning domains with highly dynamic content. In the paper we propose an experiment aimed at constructing ontology of interests based on the data provided by the Delicious online social service. This ontology will then be used as raw material for our main goal of addressing the challenge of improving or enriching the ontological structure by developing techniques and mechanisms for capturing and representing the "hidden information" in ontology.

Download
Abstract: one of the tasks of the computational linguistics is a detection of the structure (more general – a grammar of the language) of texts from some corpus to get an opportunity to generate new texts not seen by the analyzer. In this work we distil the structure of the given text using the description of the ADIOS algorithm (Automatic Distillation of Structure) and test it on russian texts. The steps include dividing the given text into minimal morphological parts (MMP) artificially constructed from words using the list of possible russian morphs or into words; constructing the graph of constituents; processing of significant patterns and detecting equivalence classes; visualizing the distilled hierarchical structure.

Download
Abstract: объектом исследования являются тексты, разработанные в рамках проекта «Национальный корпус русского языка».
Цель работы – создание программы-конкордансера для построения конкорданса текстов указанного выше проекта. В процессе работы строились веб-интерфейсы для взаимодействия с пользователем и шаблоны на языке регулярных выражений для нахождения совпадений в текстах корпуса с запрошенной пользователем словоформой (веб-интерфейс).

Download
Abstract: объектом исследования являются тексты, входящие в состав «Национального корпуса русского языка».
Цель работы: создание программы сегментации слитного текста с удаленными пробелами и знаками препинания, реализующей алгоритм когорты и расширенный алгоритм когорты, использующий частотные данные. В процессе работы создавались интерфейсы взаимодействия с пользователем и обрабатывались данные о частотности встречаемости последовательностей букв в начале и в конце слов.

Download


TWIN PERSONA © 2011-2017
Created by Lory Rosh